DeepSeek Opens the AI Application Layer - Battle Shifts from Best Model to Best Ecosystem.

 What happens when models, search, files, voice and application scaffolding become composable?

There are moments when a product announcement looks tactical but is actually changing the structure of an industry. DeepSeek’s decision to expose the capabilities behind its application through APIs may be one of those moments.

The superficial reading is that DeepSeek has “open-sourced its entire app.” That is not precise. The more consequential interpretation is that DeepSeek has combined hosted model access, application capabilities, familiar API conventions and an open reference frontend so developers can reproduce much of the application experience without rebuilding every foundational component.

This matters because the AI industry is moving beyond a contest over which model ranks first on a benchmark. The next contest concerns who controls the application stack, who owns workflow distribution, how easily customers can switch providers, and which company can turn intelligence into a dependable business outcome.

The argument in one minute

DeepSeek has not made every backend component independently deployable. The reference frontend may be open, while search, documents, OCR, speech and other capabilities can remain provider-operated services. Yet the strategic direction is clear: the application-building layer is becoming a platform rather than a collection of specialist integrations.

Portability is becoming a competitive weapon. Familiar API conventions lower migration friction. Aggressive pricing forces the market to discuss economics. Open-weight options create an alternative path for organisations that need more deployment control. None of this makes DeepSeek universally superior; it makes the architecture more contestable.

The most important consequence is the movement of value. As models, retrieval, files, voice and application scaffolding become reusable, defensible advantage shifts toward proprietary data, workflow ownership, distribution, governance, trust and measurable business outcomes.

What DeepSeek actually changed

DeepSeek has exposed a broad group of capabilities associated with its consumer experience through APIs. The reported set includes web search, deeper search, file input, URL retrieval, safety checks, OCR, speech recognition and text-to-speech. Several capabilities can also be invoked independently, which matters for developers building agents, retrieval systems and multimodal workflows.

The move turns features that might previously have required separate vendors into reusable primitives. A research assistant can combine a model with search, URL retrieval, document input and structured tool use. A customer-service product can add voice, document processing and safety services without separately constructing every foundational capability.

DeepSeek has also released a reference chat frontend. That lowers the cost of beginning a product, but it does not mean the complete production backend is open source or independently deployable. The accurate description is an open reference interface combined with hosted APIs and services.

The third change is portability. Supporting familiar API patterns lowers the work required to test an alternative provider. The strategic message is simple: organisations should not have to rewrite an application merely to evaluate different intelligence or economics.

Three ways to build the same user experience

The market is no longer choosing between only “closed AI” and “open-source AI.” Buyers are selecting among operating models. A managed platform maximises integration convenience but leaves more control with the provider. DeepSeek’s hosted toolkit offers a more composable path while retaining backend dependencies. Independent deployment provides the most control, but transfers security, monitoring, scaling, upgrades and reliability to the adopting organisation.

Three architectural choices can produce a similar user experience while allocating cost, control and operating responsibility very differently.

A user may see a chat interface, upload a document, request web research and receive a spoken response in all three cases. Underneath that surface, however, the allocation of cost, risk and accountability is materially different.

Where DeepSeek sits compared with other offerings

A 2×2 matrix is useful only if it positions offerings rather than entire companies. OpenAI, Google and DeepSeek can each participate in more than one part of the market. A consumer application is usually provider-controlled, while a self-hosted open-weight deployment gives an organisation more independent control. Hosted APIs sit between those extremes because they support business integration without transferring infrastructure operation.

DeepSeek sits close to the boundary between hosted convenience and independently deployable model options.

DeepSeek’s hosted app and hosted APIs remain provider-operated. At the same time, open-weight model options, familiar interfaces and community tooling can support more independently deployable solutions. The proposition is therefore not “everything is open.” It is “customers have more architectural choices.”

That optionality can be commercially powerful. A startup may begin with the hosted API to achieve speed, then move selected workloads into independently hosted models as volume, privacy or economics justify the additional operating effort. An enterprise may retain a managed provider for high-value reasoning while using a deployable model for internal classification, retrieval or summarisation.

The competitive landscape is splitting into different theories of value

OpenAI has expanded from models into a broad consumer product, developer platform, agents, multimodality and enterprise controls. Its theory of advantage combines product breadth, distribution and ecosystem depth.

Anthropic has concentrated its identity around reasoning, coding, dependable tool use and high-value enterprise workflows. The objective is not merely to provide a cheaper response, but to become embedded in consequential work where reliability and human time are more important than headline token rates.

Google has the broadest vertical system. Gemini, Google Cloud, TPUs, Search, Workspace, Android and global consumer distribution sit inside one economic architecture. Google can make one layer cheaper to protect value elsewhere.

DeepSeek is differentiating through aggressive economics, composability, API compatibility and a larger role for deployable or open-weight components. Its test is whether lower switching costs and an accessible application toolkit can make DeepSeek infrastructure a default ingredient inside products built by others.

A qualitative comparison across price competitiveness, usability, enterprise maturity, portability and infrastructure integration.

DeepSeek’s strengths are price pressure, portability and open or deployable options. Its challenge is enterprise maturity. Identity, permissions, audit, compliance, service assurance, data residency and long-term support can matter more than headline inference rates. OpenAI, Anthropic and Google have stronger enterprise foundations, although customers may pay for that maturity through platform dependence, premium tiers or both.

Pricing and efficiency: the wrong argument and the right one

The simplest comparison asks which provider charges the lowest input and output token rates. That is useful for screening, but it is not a complete economic model. A low-priced model can become expensive if it requires more reasoning tokens, repeated prompts, failed tool calls, external searches or human correction. A premium model may prove economical when it completes a high-value workflow with fewer attempts and less supervision.

Token rates flow through model behaviour, tools, system performance and human effort before becoming an economic outcome.

A defensible benchmark fixes the workload and measures the full path. For customer support, that means resolution rate, escalation rate, latency, tool-call reliability and human review. For coding, it means accepted changes, regression rates and the time developers spend correcting output. For document extraction, it means field-level accuracy, exceptions and reconciliation effort. For research, it means source quality, search cost and analyst verification.

The same principle applies to infrastructure. NVIDIA and AMD cannot be compared from purchase price alone because utilisation, memory, networking, software optimisation and engineering effort determine the real cost of useful tokens. Cheap components are valuable only when the system converts them into reliable output.

Portability is becoming bargaining power

For years, AI providers benefited from integration gravity. Once an application was written around one model’s message format, tool schema, authentication model and safety behaviour, switching created friction. Compatibility weakens that gravity. An enterprise can introduce an orchestration layer that routes workloads among providers, uses a premium model for complex cases and a lower-cost model for routine tasks.

This architecture does not remove complexity. It adds model evaluation, routing policies, observability and the possibility of inconsistent behaviour. But it creates commercial and operational leverage. Procurement can compare providers using real workloads rather than headline benchmarks. Engineering can replace a model without replacing the whole application. Risk teams can maintain a fallback provider for resilience.

DeepSeek’s strategic opportunity is therefore larger than the number of people using its consumer app. The company can become infrastructure inside someone else’s product. That broadens the addressable market and applies pressure to the margins of generic AI applications.

What becomes commoditised, and what becomes more valuable

Search once required specialist infrastructure. OCR was a distinct product category. Speech recognition and synthesis required dedicated vendors. File processing, safety checks and tool orchestration each represented meaningful development work. As these capabilities become standardised APIs, the cost of assembling an AI application falls.

That does not make every application worthless. It changes what an application must own. A generic chat interface with search and documents becomes easier to reproduce. A deeply integrated claims workflow, clinical pathway, engineering system or sales process remains difficult because the value lies in permissions, proprietary knowledge, exception handling, adoption and accountability.

As components become composable, defensible value migrates toward workflow, data, governance, distribution and outcomes.

This is a familiar pattern in technology markets. Commoditisation at one layer expands demand for the layers above it. Cheaper models increase application experimentation. Easier application assembly increases demand for governance and integration. More open weights increase demand for infrastructure, monitoring and support. Value is not destroyed; it is redistributed.

What this means for the major companies

OpenAI

OpenAI is unlikely to be displaced simply because another provider offers cheaper APIs. Its defence is the combined value of ChatGPT distribution, developer APIs, multimodality, agents, integrations and enterprise controls. The risk is margin compression: customers may refuse to pay premium rates for routine workloads when compatible alternatives are available. OpenAI must show that premium models reduce total task cost rather than merely producing higher benchmark scores.

Anthropic

Anthropic’s strongest position is high-value work: coding, reasoning, research and agentic enterprise workflows. Generic chat functionality is vulnerable to commoditisation, but dependable performance inside complex work can remain differentiated. Anthropic’s challenge is to translate model quality into workflow ownership before open toolkits make the surrounding application layer easier to reproduce.

Google

Google has the most complete vertical hedge. It can capture value through TPUs, Google Cloud, Gemini, Search, Workspace, Android and advertising. If model margins decline, Google can still benefit from cloud consumption and distribution. Its challenge is external trust: enterprise buyers need confidence that Google will support third-party workloads as consistently as its own integrated ecosystem.

DeepSeek

DeepSeek can become a significant infrastructure provider if it combines competitive models, aggressive pricing, portable interfaces and a broad toolkit. The opportunity is to become the default backend for products that do not want to rebuild foundational capabilities. The challenges are enterprise assurance, geopolitical risk, licensing clarity, service availability and the difference between accessible APIs and independently controlled infrastructure.

Application start-ups

The pressure is greatest on startups whose differentiation is a thin layer over a model, search, documents or voice. If the underlying platform exposes those features directly, the startup must move deeper into proprietary workflows, regulated expertise, distribution or systems of record. “We added AI” is not a moat. Owning the customer’s process can be.

What enterprise buyers should do now

First, separate the application from the model provider. Use an orchestration layer, internal interface or gateway so the business can test alternatives without redesigning the user experience. Second, establish workload-specific evaluations. A single benchmark cannot tell procurement which model is best for coding, extraction, customer service, research and agents. Third, calculate cost per outcome, including tools, search, retries, latency, human review and operating effort.

Fourth, classify which components must remain under organisational control. Sensitive data, retrieval indexes, policy rules, audit logs and workflow state may require a different architecture from public content generation. Fifth, insist on explicit portability: data export, prompt and tool portability, fallback providers, model-change testing and exit terms. Finally, identify the proprietary layer. If competitors can buy the same model and tools, advantage must come from data, workflow integration, domain knowledge or customer access.

Conclusion

The viral interpretation says DeepSeek has open-sourced its entire application stack and closed models are finished. That conclusion is too dramatic. DeepSeek has not removed the value of managed infrastructure, frontier model quality, enterprise controls or distribution. It has, however, lowered the cost of reproducing capabilities that once made a sophisticated AI application distinctive.

