Jensen Huang · NVIDIA
NVIDIA's fiscal 2026 letter from Jensen Huang
Covers FY2026Published 12 May 2026Available at SEC.gov, link checked 7 Oct 2026Shareholder letter · Banking & Technology
Letter begins on page 12 of 175. A large file (23.3 MB): on mobile data, it may take a while to open.
The view as filed on 12 May 2026, not a statement of what the author thinks today.
What to notice
- 1
Huang says inference has overtaken training as the dominant workload, and describes a flywheel: capable agents generate more tokens, more tokens need more compute, and more compute produces smarter AI and further demand.
Continue with L3VLUP analysis
Unlock the 3 remaining takeaways on this letter, and every analysis across the full Letters & Theses library.
Checking what you can open…
“A computer used to be a tool. In the age of AI, it is manufacturing equipment.”
Why read it
NVIDIA's founder sets out how he thinks about the economics of AI factories and tokens, direct context for anyone covering semiconductors, data centres or technology investing.
About this letter
Jensen Huang, NVIDIA's founder and chief executive, wrote this letter to 'NVIDIANs and Stakeholders' for the company's annual report, which brings together the notice of its 2026 annual meeting, the proxy statement and the form 10-K. The letter is dated May 2026 and covers fiscal 2026.
Huang argues that computing is undergoing a complete reset not seen since IBM introduced the System/360 in 1964, moving from computers as tools to a new kind of infrastructure that manufactures intelligence. He describes AI as a new industry built in five layers, from energy and chips through infrastructure and models to applications at the top, and reports a year in which revenue grew 65 percent to $216 billion. He then turns to AI factories, the American manufacturing buildout, enterprise agents, open models, AI as national infrastructure and physical AI.
Companies
- NVIDIA NVDA
Themes
- ai
- technology shifts
- strategy
- competitive advantage
- growth
A large file (23.3 MB): on mobile data, it may take a while to open.
L3VLUP does not host this document. It belongs to NVIDIA and is a public SEC filing; the notes above are L3VLUP’s reading of it.
The same argument, elsewhere
- Nscale's 2026 IPO letter from founder Josh Payne, on AI infrastructure · Josh Payne, Nscale
- Charles Schwab's 2025 letter from Rick Wurster · Rick Wurster, Charles Schwab
- Capital One's 2024 letter from Rich Fairbank · Rich Fairbank, Capital One
- Stripe's 2023 annual letter · Patrick Collison and John Collison, Stripe
The ideas in this letter, on Atlas
- Coding is a task and innovation is the purpose: automate the task, keep the jobName the purpose of a job separately from its tasks; automation changes the tasks and leaves the purpose to be done better or worse.
- Inference has overtaken training, and the flywheel runs on tokensA compute vendor's case is a flywheel from agents to tokens to compute; the buyer's case is the return on that compute, and both have to be read together.