NVIDIA CEO Jensen Huang described a strategic shift from generative AI to "agenting" and physical AI, underpinned by a "full stack" architecture that integrates chips, systems, and software. Huang reported a record 46 billion net income quarter, asserting that "compute equals revenues" and "compute equals GDP" as data centers evolve into "AI factories" producing monetizable tokens. He confirmed massive strategic investments—30 billion in OpenAI and $10 billion in Anthropic—ahead of anticipated IPOs, and highlighted the emergence of robotics and digital biology as the next major growth frontiers.

Key Takeaways

• Financial Performance: NVIDIA reported a "70 billion revenue quarter and a 46 billion net income quarter," citing unprecedented scale and growth.
• Strategic Investments: Huang confirmed a finalized "30 billion investment in OpenAI and a 10 billion investment in Anthropic," noting OpenAI is expected to go public "towards the end of the year."
• AI Evolution: The industry has moved from generative AI (Chat) to reasoning, and now to "agenting" (actions/creating). A new open-source software, "open claaw," surpassed Linux in downloads in just three weeks.
• Factory Economics: Data centers are now "AI factories" where "tokens per watt" is the critical metric; NVIDIA claims its architecture offers an "order of magnitude" better performance per watt and dollar than alternatives.
• New Demand Sources: Beyond existing hyperscalers, Huang identified a "brand new lab," referred to as "MSL," which requires millions of GPUs, adding to demand from OpenAI and Anthropic.
• Future Frontiers: The company is aggressively targeting physical AI and robotics, with its "Groot" model becoming the number one human robotics model; Huang predicts the industry will focus on physical AI in "two years time."
• Market Transformation: Huang predicts the entire $2 trillion IT industry will transition from renting tools to consuming tokens, as every software company becomes an "agency company."

Q&A

What strategically and technically had to come together to deliver this type of hypergrowth, moving from millions to billions in net income?

• Jensen Huang: NVIDIA succeeded by adopting a "full stack" approach over 33 years—owning the CPU, GPU, networking (NVLink/Spectrum X), and software. This control allows them to "build an entire infrastructure each year" rather than just a chip, ensuring they stay ahead of competitors who are "connecting too many cats and dogs."

How do you see the enterprise market evolving from generative AI models to reasoning and agenting?

• Jensen Huang: The industry has moved through three inflection points: generative (ChatGPT), reasoning (grounded on truth), and now agents that can "use tools," "do search," and "do planning." This shift increases compute demand exponentially, as agents run continuously in the background consuming "a million times more tokens" than simple queries.

With the extraordinary demand for tokens, how does the massive CapEx required to build these AI factories get financed?

• Jensen Huang: Financing is driven by the fact that "compute equals revenues" and "compute equals GDP." Companies know that if they have "ten times the alternative" in tokens per watt, their revenues will be higher; therefore, the entire IT industry's capital expenditure is shifting toward AI because it is the only way to power the future "digital workforce."

How do constraints regarding memory, power, and permitting play out, and is it negative if the build-out cycle takes longer?

• Jensen Huang: Constraints are beneficial because they force customers to "choose the best" architecture to maximize revenue per megawatt. Because power is limited, companies cannot risk "choosing the wrong foundry" or system; they must select NVIDIA’s architecture to ensure their "revenues next year" are not compromised.