DeepSeek’s biggest disruption may not be making intelligence cheaper. It may be making the AI application easier to reproduce.

The industry is no longer asking only who has the best model or the lowest token price. It is asking who can combine intelligence, tools, portability, governance and distribution into an ecosystem that customers can adopt without becoming trapped.

Open and closed systems will coexist. Many winning products will combine open-weight models, hosted proprietary services and organisation-controlled workflows. The durable moat will not be ideological purity. It will be architectural flexibility paired with trust, operating excellence and ownership of a valuable business process.

When models, search, voice, documents and application scaffolding become reusable components, the scarce asset becomes the relationship between the technology and the customer’s work. That is where the next phase of AI value creation will be fought.

References

1. DeepSeek API documentation: application capabilities, Search, Fetch, OCR, ASR and TTS services, November 2026.

2. DeepSeek API documentation: model pricing and API compatibility conventions, November 2026.

3. OpenAI: official API and business pricing pages, November 2026.

4. Anthropic: official Claude API pricing and enterprise documentation, November 2026.

5. Google: Gemini Developer API pricing and Gemini model documentation, November 2026.

6. PyTorch and Google Cloud: TorchTPU announcements and native PyTorch support for Google TPUs, October 2026.

7. Public model documentation and release material for Qwen, DeepSeek, Kimi, GLM, Llama and Gemma.

8. Vishal Sharma, “Atlassian’s Evolution: Navigating Market Pressure and the AI Revolution,” LinkedIn.

9. Vishal Sharma, Strategic Foresight articles and AI value-chain analysis.

10. Pricing and feature availability change frequently. Recheck provider documentation before procurement or production deployment.

Jensen Is Out What happens next for Nvidia, AI and the US-China chip war

 The Architect of the AI Boom - Jensen Huang at a Crossroads

For three decades, Jensen Huang has been Nvidia. He co-founded the company in 1993 and led it as it became the supplier that almost every AI lab, cloud provider and government depends on. By most estimates it generates somewhere between 75% and 90% of the world's AI-chip revenue, and at its early-October record it was worth about $5.8 trillion.

On 12 November 2026, he resigned.

But here's the uncomfortable truth: the scandal is only the trigger. Huang left after admitting an affair with a former executive of a company US officials suspect of diverting chips to China, and after reports of a relationship with a direct report. What he leaves behind is a flat organisation of about 60 direct reports, no named successor, a board that had no permanent chair until this week, and a web of financing commitments that has earned Nvidia the nickname “central bank of AI”. The stock closed down 4.3% on the day and sits roughly 30% below its record.
So what's really going on? Let's dig into what happened, what the ChinaTalk panel made of it in the first hours, whether Huang can come back, what Washington is doing, who wins and loses across the AI stack, and what it all means for the US-China race.

A quick note before we start. This article extends the Nvidia chapter of my book, Gen AI - The New Reality (book pp. 22–29, with related material on pp. 42, 73, 97-98 and 117). When the book was written in early 2024, Nvidia was “Godfather of AI”, a US$1.12 trillion company with a CUDA moat, a Taiwan link and a long list of risks. Two years on, the biggest risk on that list was one the book never put on a slide: the founder himself. The ChinaTalk podcast transcript does not label speakers, so I attribute quotes to “the panel”, or to Dylan or Jordan where the conversational cues make that clear. And much of what is reported about Huang is allegation. He has not been charged with any offence, Nvidia has said it had no reason to believe national security or company secrets were compromised, and the Chinese government rejects the espionage claims. Where this article says “reported” or “alleged”, it means exactly that. Figures are as at 13 November 2026.

1. Executive summary

Short version: a founder, a scandal and a structural problem arrived in the same week.

Jensen Huang is out, the AI trade has lost its figurehead, and Washington has just been handed the political momentum it needed to tighten chip controls. Those three facts drive everything else in this article.

• What happened. Huang resigned on 12 November after he admitted an affair with Alice Huang, a former Megaspeed executive whom US intelligence reportedly suspects of working for China’s Ministry of State Security, and after the New York Times reported a relationship with Kimberly Powell, a senior Nvidia executive who reported to him. CFO Colette Kress is interim CEO while the board searches; Harvey C. Jones becomes Nvidia’s first-ever board chair.
• Why it matters more than a scandal. Nvidia is the “central bank of AI”: it supplies the chips and also finances the customers who buy them, through equity stakes, lease guarantees and off-balance-sheet backstops. A founder who personally ran roughly 60 direct reports and the key relationships has left with no named successor [3].
• The market reaction. Shares opened about 13% lower on 12 November, touched $148.80, and closed at $166.19, down 4.3%. The stock is roughly 30% below its early-October record, when the company was worth about $5.8 trillion.
• The podcast’s core argument. Export controls should target chipmaking tools (the MATCH Act) more than remote cloud access (the RASA bill), because an outright ban on rental invites retaliation and more smuggling. Nvidia will be fine in the short term but faces a harder medium-term test if executives leave and decisions start going to committees.
• Can he return? Possible in theory, unlikely in practice. Steve Jobs and Sam Altman both came back, but neither left under a national-security cloud. Huang resigned from the board too, faces a law-firm investigation, congressional interviews and reported Justice Department probes into companies around him.
• Washington. Trump praised Huang and suggested Tim Cook as successor; Treasury Secretary Scott Bessent publicly blamed Commerce, which “has the pen”, for not blacklisting Megaspeed; Commerce Secretary Howard Lutnick will testify to the Senate Banking Committee on 10 December while advisers discuss whether he should go. Democrats are pressing on oversight and on Trump’s handling of the case rather than on antitrust.
• Winners and losers. AMD, Broadcom, Google and Amazon gain narrative momentum for their own silicon, and Chinese open-weight models keep eating share. Neoclouds, Supermicro and anyone relying on Nvidia’s balance sheet are most exposed. OpenAI and Anthropic are mixed: more dependent on Nvidia hardware, but freed from Huang’s open-source and China-friendly lobbying.

Nvidia and the AI race: key numbers at 13 November 2026

2. Nine days that ended a 33-year run

It took nine days to go from record high to resignation.

The speed of the collapse is part of the story. Nvidia touched a record close of roughly $5.78 trillion on 6 October. By late October the New York Times had reported that Huang was the target of a Chinese intelligence operation run through his relationship with Alice Huang; Wired dates that report to just under two weeks before the board opened its investigation on 8 November. Huang told employees that same Sunday that he had been unfaithful to his wife, adding that there was “no excuse” for what he did.

Nine days: from record high to resignation

The board first stood by him, then reversed. It retained Wachtell, Lipton, Rosen & Katz, led by a former acting US attorney, to investigate; on 11 November the New York Times reported a second relationship, this time with an executive who reported to him; and by the evening the board was meeting to consider his removal. Reuters later reported a third, unnamed relationship with another employee. Nvidia announced the resignation about 42 minutes before the US market opened on 12 November.

Wired’s source offers a useful reading of the board’s thinking: the alleged relationship with a subordinate, together with Huang’s business dealings with Megaspeed, probably weighed more heavily than the espionage allegations themselves. That distinction matters later in this article, because it separates a governance failure (which boards can act on quickly) from a national-security failure (which prosecutors and Congress will take months to unpick).

The market moved violently and then partly retraced. Shares fell as much as 14% on the open on 12 November, which would have been the largest one-day market-value loss on record had it held, then recovered to close down 4.3%. Retail investors poured in a record US$111 million in the first half-hour of trading, according to Vanda Research. On 13 November the stock slid roughly another 4%, back under $4 trillion. By one count about $550 billion of market value was lost in the week and over $1 trillion across five weeks; by another the week’s damage exceeded $600 billion. Treat these as approximate: the numbers moved by the hour.

Nvidia’s market value and share price, from the book’s 2023 snapshot to exit week

Analysts had been bracing. Several told CNBC before the resignation that losing Huang could cost the stock about 20%, and at least seven firms downgraded Nvidia in the preceding week. Forbes contributor Peter Cohan put the value at risk from losing Huang at between $700 billion and $1.1 trillion, around $600 billion of which had already gone. Against that, the fundamentals the market is pricing are still extraordinary. Q2 fiscal 2027 revenue was $96.2 billion, up 106% year on year, with Data Center revenue of $89.0 billion and a 75% gross margin.

Nvidia quarterly revenue, from $13.5B to a $108B guide

To put that in the book’s terms: page 25 celebrated $13.5 billion in a record quarter and a 141% jump in Data Center revenue between quarters. Nvidia now books roughly seven times that in a single quarter. The business the next CEO inherits is an order of magnitude larger than the one the book described, and that scale is exactly why the succession question is so uncomfortable.

3. What the ChinaTalk panel said, chapter by chapter

What the sharpest voices in chips said in the first hours after the news.

The podcast was recorded in the immediate aftermath of the news. The panel (the host plus Dylan and Jordan of SemiAnalysis) debates two things at once: what Huang’s exit means for US export policy, and what it means for Nvidia and AI. Here is the argument in order, with the strongest lines kept verbatim.

3.1 Chapters 1–2: Jensen is out, and the legislative fork

The opening is half gallows humour, half policy alarm. Dylan’s first worry is not Nvidia but the political reaction: he fears the scandal “is going to be used as cover to pass really stupid laws.” He then draws a line between two bills. The MATCH Act would extend controls on advanced chipmaking tools such as deep-ultraviolet lithography to China; the Remote Access Security Act (RASA) would close the cloud loophole that lets Chinese companies rent banned GPUs through cutouts in places like Malaysia. Dylan wants the first and is hostile to the second: “RASA’s pretty dumb bill.” He also supports Biden-style diffusion controls, and the panel notes that framework was issued but never enforced.

MATCH “is far more important for US national interest than the RASA Act.” - Dylan, ChinaTalk

3.2 Chapters 3-4: why not ban everything, and will China retaliate?

Dylan’s game-theory case is the intellectual centre of the episode. His claim: if the US bans every route to compute, including rental, China has no reason not to “blow the whole thing up”, because Chinese suppliers sit inside American supply chains. He points to the largest PCB supplier to Nvidia being a Chinese company. The optimum, he says, is a regime that minimises what China can access and maximises American oversight.

