Key Message
- Q2 FY23 - The data centre business, driven by demand from cloud providers and internet companies, played a vital role, contributing $10.32 billion (76% of total revenue). Gaming revenue also grew positively to $2.49 billion.
- Products like the H100 tensor core GPU, DGX supercomputers, inference platforms, and AI Infrastructure-as-a-Service in the cloud are poised to transform AI delivery, partnered with major cloud providers.
- Today, Nvidia commands 70% of the AI Chip market.
- Other Key players are AMD, ARM (Softbank) and Intel.
Where is the Spend
If you take the other Nvidia divisions, separate them out and add them together, revenues were up by 9.9% to $3.18Bn. That’s another way of saying that the Nvidia datacentre business is now 3.2x bigger than the rest of Nvidia, and the datacentre business now accounts for 76.4% of the overall sales for the company.
What is their Economic Moat
A single chip can cost upwards of US$40K, and OpenAI used about 10,000 of them to train ChatGPT. If you want to train a GPT5-level model, you apparently need somewhere on the order of 20K to 25K H100 GPU accelerators, and that is somewhere around two dozen $1 billion machines
Nvidia's Economic Moat is driven by its 1st mover advantage in the AI Chip segment, supported by the network effects of a multisided platform and a frictionless Flywheel. 
Nvidia's Economic Moat is driven by its 1st mover advantage in the AI Chip segment, supported by the network effects of a multisided platform and a frictionless Flywheel.
My other posts on AI Value Chain and how Generative AI can transform Telecoms, Energy and Utilities
Source: Seekingalpha, Bloomberg, Martinfowler.com, Databricks.com Linkedin, Nvidia, Google, AWS
No comments:
Post a Comment