“What is the minimum amount of compute that China can have with the most amount of American oversight to where they will not blow up the supply chain.” - Dylan, ChinaTalk

“We do not have the stomach to disentangle the supply chain, very clearly.” - Dylan, ChinaTalk

His practical conclusion is counter-intuitive for hawks: “the best thing for America to do is allow remote access of GPUs because the alternatives are much worse”. Hyperscaler-controlled rental gives Washington a kill switch, keeps Chinese data visible to US intelligence, and sends enormous margin dollars to Nvidia and the US clouds. He cites roughly $7 million of Nvidia profit per megawatt deployed plus another $7 to $10 million for the cloud provider at current pricing, a back-of-envelope figure rather than a reported one.

The host pushes back on several fronts. Beijing may actually welcome a ban, because parts of the Chinese state are embarrassed that its champions want Nvidia chips and would like them forced into the domestic ecosystem. Xi Jinping also has reasons not to start an economic war with an economy that, as one speaker puts it, isn’t growing. Others counter that earlier rounds of tit-for-tat ended in compromise, that Beijing did not retaliate against the 2022, 2023 or 2024 rules, and that this administration has shown it will absorb political pain (the speaker cites the Iran war). The disagreement is not resolved, and it is the most honest part of the episode.

3.3 Chapter 5: the smuggling evidence

This is where the conversation turns from policy to Nvidia’s proximity to wrongdoing. The panel lists what was then public: Supermicro’s co-founder arrested, a Supermicro Taiwan sales manager a fugitive, and the now-famous detail of a staff member moving serial-number stickers between servers during an audit. One panelist asks how Nvidia could stop reporting Singapore revenue, which I have not been able to verify. One panelist is careful on the legal point: the bar for personal complicity would be Huang actually knowing and acting, but “either way” the pattern around him is damaging.

“The picture of the Supermicro Taiwan guy with the hair dryer in his hand moving the stickers… It’s pretty like damning evidence that they’re aware of this.” - ChinaTalk panel

Two further observations are worth keeping. First, the panel notes that the timing coincides with the midterms: “the cover is gone”, as one speaker puts it, and Lutnick and Huang lose protection together. Second, they point out that the Biden diffusion framework was never repealed, only waved away in a footnote in May 2025. The enforcement gap, not the rule book, is the recurring theme.

3.4 Chapter 6: China is getting off GPUs anyway

Dylan and Jordan argue that policy is arriving late to a trend already under way. GLM shipped with support for seven domestic accelerators; DeepSeek published a megakernel for Huawei Ascend for its V4 model; Meituan claims it trained LongCat entirely on Ascend. One panelist’s summary is the most quotable line about the state of play: “the gap in the model capabilities is wider at this point, but the gap in chip capabilities is like as close as it’s ever been.” Both observations are borne out by the later data in this article.

3.5 Chapter 7: flip the table

Dylan’s strategic alternative is to go after the bottom of the stack rather than the top. Remove one jigsaw piece and China will engineer around it; blow up the board and it is harder. So push China up the value chain into buying expensive things it would rather make cheaply.

“Don’t let them buy chips, make them rent the chips. Don’t let them rent the chips, make them buy the AI.” - Dylan, ChinaTalk

The panel converges on equipment: expand MATCH beyond EUV so the Japanese, Dutch and Koreans cannot backfill American restrictions, and make China build lithography from scratch. The panel also stresses scale: it is not enough for Huawei to match a phone chip or a Hopper-class accelerator; the test is whether China can make hundreds of thousands of them.

3.6 Chapter 8: open models and talent

The conversation then widens to why US labs have not simply out-spent Chinese open-weight models. Dylan’s answer is blunt: “It’s not money, it’s talent.” He argues that Moonshot is, skill for skill, as good as the US labs and is held back by compute and deployment. A panelist adds that Nvidia itself spent heavily on its Nemotron open models without beating the Chinese open releases. Dylan, playing devil’s advocate for Huang’s own argument, makes a point that cuts against the hawkish framing: American inference companies earn real money serving Chinese models. “America is making so much money off of the Chinese models.”

3.7 Chapters 9–10: Nvidia without Jensen, and who runs it next

This is the most useful part for investors. One panelist separates the short term from the medium term. In the short term NVIDIA is fine: it has commitments to meet and a supply chain to execute on Rubin and Rubin Ultra, and “nobody set them up for success as well as Jensen could, but now they just need to execute on the stuff he set up.” The test comes later and it is about people. Jensen has been unusually good at keeping executives, and without him, “this is probably when the sharks are in the water”: venture capital will pay billions to anyone who wants to leave and start a chip company or a neolab.

The bear case is operational. Dylan says that Nvidia’s speed depends on Jensen being a decision engine: “Things get solved slower, things get solved by committees, things don’t get escalated up.” He recalls the story that the Volta GPU had no Tensor Cores until three months before tape-out, when Huang ordered them added. Stack up a hundred decisions like that, he says, and Feynman could slip, Rubin yield could ramp slower, and by 2029 the best chips may not be Nvidia’s. The rebuttal from the others is that this is 2026, not 2022: there are “lots of other competent teams” and Nvidia remains the leader.

“If NVIDIA slows down, it’s not like someone else accelerated.” - ChinaTalk panel

“Hardware, it progresses slower without Jensen at the helm of NVIDIA, period.” - ChinaTalk panel

On the Steve Jobs analogy the panel splits. One side argues that buying Apple after Jobs stepped back would have made a lot of money. Another replies that “I don’t think Steve Jobs exerted as much control over Apple as Jensen does over NVIDIA or Elon does over his companies”, because Jensen negotiates supply contracts himself. The panel’s dream CEOs are Lisa Su, Charlie Kawwas of Broadcom and, jokingly, Hock Tan. On Kress, one panelist doubts that any senior executive can be wholly unaware of what happened with the chips, while others note she is not the one in Asia making agreements. I have found no report alleging wrongdoing by Kress, and the panel’s remarks should be read as speculation about congressional testimony, not as a factual claim.

3.8 Chapters 11-12: finance, sales and technology, and the nationalisation debate

One panelist closes with a governance observation that deserves to be in every management textbook: every company has a permanent tension between sales (which wants deals), technology (which wants big bets) and finance (which has to pay for it). With Kress as interim CEO, “Finance is in charge now”, which probably means a profitable stretch followed by a different company in the medium term. The final minutes become an argument over whether the US government is effectively nationalising Nvidia, a theory one panelist floats and the others reject. The pushback is that “there was like so much between don’t break the law and nationalized”, and that any of the CEOs being floated are “just logical CEOs”. It is the least evidence-based exchange of the episode and is included here for completeness; nothing in the reporting supports state control of Nvidia.

Where the podcast and the reporting agree. Almost every strategic claim in the episode survives contact with the newest news: Congress is now moving on exactly the three bills the panel discussed, Nvidia’s China revenue has collapsed while Chinese models have surged, and the succession vacuum is the central fear of analysts. Where the panel was ahead of the news was the sequence: it predicted the legislative reaction before the House speaker, the Democratic leader and the Senate majority leader all said so on the record.

4. It is economic warfare, and smuggling is the front line

This is not just a scandal. It is a front in an economic war.

The most useful sentence in the podcast is also the simplest: the US and China are in an economic war. Dylan says he wrote that into his reading of the October 2022 chip rules and still believes it. Read that way, three things that look unrelated are one story: a founder CEO who lobbied for access to the Chinese market, a procurement chain that leaked restricted hardware, and a customer base that learned to rent what it could not buy.

4.1 What the record shows so far

The strongest public evidence of diversion is not about Huang at all. In March 2026 the Southern District of New York charged Supermicro co-founder Wally Liaw, a Taiwan sales manager and a contractor with conspiring to divert about $2.5 billion of servers containing Nvidia chips to China. Prosecutors said the group used fabricated paperwork, a pass-through company and staged non-working dummy servers to satisfy audits. The contractor pleaded guilty in October; Liaw and the Taiwan manager remain charged, the manager a fugitive when last reported, and Supermicro itself is not a defendant.

The Megaspeed thread is newer and murkier. Megaspeed is a Singapore data-centre company that became one of Nvidia’s largest buyers in Southeast Asia. Semafor reports Nvidia sold it about $4 billion of chips. The US government suspects Megaspeed of directing advanced chips to China, and Alice Huang is a former Megaspeed executive. Bloomberg reports that the Justice Department is running three probes: an antitrust inquiry into Nvidia and two others on Megaspeed and Supermicro, while Manhattan prosecutors are examining whether entities linked to Alice Huang were involved in moving Supermicro hardware to China. Neither Nvidia, Huang nor Supermicro has been charged. Supermicro says Megaspeed has never been its customer.

The scale estimates are large but come from analysts, not courts. SemiAnalysis, the firm behind the podcast guests, estimates that more than 1.5 gigawatts of restricted Nvidia hardware has reached China, excluding chips designed to comply with the rules such as the H20; Tom’s Hardware notes the firm does not disclose unit counts and says the Commerce Department has reached a similar conclusion about scale. Convert that at 700 to 1,000 watts per accelerator and you get something between 1.5 and 2.1 million GPUs, which is the right order of magnitude to explain why Chinese labs have been able to train frontier models.

4.2 The legal loophole the book already flagged

Smuggling is only half the story. The other half is perfectly legal. Page 27 of the book warned that, despite controls aimed at physical exports, “most Chinese technology firms access chips virtually using services offered by cloud computing companies”, and that Nvidia had called out intentions to expand cloud supercomputing in China. That is exactly the gap RASA is meant to close. The bill passed the House 369 to 22 on 12 January 2026 and has sat in the Senate Banking Committee since; the Senate text (S. 3519) is narrower than the House version and covers only infrastructure-as-a-service. Commerce lawyers say they lack authority to police rented compute until Congress acts, and one analyst note estimates that the compute China reportedly accesses remotely could lift its capacity by at least 60% in 2026.

The headline case study is Moonshot. A White House science official accused the lab in July of running workloads on Nvidia GB300 servers in a Thailand data centre, and the lab’s Kimi K3 model reportedly drew on around 20,000 H200 accelerators accessed through an Alibaba compute-sharing arrangement. That is a single secondary source and should be read with care, but it matches what the panel described.

The four ways Chinese labs reach Nvidia compute

4.3 Misjudgement, wilful blindness or something worse?

The question the panel keeps returning to is what Huang knew. The evidence supports three distinguishable readings.

• Misjudgement. Huang genuinely believed that selling to the world, including China’s approved customers, was in America’s interest and that diversion was rare. Reuters notes that until the Justice Department brought indictments last year, Huang and Nvidia had argued there was no chip smuggling and that the company complied with export laws. Trump said Huang told him he had no idea the woman was a spy.
• Wilful blindness. Nvidia’s own customers and partners were the vector. A company that books $4 billion from a single Singapore buyer and sees its partner’s staff move audit stickers is, on this reading, choosing not to look. The New York Times reported that Nvidia’s general counsel intervened in a Commerce Department investigation into Megaspeed, a claim reported in a newsletter republished by Senator Banks’s office and not independently confirmed here.
• Active complicity. No public evidence supports this. The legal test the panel identified, Huang knowing and acting, has not been met by anything on the public record, and he has not been charged.

The honest position is to say that the first reading is not disproved, the second is plausible, and the third is unproven. What is not in doubt is the governance cost. A CEO who lobbied for China access, who was the target of an intelligence operation, and whose closest commercial partners were under investigation had given his board a problem it could not contain.

4.4 The workplace affair is a separate, and fully sufficient, problem

There is a second, simpler thread, and it may have been the decisive one. A relationship with a direct report is an ethics and governance issue wherever it occurs. Huang admitted unfaithfulness to his wife on 8 November; the allegation about Powell, reported by the Times on the word of five Nvidia employees and by Reuters for a second employee, concerns someone who reported to him. Powell has not commented publicly on the report. Wired’s source says this, together with the Megaspeed dealings, likely played a larger role in the board’s decision than the espionage allegations. The Times has said it has not seen the video it describes, and CBS has not verified the report.

What the board was dealing with: type of issue vs strength of public evidence

5. What is ahead for Nvidia: the control problem

One man, sixty direct reports, and no succession plan.

The book rightly describes Huang as a visionary who “invented the GPU” and bet early on CUDA (book pp. 22–23). What it could not show is how the company was actually run. The reporting of the past week fills the gap, and it explains why this transition is harder than Apple’s in 2011 or Google’s in 2019.

5.1 A flat structure with Jensen as the integration layer

Huang held the roles of co-founder, president and CEO, and ran an unusually flat organisation with about 60 executives reporting directly to him. Reuters describes a leader who “dug deep into startups’ technology”, helped them win customers inside Nvidia and even joined fundraising calls, personally overseeing the 60 reports and diving into problems across the company. The panel’s anecdote that SemiAnalysis has had to email Jensen to get a fix to some open-source software is the same observation from the outside: in Nvidia, escalation ends at one person.

Two structural details compound the problem. The WSJ reported that Huang once said he did not believe in succession planning . And he was the only Nvidia insider on a board that had no permanent chair until Harvey C. Jones was named on 12 November Nvidia’s own annual report warns that it is “highly dependent” on its long-standing executive team and that failure to ensure effective succession planning could hurt long-term success . The risk was disclosed. It was not mitigated.

“The secret for being CEO for this long is 1) don’t get fired and 2) don’t get bored… How long [will I remain CEO]? For as long as I deserve it.” - Jensen Huang, January 2026.

Forbes contributor Peter Cohan makes the sharpest point: with 60 direct reports each running a slice of strategy, “it is hard to see how any of those direct reports … could have enough of a view of the big picture to take over”. That is the cost of a flat structure. It maximises speed and information flow while the leader is present and leaves no deputy with the integrated view when he is not.

5.2 Family firm characteristics

Forbes also draws attention to the family dimensions of the company: Huang’s wife co-owns the family stake through trusts, and both of their children work at Nvidia [42]. Huang is the fourth-largest investor with a stake of about 3.61%. None of this is improper, but it complicates a clean break. A departing founder with a 3.6% holding, family employees and a record of personally brokering the company’s biggest relationships will remain a gravitational presence whatever the board says.

5.3 Who could run it?

The names in circulation show how thin the bench looks. Internally, Kress, who has been CFO since 2013 and spent nearly 25 years at Cisco and Microsoft before that, is the interim CEO and keeps the CFO job. Forbes lists Michael Kagan, the CTO who founded Mellanox, and Jay Puri, who retired in July, and notes that Ian Buck, who helped develop CUDA and runs hyperscale and high-performance computing, has been floated although analysts say he lacks broad operational experience. Externally, The Information floated Lisa Su of AMD and TechCrunch floated Eric Schmidt. Trump named Tim Cook. The panel’s own wish-list was Lisa Su, Charlie Kawwas and Hock Tan.

The board has said it will consider internal and external candidates and has not given a timeline [1]. In the meantime the first test is the 17 November earnings call, the first under Kress, where Nvidia has guided to $108 billion of revenue, plus or minus 2%, against a consensus of about $109 billion. The quarter closed before the scandal broke, so the numbers will describe the business as it was; the guidance and tone will show how the company is coping without its founder.

What the book would now add to its Nvidia risk page. Page 29 listed customer concentration, supply-chain bottlenecks, Taiwan geopolitics, low-cost ASIC rivals and the CUDA moat. The 2026 version of that slide needs five new rows: key-person risk, governance (no chair, one insider), export-control enforcement, counterparty and financing risk from Nvidia’s own investments, and customer defection to in-house silicon.

6. Can he come back? Jobs, Altman and the limits of the analogy

Jobs came back. Altman came back. Will Jensen?

The question is natural because the book itself contains the best-known recent boomerang. Page 98 records that Sam Altman was sacked as OpenAI CEO on 17 November 2023 and reappointed on 22 November 2023, five days later, after a boardroom crisis that remade the board. And Steve Jobs, pushed out of Apple in 1985, returned in 1997 and built the company that Tim Cook then ran for 15 years. Founders can come back. So why is Huang’s case different?

6.1 The three differences

• The reason for leaving. Altman’s exit was a governance dispute inside a board that could not explain itself; Jobs’s first exit was a strategic fight. Huang left after an admitted affair, a reported relationship with a direct report, and a reported intelligence operation. Those are conduct and national-security issues, and they are not resolved by winning an internal argument.
• Who holds the cards. In 2023 OpenAI’s staff and investors held the leverage and used it. At Nvidia the board moved within four days of opening its investigation, and the cards now sit with the board, a law firm, prosecutors and Congress. Huang has been asked to give a closed-door transcribed interview to three House committees on 1 December and to the Senate Banking Committee, whose chair has warned of a subpoena; Reuters reported through CNBC that he agreed to testify before Thanksgiving.
• The legal perimeter. Bloomberg reports multiple Justice Department probes touching Nvidia, Megaspeed and Supermicro, though no one at Nvidia has been charged. Trump has said the FBI is looking at the romantic and business ties. A CEO who returns while those matters are live is a standing invitation to subpoenas.

 Founder exits: why they left and who held the cards

6.2 Scenarios for Nvidia’s leadership

Here is how I see the five realistic paths. These are qualitative judgements, not forecasts; prediction markets on the next CEO exist but were not used here.

• A. Internal successor stabilises the company. Kress or a technical leader such as Buck wins the permanent job; top-60 talent stays; Rubin ships on time; customers keep ordering.
• B. External heavyweight appointed. The board concludes it needs an outsider with chip credibility and Washington standing; the names floated are Lisa Su, Eric Schmidt and Tim Cook, each with obvious obstacles.
• C. Huang in a founder-emeritus or adviser role. Investigations clear national-security concerns; the board wants his relationships without his authority.
• D. Huang returns as CEO within 12 months. Investigations conclude with no findings against him; staff and major customers revolt; Washington’s posture softens.
• E. Huang joins a Musk company. No formal offer or talks exist today; see section 8.

Huang’s own insistence that he would stay “for as long as I deserve it” is a clue to the board’s calculus. The company’s independent directors were willing to remove a founder who owns about 3.6% of the stock and whom the President of the United States urged to fight for his job [29]. A board that has made that call is unlikely to reverse it within a year.

There is also a market-structure argument against a comeback. The scandal has done real damage to Nvidia’s relationships in Washington. Reuters notes that the crisis threatened to implicate administration members, including Lutnick, who pushed with Huang for relaxing chip sales, and that Huang’s influence over US export policy was part of his value. A newsletter republished by Senator Banks’s office said the reports are setting back Nvidia’s policy-and-influence operation in Washington “in a very real way”. A CEO whose principal political asset has become a liability is a hard return to sell.

7. Washington: Trump, Lutnick, Bessent and the Democrats

Washington has found its villain, and is arguing about which one.

7.1 Trump: loyal until the end, then a successor suggestion

Trump was Huang’s most prominent defender. On the Sunday that Huang admitted the affair, he said Huang had “done an amazing job for our country” and that Huang had told him he did not know the woman was a spy. The New York Times reported that on Tuesday he urged Huang to fight for his job. Wired and Forbes report that US intelligence officials briefed the President on the relationship last year, and he did not act on it. He has also said the FBI is looking into the relationship, while adding, “I think he’ll be in good shape”.

After the resignation he went straight to the succession question on Truth Social:

“…there is only one man to take his place - Tim Cook of Apple. A Legend replaces a Legend!” - Donald Trump

Trump described Cook as Apple’s executive chairman, and nothing in the reporting suggests Cook is interested. The suggestion is best read as a signal about the kind of successor the White House wants: a supply-chain operator and government interlocutor rather than a technologist. The point matters because the next Nvidia CEO will be chosen by a board that has to answer to Congress, not to a President who is himself under questioning about how much he knew.

7.2 Lutnick: the friend who held the pen

Commerce Secretary Howard Lutnick is the administration figure most exposed. CBS reports he was the chairman and CEO of Cantor Fitzgerald, which held about $1.2 billion of Nvidia stock before he joined government, and that he developed a warm relationship with Huang. He frequently consulted Huang on AI policy and praised him as one of America’s great executives, according to a Reuters report carried by the AFR. Critics point to specifics. In 2024 the Commerce Department’s Bureau of Industry and Security opened an investigation into Nvidia’s sales to Megaspeed; according to the New York Times, it was abruptly transferred to the Justice Department. Commerce has not added a company to the Entity List since October 2025, and in June it reportedly held off adding DeepSeek, memory-chipmaker CXMT and more than 100 other firms.

“The real scandal here is the lack of enforcement of export controls.” - Chris McGuire, Council on Foreign Relations, via CBS News

The White House says Trump “maintains complete confidence” in Lutnick, but senior officials are discussing whether he should be pushed out, according to Reuters and CNN, and Trump has fielded calls raising alarm about him Lutnick has agreed to testify to the Senate Banking Committee on 10 December. Republican Senator Jim Banks said that “it appears” Commerce “used loopholes” to let chip sales occur despite the President’s direction, and Warren has demanded that Lutnick and Under Secretary Jeffrey Kessler explain why they have not closed export-control loopholes.

7.3 Bessent: “Commerce has the pen”

Treasury Secretary Scott Bessent went further than any other cabinet member on Friday 13 November. On CNBC he said Treasury, War, the US Trade Representative and State had favoured putting Megaspeed on the Entity List “for months, but Commerce has the pen”, and that he was “frustrated” that leakage of Nvidia chips to China “hurts our great companies in the AI race, gives the Chinese an advantage”. Outlets differ on whether he cited A100 or H100 chips as the ceiling the President had set. Axios reported that Bessent had confronted Lutnick the previous week; Bessent declined to comment on reported tensions between them.

This is an unusual public break. A Treasury Secretary assigning blame to a colleague, on air, for an enforcement failure is a sign that the administration has decided the scandal is best survived by finding a Commerce-shaped culprit. It also gives Congress a ready-made narrative: the policy was clear, the agency with the pen did not use it.

7.4 Congress: bipartisan, and moving

The most consequential reaction is legislative. House Speaker Mike Johnson said there is a growing consensus to limit chip exports to China, and Democratic leader Hakeem Jeffries said he would work to attach the bills to the defence bill. Senate Majority Leader John Thune told Semafor there is a “high probability” of new restrictions, and Senator John Kennedy put the odds at “better than even”, adding, “if you’re going to do it, try not to have an affair with a Chinese spy”. Three measures are in play:

• Chip Security Act. Location verification for exported AI chips, to detect diversion.
• AI Overwatch Act. Congressional review of advanced-chip export licences, and codification of the President’s policy against exporting the most advanced chips to countries of concern. The Senate version, in the Senate NDAA, extends that policy for 18 months and requires licences rather than blanket approvals.
• MATCH Act. Pushes allies to align their chipmaking-equipment controls with US levels. A lobbyist told Banks’s newsletter it is “a tougher one” because of concerns it could aggravate the global supply chain.

Nvidia was the biggest opponent of the Chip Security and AI Overwatch acts, according to a former congressional staffer cited by Reuters. A senior House Republican aide called the moment “a big ‘I told you so’ moment to everyone peddling Nvidia talking points”. Not everyone is cheering: a policy commentary argues the three bills risk handing China the AI race by making US suppliers less reliable and alienating allies. And there is no deal yet to include them in the final NDAA.

7.5 The Democrats: oversight first, antitrust not yet

A natural question is whether any Democrat is pushing this toward an antitrust case. In the reporting reviewed here the answer is no: the Democratic effort is about oversight, records and the President’s conduct. Elizabeth Warren says Huang’s resignation “does not absolve him, other executives, or the company” and that they “still owe Congress answers” about export-control compliance. Warren and Chris Coons asked Rubio, Stephen Miller and Lutnick to preserve records on Nvidia’s chip exports. Democrats are targeting not only Commerce but Trump himself for defending Huang, and Jeffries is aiming to put the chip bills in the defence bill. Chuck Schumer wants all three bills passed. Warren has also co-signed with Republican Banks, which is why this is not a partisan story.

Antitrust does exist as a track, but at the Justice Department: Bloomberg reports that one of its three probes is an antitrust inquiry into Nvidia. Warren-style language about “Big Tech billionaire CEOs” suggests the political appetite is there, but I found no Democratic call for a competition case. That is a gap to watch: if Nvidia’s financing web becomes the focus, an antitrust framing would be a natural extension.

8. Could Jensen end up at SpaceX, and would Musk be fine with it?

Musk says it would be an honor. That is not the same as an offer.

The short answer: Musk has said he would be, there is no sign of an offer, and the practical obstacles are bigger than the headlines suggest.

8.1 What Musk actually said

Late on 11 November, before Nvidia announced anything, the financial blog ZeroHedge posted “Elon please hire Jensen”. Musk replied: “That would be an honor”. Earlier in the week he had agreed with a Sequoia partner who called Huang “irreplaceable for America’s AI leadership”, and after the resignation he replied “True words” to a post saying one does not simply replace a founder. On Friday he reposted the “Air” monologue about Michael Jordan with a one-word caption: “Jensen”. Business Insider and others stress that this is an expression of interest, not a job offer; IBTimes reports no sign that the two have discussed an arrangement.

Investors read it as a signal. SpaceX shares rose roughly 3% to 4% on the day Nvidia fell, at a market value of about $1.96 trillion.

A petition is underway to reinstate Jensen

8.2 Why the pairing makes commercial sense

• Nvidia owns a slice of Musk’s empire. Nvidia holds about 122.8 million SpaceX shares, received through its investment in xAI before SpaceX acquired xAI, valued at about $21 billion in its August 13F filing.
• SpaceX is a very large Nvidia customer. SpaceX has committed to a $40 billion lease for Nvidia chips and equipment, told investors in August it would use Nvidia GPUs exclusively, and has set a 10-gigawatt compute target for the end of 2027, a build-out reportedly costing about $500 billion.

8.3 Why it is harder than it looks

• Legal overhang. Huang faces a law-firm investigation, congressional interviews and a political environment in which any employer would draw scrutiny. SpaceX is a public company with government contracts, including NASA lunar work. Any role touching national-security work would raise vetting questions. That last point is my inference, not a reported fact.
• Conflict of interest. Nvidia is simultaneously SpaceX’s key supplier and a shareholder. A former CEO working for the customer on the other side of supply negotiations would need careful structuring.
• Musk’s own direction. One outlet reports Musk is working to reduce reliance on third-party chips through Tesla’s custom accelerators and a long-term “Terafab” fabrication project. A Huang hire would sit awkwardly with a strategy of reducing dependency on Huang’s former company.
• Precedent cuts both ways. Business Insider notes the 2010 Ellison-Hurd precedent: HP’s board forced Hurd out and Ellison hired him within weeks. It shows speed is possible. It also shows the hire came from a founder-controlled company with a friendly board.

My judgement: an operating role at SpaceX or xAI within the next six months is unlikely. A looser relationship (adviser, board seat or investment partner) is more plausible after the investigations settle. Musk’s enthusiasm is real, but it is a signal about loyalty and positioning more than about an imminent appointment.

9. The circular dependency: Nvidia as the “central bank of AI”

Nvidia doesn’t just sell the picks and shovels. It lends the miners the money.

Circular dependency is the AI-relevant theme, and it deserves its own section because it is where Huang’s exit can travel furthest beyond Nvidia. This section uses “moving into the middle” to mean Nvidia’s push beyond selling chips into the financing and cloud layers that sit between silicon and applications.

9.1 How the loop works

Huang used Nvidia’s swelling cash pile to bankroll the industry that buys its chips. Reuters lists investments spanning AI cloud providers such as CoreWeave and Nebius and suppliers to the biggest labs. CNBC calls the result the “central bank of AI”: Nvidia has made hundreds of billions of dollars of commitments, including cloud-service purchases and project guarantees, and the Federal Reserve is watching the interconnected financing for the possibility that an unexpected event could trigger what Fed officials discussed in July as a “broad-based repricing of assets”. Jon Peddie Research is blunter: Nvidia “is shifting from a tech company to a finance company”, with $279 billion of supply and capacity commitments, $29 billion in cloud agreements, and credit support of up to $105 billion tied to an Ohio campus for OpenAI.

Nvidia’s balance-sheet web (bars are not additive)

In August Nvidia also signed memorandums of understanding with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR to mobilise more than $500 billion of third-party capital, using GPUs as collateral in structures resembling real-estate or toll-road securitisations. The arrangements are preliminary. The vision depends on GPUs being bankable assets, and a founder with the credibility to persuade asset managers that they are.

9.2 What the exit does to the loop

Goldman Sachs analyst James Schneider expects Nvidia to honour all announced commitments but says it “may be more difficult for the company to negotiate new multi-party investments among ecosystem partners at the same scale” [8]. The market is already pricing stress. The yield on Nvidia’s 4.5% bond due 2031 was 5.77% on Friday, nearly 6% higher on the week, and credit-default-swap spreads have widened, most in lower-rated tech names [8]. The smaller counterparties are the weak link: Nvidia had $36 billion of off-balance-sheet cloud agreements, or backstops, for neoclouds at its last quarterly filing, and CoreWeave, Nebius and IREN all fell sharply [8].

Not everyone sees a systemic threat. Former FDIC chair Sheila Bair told CNBC Nvidia is not “too big to fail in a bank sense at all”, because it is not exposed to a run and would still make a product customers want, though “Nvidia going down would be a pretty seismic event” and regulators should watch bank exposure. Ben Barringer of Quilter Cheviot made the same split between short and long term: the news “doesn’t necessarily harm Nvidia in the short-term”, but “what this news threatens to do is harm the wider AI ecosystem in the longer-term” [60].

Sovereign deals are the least transparent part of the web. The Guardian notes that Huang spent 2025 persuading world leaders to develop “sovereign AI” with Nvidia chips, with deals that were not particularly transparent; Nvidia appears to have promised 260,000 advanced chips to South Korea without it being clear when they would be delivered or who would pay [61]. Those are relationship-based promises, and relationships were Huang’s personal asset.

9.3 The Anthropic IPO is the first live test

Anthropic is days from an IPO targeting about a $1.4 trillion valuation, and had said it would take $10 billion from Nvidia. Bloomberg Intelligence’s Mandeep Singh wrote before the resignation that Nvidia’s share of an IPO-day investment “could face heightened scrutiny” if Huang were removed and that a transition “could prompt Nvidia’s board to reassess the size, timing or strategic rationale of the commitment”. PitchBook’s Harrison Rolfes disagrees: “It has nothing to do with Jensen’s overall control. It’s just: ‘We made the investment, and we’re good to go’” [19]. Whichever is right, the Anthropic outcome will be read as a signal of whether the loop survives its architect.

The book anticipated the shape of this risk without naming it. Page 26 compared the AI boom with the dot-com bubble: every start-up with venture funding bought servers from Sun, storage from EMC and databases from Oracle, those three “made a lot of money for several years”, then the bubble burst and the survivors were those on cheaper open technologies. The 2026 twist is that the vendor is also the lender. In the dot-com era Sun did not underwrite its customers’ data centres.

Nvidia’s move up the stack: where the balance sheet reaches

10. Who gains and who loses: AMD, Google, Amazon, Chinese models, OpenAI, Anthropic and the hyperscalers

Every crisis creates a leaderboard.

The right order is to identify the dependency first, then see who can exploit it. The answer is that Huang’s exit helps challengers in narrative and may help them in substance, but nobody escapes Nvidia’s gravity in the next two years.

10.1 The scale of the dependency

Estimates of Nvidia’s share vary by definition. Forbes says it generated about 85% of global AI-chip revenue, with share expected to fall to about 75% this year; Al Jazeera cites about 90% of the market for advanced AI chips [63]; the book put it at more than 70% in 2023 (p.22). The direction is down, not the level. One industry analysis projects Nvidia’s inference share falling from about 90% to 20–30% by 2028, but notes the key custom chips (Trainium, TPU, Maia, MTIA) are captive to their builders and cannot be rented outside them. That qualification matters: it limits how much of Nvidia’s business the hyperscalers’ chips can take directly.

10.2 The first-day market read

On the day of the resignation AMD and Broadcom rose about 4.7% and Qualcomm about 2.7%, while Intel slipped 0.8%, all by 11 a.m. ET. Elsewhere, Marvell fell 2.6%, Micron 1.6% and Dell 0.7%, and Supermicro dropped about 11% on a separate DOJ report. The Seoul Economic Daily lists Samsung, AMD, Broadcom and Google as potential beneficiaries of an unwinding of Nvidia-centred circular deals. Analysts quoted by Mint expect AMD’s Instinct accelerators to compete head-on with Nvidia and Broadcom to gain from big tech’s preference for custom accelerators.

10.3 Google: the TPU is the credible alternative

Google’s case is the strongest because it owns both the chip and the model. In April, Alphabet reported Google Cloud revenue up 63% to $20 billion with backlog nearly doubled to $462 billion; the eighth-generation TPU offers three times the processing power of Ironwood; and Sundar Pichai said Google will deliver TPUs to a select group of customers in their own data centres this year, with most revenue from 2027. Morgan Stanley estimated that selling 500,000 TPUs could add about $13 billion to Google’s 2027 revenue. On 20 October Google released TorchTPU as open source, letting PyTorch applications move to TPUs with a configuration change.

This is the scenario the book flagged in 2024. Page 29 identified Chris Lattner’s Modular as a threat to Nvidia’s software moat, and page 42 argued Google’s TPUs could deliver 40-70% cost savings on some workloads. TorchTPU and Amazon’s TorchNeuron are the industrial version of that argument: reduce the switching cost that made CUDA a mousetrap.

10.4 Amazon: Trainium sold to Nvidia’s own customers

Amazon’s chips business topped a $20 billion annual revenue run-rate, growing at triple digits, with Trainium commitments above $225 billion; Trainium2 is about 30% cheaper per unit of performance than comparable GPUs, and Anthropic, OpenAI and Meta are named customers. AWS grew 28%, its fastest in 15 quarters. Andy Jassy said there is a good chance Amazon will offer full racks of Trainium beyond its own cloud over the next couple of years [68]. Amazon released TorchNeuron on 19 October and is submitting code to decouple PyTorch features from Nvidia hardware.

10.5 OpenAI and Anthropic: more dependent on Nvidia, less constrained by Huang

These are the labs most exposed to the financing loop and also those with the most to gain from the political reset. Business Insider’s reading of Anthropic is the clearest: Dario Amodei has battled with Huang for influence in Washington over chip exports, Huang used his perch to support open-source models that compete with Anthropic’s, and Anthropic’s diversified chip strategy now looks wise. The risk is that its race for compute relies in part on Nvidia-backed neoclouds [19]. OpenAI reportedly told lawmakers it supports the Chip Security, AI Overwatch and MATCH acts, a pivot from a company with deepening financial ties to Nvidia, according to a trade-press summary of reporting by Politico and Bloomberg Law that I have not independently confirmed. It also has up to $105 billion of Nvidia credit support attached to its Ohio campus.

10.6 The hyperscalers: dependent customers who are also competitors

Forbes notes the exit “generated great uncertainty for hyperscalers and neoclouds” that rely on Nvidia GPUs. Hyperscaler bond yields rose slightly on Thursday on the persistent concern about AI deb. But the hyperscalers are in the best position of any group: they buy Nvidia chips, build their own, and increasingly sell the alternatives. Their bargaining position with Nvidia improves with every month Nvidia lacks a permanent CEO, because the person who set prices, allocations and financing terms is gone.

10.7 Chinese models: a nuanced winner

A common thesis is that Huang’s loss is good for Chinese models. The reality is more complicated. Narratively, yes: every report about Nvidia’s weakness strengthens the case for Huawei-based stacks and for open-weight models trained on domestic chips, and Huang was the most effective lobbyist for access to the Chinese market. Practically, no: the immediate legislative effect of his exit is tighter controls on exactly the compute Chinese labs rent offshore, and that is bad for Moonshot and its peers in the short run. The panel’s argument is that the substitution toward domestic silicon is happening “with or without the remote access laws being passed”, so the controls arrive late and mostly raise the cost of the transition.

Winners and losers from Huang’s exit

One closing observation from the panel is worth keeping. One panelist’s summary of the industry impact was brief: “this is a bad day for AI progress.” The case for it is that US, Chinese and Taiwanese engineers collaborate across Nvidia, AMD and the labs, and an escalation breaks those links. The case against it is that the technology is now deeper than any one person. Both can be true.

11. China versus the US: the chip gap closes, the model gap is contested

The chip gap is closing faster than the model gap.

11.1 The panel’s framing, tested

The panel’s clearest empirical claim is that the model gap is wider than it looks but the chip gap is “as close as it’s ever been”. Reporting a few months old supports both halves, with caveats.

• Chips. Huawei’s Ascend line sits at the centre of China’s stack. DeepSeek’s V4, released on 24 April, was adapted for Ascend, and Alibaba says its in-house T-Head GPUs have reached scaled mass production. The panel’s own list of domestic accelerators supported by GLM (Moore Threads, Hygon, Cambricon, Kunlunxin, MetaX, Enflame) is about breadth.
• Models. Moonshot’s Kimi K3, released on 16 July, is a 2.8-trillion-parameter model with a one-million-token context window, which the company says benchmarks neck-and-neck with the top proprietary US systems; benchmarks are lab-reported and not independently reproduced [73]. DeepSeek V4-Pro has 1.6 trillion total parameters (49 billion active), an MIT licence and a $0.44 per million input-token price.
• Adoption. On Vercel’s enterprise developer gateway, Chinese models rose from 11% to 55% of token traffic between January and August 2026, driven mainly by DeepSeek V4 and Kimi K3. That is a single platform, but the direction is unmistakable.

Chinese models’ share of Vercel traffic, and parameters of leading open-weight models

The US response is not standing still: Thinking Machines’ Inkling and Meta’s Apache-licensed models are US open-weight answers, according to a November overview, and Beijing was drafting its own restrictions on exports of advanced AI models on 16 October, with already-released open weights exempt. The race has become two-sided, which is why the panel’s “flip the table” logic is about equipment and value-chain position rather than any single chip.

11.2 The H200 story: licences granted, deliveries throttled

Washington authorised H200 sales to China in December 2025, with 25% of revenue going to the US government. By mid-May about ten Chinese companies, including Alibaba, Tencent, ByteDance and JD.com, were licensed to buy up to 75,000 each, yet no chips had been delivered. The reason was Beijing, not Washington: Chinese authorities told companies to limit their use of Nvidia chips to overseas operations and support domestic suppliers, while US rules require H200s to be used in China. In August, ByteDance and Tencent each received about 10,000 units, around 13% of their quota, and every further purchase needs individual approval from China’s National Development and Reform Commission, which also forces buyers to justify why they cannot use domestic alternatives.

China’s share of Nvidia revenue, from 21% to a zero-China guide

The numbers explain why the panel thinks Nvidia’s China business is largely gone for the purposes of valuation. The book (p.29) quoted China at around 21% of revenue; reports now put it at roughly 5% in recent quarters, China Hopper shipments were under 1% of Data Center revenue in Q2 FY27, and Nvidia’s Q3 outlook assumes no China Data Center compute revenue at all. What remains is indirect exposure: offshore rental, Southeast Asia and whatever the diversion estimates imply.

11.3 Policy matrix: where the controls bite

Dylan’s argument can be drawn as a matrix. The horizontal axis is the control point: restricting who may use compute, or who may build it. The vertical axis is blowback to US firms and supply chains. His preference is to apply pressure where the US already has leverage and the blowback is lowest.

Policy levers: control point vs blowback to US firms

11.4 Shenzhen next week

The politics converge on 18–19 November, when Trump is expected to meet Xi around the APEC summit in Shenzhen with trade, rare earths and advanced chips on the table. Outlook India notes that Washington has allowed case-by-case H200 licensing to approved Chinese customers while Congress pushes a law to regulate remote access to US-controlled compute, and that the Huang scandal has added urgency. The previous Trump-Xi summit, in Beijing in May, took Huang along at the last minute and produced no chip deliveries. This one will take place without him and with a Congress that has more leverage.

12. What to watch next

The dates that will decide the next chapter.

Dates are from reporting to 13 November 2026 and are subject to change.

• 17 Nov: Nvidia Q3 FY27 results (about 1:20pm PT) and call (2pm PT); guidance of $108B ±2% against consensus near $109B. Why it matters: first report under Kress; guidance and tone on supply, financing and succession.
• Early the following week: Anthropic IPO targeted at about $1.4T. Why it matters: tests Nvidia’s planned $10B participation.
• 18 Nov: House GOP chairs’ requested closed-door briefing with Huang. Why it matters: first formal congressional exposure.
• 18–19 Nov: Trump–Xi meeting around APEC, Shenzhen. Why it matters: H200, remote access and rare earths.
• Before Thanksgiving: Huang reportedly agreed to testify. Date and format to be confirmed.
• 1 Dec: Closed-door transcribed interview requested by House Foreign Affairs, Intelligence and China committees. Subpoena threatened if he declines.
• 10 Dec: Lutnick to testify to the Senate Banking Committee. Entity List delay; his future as Commerce Secretary.
• Rolling: Senate-House NDAA negotiations on the Chip Security, AI Overwatch and MATCH acts. Whether the bills reach law this year.
• Unscheduled: Wachtell investigation findings; DOJ probes of Megaspeed and Supermicro; permanent CEO announcement. Determines whether any Huang return is even discussable.

12.1 Five predictions, labelled as judgements

• Congress passes at least the Chip Security Act and the AI Overwatch Act in some form this year; MATCH is the one most likely to be diluted or delayed. Thune’s “high probability” and Kennedy’s “better than even” support this, but no deal exists yet.
• Nvidia’s Q3 numbers will be strong, because the quarter closed before the scandal, and the stock will react to the guidance, supply commentary and any financing-commitment language rather than to the print.
• Nvidia names a permanent CEO later rather than sooner. A board that has just created a chair and promised “internal and external” candidates is not looking to rush, and an interim CEO who is also CFO is a stabilising rather than strategic choice.
• Lutnick’s position is the most fragile in the cabinet; a departure before or after 10 December is plausible, given Bessent’s public blame and reports of internal discussions.
• Hyperscaler and neocloud financing terms tighten at the margin; the first visible sign will be in credit spreads and the next multi-party deal that fails to close rather than in any single announcement.

13. Takeaways

If you remember only six things.

• The scandal is the trigger; the structure is the cause. A 60-report flat organisation, no chair, one insider and no succession plan turned a personal crisis into an industry event .
• Export policy has changed direction. The reaction is bipartisan and aimed at enforcement. Expect the bills the panel debated to move, and expect Commerce to be on the defensive.
• Economic war is the correct frame. China is substituting domestic chips and open models faster than controls can act, and the legal rental loophole matters as much as illegal smuggling.
• Circularity is the amplifier. Nvidia’s role as financier makes its governance a systemic question, which is why a CEO resignation moved credit spreads.
• A comeback is possible in theory and improbable in practice. Jobs and Altman returned because of leverage and time; Huang is in a different category.
• The challengers are real but constrained. Google and Amazon have the chips and the software bridges; AMD and Broadcom have the narrative; none can take over Nvidia’s role in two years.

“The graveyards are filled with indispensable men.” - Lloyd Blankfein, quoted by CNBC

What do you think?

Is Nvidia’s strength really in the founder, or in the system he built? And does tighter export policy make the US more secure, or just accelerate China’s move to its own stack? I’d like to hear where you land.

If you found this useful, follow along: I’m tracking how this story and the AI chip race develop. This article is analysis, not financial or legal advice, and it describes allegations as allegations.

Sources and further reading

Book page references (“book p.X”) refer to my book, Gen AI - The New Reality.

1. NVIDIA, “NVIDIA Announces CEO Transition”, 12 Nov 2026 – https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-CEO-Transition/
2. ChinaTalk podcast, emergency episode on Jensen Huang’s resignation (transcript supplied by the author; speakers not labelled) – https://www.youtube.com/watch?v=2jqSNHWdHEY
3. ABC News (Australia), “Nvidia without succession plan as CEO Jensen Huang resigns”, 12 Nov 2026 – https://www.abc.net.au/news/2026-11-12/jensen-huang-nvidia-ceo-succession-plan-after-spy-plot/107376722
4. Fortune via Yahoo Finance, “Before Megaspeed, there was Supermicro”, 12 Nov 2026 – https://finance.yahoo.com/markets/stocks/articles/megaspeed-supermicro-other-nvidia-partner-200126748.html
5. WIRED, “Jensen Huang Resigns as Nvidia CEO”, 12 Nov 2026 – https://www.wired.com/story/nvidia-ceo-jensen-huang-steps-down/
6. CBS News, “Nvidia CEO Jensen Huang resigns in wake of affair with woman suspected of spying for China”, 12 Nov 2026 – https://www.cbsnews.com/news/nvidia-ceo-jensen-huang-resigns/
7. CNBC, “Nvidia’s leadership crisis tests the chipmaker’s role as the ‘central bank of AI’”, 11 Nov 2026 – https://www.cnbc.com/2026/11/11/nvidia-jensen-huang-ai-financial-system-risk.html
8. CNBC, “What’s ahead for AI financing after Nvidia CEO Jensen Huang departs”, 13 Nov 2026 – https://www.cnbc.com/2026/11/13/nvidia-jensen-huang-resignation-ai-financing.html
9. CNBC, “Nvidia stock closes down over 4% after CEO Jensen Huang announces resignation”, 12 Nov 2026 – https://www.cnbc.com/2026/11/12/nvidia-nvda-jensen-huang-second-affair.html
10. The Motley Fool, “Nvidia Stock Has Wiped Out $1.7 Trillion in Value Since October”, 10 Nov 2026 – https://www.fool.com/investing/2026/11/10/nvidia-stock-has-wiped-out-usd1-7-trillion-in-value-since-october-what-could-happen-if-jensen-huang-steps-down/
11. RoboForex, “Nvidia Earnings Date: 17 November Scenarios With an Interim CEO”, 13 Nov 2026 – https://roboforex.com/blog/analysis/nvidia-earnings-date-interim-ceo-scenarios/
12. Euronews, “Ex-Nvidia CEO Huang and US Commerce Secretary Lutnick summoned to testify before Congress”, 13 Nov 2026 – https://www.euronews.com/2026/11/13/ex-nvidia-ceo-huang-and-us-commerce-secretary-howard-lutnick-summoned-to-testify-before-co
13. Forbes, “Jensen Huang Resigns From Nvidia—Trump Says Tim Cook Should Replace Him”, 12 Nov 2026 – https://www.forbes.com/sites/maryroeloffs/2026/11/12/jensen-huang-resigns-from-nvidia-trump-says-tim-cook-should-replace-him/
14. CBS News, “Lutnick agrees to testify as Bessent blames him for not banning Nvidia chip exports to China-linked firm”, 13 Nov 2026 – https://www.cbsnews.com/news/howard-lutnick-scott-bessent-nvidia-megaspeed-testimony/
15. CNN via Yahoo News, “Bessent points finger at Commerce Department in Nvidia scandal as Trump advisers discuss possible Lutnick ouster”, 13 Nov 2026 – https://www.yahoo.com/news/politics/articles/bessent-points-finger-commerce-department-140048324.html
16. Fox Business, “Democratic senators demand NVIDIA records from Trump administration over alleged Chinese spy ties”, 12 Nov 2026 – https://www.foxbusiness.com/politics/democratic-senators-demand-nvidia-records-from-trump-administration-over-alleged-chinese-spy-ties
17. Seoul Economic Daily, “‘Huang Scandal’ Widens as Republicans Join Call to Halt AI Chip Exports to China”, 10 Nov 2026 – https://en.sedaily.com/world/2026/11/10/huang-scandal-widens-as-republicans-join-call-to-halt-ai
18. Forbes (Fiona Riley), “Nvidia Sinks 11% At Open After Jensen Huang Resigns”, 12 Nov 2026 – https://www.forbes.com/sites/fionariley/2026/11/12/nvidia-sinks-11-at-open-after-jensen-huang-resigns-pushing-market-cap-below-microsoft/
19. Business Insider, “How Jensen Huang’s dramatic ouster from Nvidia affects Anthropic’s $1.4 trillion IPO”, 13 Nov 2026 – https://www.businessinsider.com/jensen-huang-fall-could-reshape-anthropic-ipo-ai-chip-race-2026-11
20. In Plain English, “OpenAI Backs Chip Security Act, AI Overwatch, and MATCH Act”, 13 Nov 2026 – https://plainenglish.io/news/openai-chip-security-act-ai-overwatch-match-act
21. NVIDIA, “NVIDIA Announces Financial Results for Second Quarter Fiscal 2027”, 26 Aug 2026 – https://investor.nvidia.com/news/press-release-details/2026/NVIDIA-Announces-Financial-Results-for-Second-Quarter-Fiscal-2027/
22. Reuters, “Jensen Huang steps down: the hands-on kingmaker behind the AI boom”, 12 Nov 2026 – https://www.reuters.com/business/jensen-huang-steps-down-hands-on-kingmaker-behind-ai-boom-2026-11-12/
23. Economic Times (Reuters), “Nvidia CEO Jensen Huang resigns, Colette Kress appointed interim CEO”, 12 Nov 2026 – https://economictimes.indiatimes.com/tech/technology/nvidia-ceo-jensen-huang-resigns-colette-kress-appointed-interim-ceo/articleshow/135772332.cms
24. AP News, “3 men are charged with conspiring to smuggle US artificial intelligence to China”, 19 Mar 2026 – https://apnews.com/article/artificial-intelligence-china-charges-e8f5135a71b8863c66b9c73d04cb0eb2
25. Tom’s Hardware via Yahoo, “Mass smuggling of Nvidia AI GPUs could be the underlying cause of Jensen Huang’s departure, analysts claim”, 13 Nov 2026 – https://www.yahoo.com/news/politics/articles/mass-smuggling-nvidia-ai-gpus-114128089.html
26. Latin Times, “DeepSeek and Kimi K3: The Open-Source Strategy Turning Chinese AI Into a Geopolitical Force”, 22 Oct 2026 – https://www.latintimes.com/deepseek-kimi-k3-open-source-strategy-turning-chinese-ai-geopolitical-force-600424
27. SAYS, “Jensen Huang Resigns As NVIDIA CEO As Report Of 3rd Alleged Affair Emerges”, 13 Nov 2026 – https://says.com/my/news/jensen-huang-resigns-as-nvidia-ceo-as-report-of-3rd-alleged-affair-emerges
28. The Guardian, “Jensen Huang resigns as Nvidia CEO after alleged affairs”, 12 Nov 2026 – https://www.theguardian.com/technology/2026/nov/12/nvidia-ceo-jensen-huang-resigns
29. Reuters, “Nvidia CEO Jensen Huang resigns in blow to AI boom he helped foster”, 12 Nov 2026 – https://www.reuters.com/business/retail-consumer/nvidia-ceo-jensen-huang-step-down-big-blow-ai-boom-2026-11-12/
30. Quartz, “Nvidia market cap faces biggest one-day drop ever after Jensen Huang resigns”, 12 Nov 2026 – https://qz.com/nvidia-huang-resigns-largest-market-value-wipeout-111226
31. Korea Daily, “Nvidia Stock — US Retail Investors Pour $110M Following Crash”, 13 Nov 2026 – https://en.koreadaily.com/nvidia-stock-us-retail-investors-buy-the-dip/
32. CNBC, “Nvidia stock sinks 4% in fallout from Jensen Huang resignation”, 13 Nov 2026 – https://www.cnbc.com/2026/11/13/nvidia-nvda-stock-jensen-huang-transition.html
33. Forbes (Peter Cohan), “Jensen Huang Is Out. Can Nvidia Protect $3.3 Trillion In Gains?”, 12 Nov 2026 – https://www.forbes.com/sites/petercohan/2026/11/12/jensen-huang-out-after-33-trillion-made-how-much-value-will-be-lost/
34. Semafor, “‘A high probability’: Congress pushes tighter export controls following Nvidia scandal”, 12 Nov 2026 – https://www.semafor.com/article/11/12/2026/a-high-probability-congress-pushes-tighter-export-controls-following-nvidia-scandal
35. Yahoo Finance, “Super Micro Contractor Pleads Guilty in $2.5 Billion AI Server Smuggling Case”, 10 Oct 2026 – https://finance.yahoo.com/technology/ai/articles/super-micro-contractor-pleads-guilty-062049353.html
36. Stocks Down Under, “Why Is Super Micro Stock Crashing 11%? DOJ Probes China Chip Smuggling Allegations”, 13 Nov 2026 – https://stocksdownunder.com/why-is-super-micro-stock-crashing-doj/
37. Crowell & Moring, “House Passes Remote Access Security Act”, 14 Jan 2026 – https://www.cmtradelaw.com/2026/01/house-passes-remote-access-security-act-to-limit-adversaries-remote-access-to-critical-technology/
38. Institute for AI Policy and Strategy, “Remote Access Security Act (RASA)”, 15 Jul 2026 – https://www.iaps.ai/research/remote-access-security-act
39. NewsNYou, “The US Banned These Chips From China. China Rented Them Instead.”, 2 Sep 2026 – https://newsnyou.com/analysis/nvidia-chip-remote-access-loophole/
40. Reuters via Deccan Chronicle, “Nvidia Spy Allegations Likely To Spur Laws To Tighten China’s Access To AI Chips”, 10 Nov 2026 – https://www.deccanchronicle.com/technology/nvidia-spy-allegations-likely-to-spur-laws-to-tighten-chinas-access-to-ai-chips-2003159
41. Sen. Jim Banks, “AI & Tech Brief: Nvidia’s Capitol Hill fallout”, 11 Nov 2026 – https://www.banks.senate.gov/news/in-the-news/ai-tech-brief-nvidias-capitol-hill-fallout/
42. Forbes (Emma Waldman), “What Jensen Huang’s Exit Means For Nvidia And Investors”, 12 Nov 2026 – https://www.forbes.com/sites/emmawaldman/2026/11/12/what-jensen-huangs-exit-means-for-nvidia-and-investors/
43. CNBC-TV18, “Jensen Huang’s resignation: From Elon Musk to Donald Trump, what top leaders said”, 13 Nov 2026 – https://www.cnbctv18.com/world/jensen-huangs-resignation-from-elon-musk-to-donald-trump-what-top-leaders-said-on-nvidia-ceos-exit-20033522.htm
44. Invezz, “Nvidia earnings preview: what to expect as Kress faces first test without Huang”, 13 Nov 2026 – https://invezz.com/news/2026/11/13/nvidia-earnings-preview-q3-kress-huang/
45. Business Insider, “Elon Musk says he would hire Jensen Huang if Nvidia’s board removes him”, 11 Nov 2026 – https://www.businessinsider.com/elon-musk-hire-nvidia-ceo-jensen-huang-if-board-ousts-2026-11
46. International Business Times (Australia), “Will Jensen Huang Join Elon Musk?”, 13 Nov 2026 – https://www.ibtimes.com.au/jensen-huang-nvidia-resignation-elon-musk-speculation-1878504
47. CBS News, “Lutnick should have prevented the diversion of sensitive Nvidia AI chips to China, critics charge”, 13 Nov 2026 – https://www.cbsnews.com/news/lutnick-nvidia-ai-chips-china/
48. Australian Financial Review, “Lutnick fights for his job in Nvidia spy scandal fallout”, 13 Nov 2026 – https://www.afr.com/world/north-america/lutnick-fights-for-his-job-in-nvidia-spy-scandal-fallout-20261113-p61dvl
49. US Senate Banking Committee (minority), “Senator Warren Statement on NVIDIA CEO Jensen Huang’s Resignation”, 12 Nov 2026 – https://www.banking.senate.gov/newsroom/minority/senator-warren-statement-on-nvidia-ceo-jensen-huangs-resignation
50. Sen. Jim Banks, “Banks Secures AI OVERWATCH Act in Senate NDAA”, 14 Jul 2026 – https://www.banks.senate.gov/news/press-releases/banks-secures-ai-overwatch-act-in-senate-ndaa/
51. International Policy Digest, “Congress Could Hand China the AI Race by Trying to Win It”, 8 Nov 2026 – https://intpolicydigest.org/congress-could-hand-china-the-ai-race-by-trying-to-win-it/
52. Sens. Banks and Warren, joint statement on NVIDIA, 9 Nov 2026 – https://www.banks.senate.gov/news/press-releases/sens-banks-warren-statement-on-national-security-concerns-surrounding-nvidia/
53. The Hill, “Elon Musk floated hiring Jensen Huang ahead of Nvidia resignation”, 12 Nov 2026 – https://thehill.com/policy/technology/6217460-elon-musk-offers-huang-job/
54. Telegraph India, “Is Elon Musk hiring Jensen Huang? Buzz in ‘Air’ after richest man posts clip”, 13 Nov 2026 – https://www.telegraphindia.com/world/is-elon-musk-hiring-jensen-huang-buzz-in-air-after-richest-man-posts-clip-for-nvidia-ex-ceo/cid/2189244
55. International Business Times (Australia), “SpaceX Stock Defies Tech Selloff, Rising 3% as Nvidia Plunges”, 13 Nov 2026 – https://www.ibtimes.com.au/spacex-shares-rise-nvidia-selloff-jensen-huang-1878498
56. 24/7 Wall St. via Yahoo Finance, “Elon Musk and Jensen Huang’s New Partnership Could Create the Next Era of Technology”, 16 Aug 2026 – https://finance.yahoo.com/technology/ai/articles/elon-musk-jensen-huang-partnership-225748686.html
57. One America News Network, “Musk: Hiring fmr Nvidia CEO Jensen Huang ‘would be an honor’”, 12 Nov 2026 – https://www.oann.com/newsroom/musk-hiring-fmr-nvidia-ceo-jensen-huang-would-be-an-honor/
58. Jon Peddie Research, “Nvidia financial results for Q2’26–FY’27”, 27 Aug 2026 – https://www.jonpeddie.com/news/nvidia-financial-results-for-q226-fy27/
59. CNBC, “Jensen Huang’s Nvidia exit casts shadow over Wall Street’s $500 billion AI financing bet”, 12 Nov 2026 – https://www.cnbc.com/2026/11/12/nvidia-ceo-exit-ai-financing.html
60. Business Insider, “Stocks drop as Jensen Huang’s resignation drags Nvidia and rattles AI bulls”, 12 Nov 2026 – https://www.businessinsider.com/jensen-huang-resigns-nvidia-stock-nvda-sp500-nasdaq100-2026-11
61. The Guardian, “Jensen Huang built an AI kingdom. Now he’s gone, will Nvidia’s ramparts be breached?”, 12 Nov 2026 – https://www.theguardian.com/technology/2026/nov/12/jensen-huang-built-an-ai-kingdom-now-hes-gone-will-nvidias-ramparts-be-breached
62. Forbes (Mike Patton), “Nvidia In Turmoil: What Jensen Huang’s Departure Means For A.I.”, 12 Nov 2026 – https://www.forbes.com/sites/mikepatton/2026/11/12/nvidia-in-turmoil-what-jensen-huangs-departure-means-for-ai/
63. Al Jazeera, “Why did Nvidia CEO Jensen Huang resign, and who is Colette Kress?”, 12 Nov 2026 – https://www.aljazeera.com/news/2026/11/12/why-did-nvidia-ceo-jensen-huang-resign-and-who-is-colette-kress
64. Spheron, “Hyperscaler Custom AI Chips in 2026”, updated 19 Oct 2026 – https://www.spheron.network/blog/hyperscaler-custom-ai-chips-2026-trainium-tpu-maia-mtia-vs-nvidia-gpu/
65. Seoul Economic Daily, “Nvidia’s Huang Forced Out, Shaking AI Circular-Deal Ecosystem”, 13 Nov 2026 – https://en.sedaily.com/world/2026/11/13/nvidias-huang-forced-out-shaking-ai-circular-deal-ecosystem
66. Livemint, “NVIDIA share price crash: Broadcom, AMD to Micron”, 13 Nov 2026 – https://www.livemint.com/market/stock-market-news/nvidia-share-price-crash-broadcom-amd-to-micron-should-you-buy-other-ai-stocks-after-jensen-huangs-resignation-11794540876728.html
67. 24/7 Wall St., “The Race to Beat Nvidia: Does Google or Amazon Have the Better In-House Silicon”, 17 Jul 2026 – https://247wallst.com/investing/2026/07/17/the-race-to-beat-nvidia-does-google-or-amazon-have-the-better-in-house-silicon/
68. Business Insider, “Nvidia’s $4.9 trillion chip empire has a new problem: its biggest customers”, 30 Apr 2026 – https://www.businessinsider.com/google-amazon-nvidia-chips-ai-tpu-trainium-customers-2026-4
69. TechCrunch, “Google has joined Amazon in attacking Nvidia’s stronghold with AI app developers”, 20 Oct 2026 – https://techcrunch.com/2026/10/20/google-has-joined-amazon-in-attacking-nvidias-stronghold-with-ai-app-developers/
70. Tech Times, “Nvidia H200 Chips Enter China at 13% of Quota; Beijing, Not Washington, Controls Rest”, 20 Aug 2026 – https://www.techtimes.com/articles/325078/20260820/nvidia-h200-chips-enter-china-13-quota-beijing-not-washington-controls-rest.htm
71. AI News, “The Nvidia H200 China deal survived the Trump-Xi summit–just not in the way anyone expected”, 19 May 2026 – https://www.artificialintelligence-news.com/news/nvidia-h200-china-deal-stalled-trump-xi-summit-2026/
72. Safeguard, “Chinese AI labs in 2026: DeepSeek, Qwen and Kimi releases and direction”, 19 Oct 2026 – https://safeguard.sh/resources/blog/chinese-ai-labs-deepseek-qwen-kimi-releases-and-stated-direction
73. VentureBeat, “China’s Moonshot AI releases Kimi K3, the largest open-source model ever”, 16 Jul 2026 – https://venturebeat.com/technology/chinas-moonshot-ai-releases-kimi-k3-the-largest-open-source-model-ever-rivaling-top-u-s-systems
74. GeoToolbox, “Best Chinese AI Models (2026): DeepSeek, Qwen, GLM, Kimi”, updated 9 Nov 2026 – https://geotoolbox.ai/blog/chinese-ai-models-compared
75. Council on Foreign Relations, “The Consequences of Exporting Nvidia’s H200 Chips to China”, 9 Dec 2025 – https://www.cfr.org/articles/consequences-exporting-nvidias-h200-chips-china
76. CNBC, “U.S. clears H200 chip sales to 10 China firms as Nvidia CEO looks for breakthrough”, 14 May 2026 – https://www.cnbc.com/2026/05/14/us-clears-h200-chip-sales-to-10-china-firms-as-nvidia-ceo-looks-for-breakthrough.html
77. Outlook India, “A New AI Dialogue, An Old Chip War: What Trump And Xi Face In Shenzhen”, 13 Nov 2026 – https://www.outlookindia.com/international/trump-xi-shenzhen-summit-nvidia-h200-ai-chips
78. CNBC, “Nvidia after Jensen Huang: What founder transitions show”, 13 Nov 2026 – https://www.cnbc.com/2026/11/13/nvidia-jensen-huang-founder-ceo-transition.html