NVIDIA's Agentic CPU Pivot: A New $200B TAM Emerges
A New Reporting Frame
NVIDIA's decision to split its data center revenue into two buckets – hyperscale and "ACIE" (AI clouds, industrial, enterprise) – and to add a third segment for edge computing is more than a reporting change. It reflects a customer base that has diversified far beyond the big cloud providers. In Q1, ACIE grew 31% quarter-over-quarter versus hyperscale's 12%, and Jensen Huang explained the logic: "we wanted to understand our business better." This "second category" – the term Jensen used repeatedly, and which became the company's top keyword this quarter (second category) – is where he sees the long-term economics. “The second category is extremely diverse. Instead of 6 or 7 companies representing the revenues associated with our first category, The second category is hundreds, thousands of companies.” — Jen-Hsun Huang, President and Chief Executive Officer · 2026-05-20 It is also where NVIDIA's full-stack advantage matters most, because these customers buy systems, not chips.
The Vera CPU: The Real News
The most consequential move is the standalone Vera Rubin CPU. Jensen framed it as the world's first CPU purpose-built for agentic AI, opening a new $200 billion TAM. He confirmed that the $20 billion visibility for this year is for standalone CPUs only, excluding the CPU embedded in Vera Rubin systems. “The 20 billion is for standalone CPU.” — Jen-Hsun Huang, President and Chief Executive Officer · 2026-05-20 The company expects to be supply constrained for the entire life of Vera Rubin. This is a strategic inflection: NVIDIA is no longer just selling GPUs with an accelerator, but is building an entire agentic computing stack. As Jensen said, "the harness runs on CPU" – and that harness is the orchestration layer for billions of future agents. This ties directly to a prior theme: last quarter Jensen argued that “compute equals revenues” — Jen-Hsun Huang, President and Chief Executive Officer · 2026-02-25. Now he's extending that logic to CPUs.
Capital Returns and China
Alongside the CPU pivot, NVIDIA sharply upgraded its capital return plan. The quarterly dividend jumps from $0.01 to $0.20 per share – Jensen corrected it to $0.25 during the call – and the board authorized an additional $80 billion buyback. The company plans to return roughly 50% of free cash flow this year. Colette Kress noted, “we do believe that one of the most important things that we can do is really supporting the extreme ecosystem” — Colette Kress, Executive Vice President and Chief Financial Officer · 2026-02-25 – but she also acknowledged the buyback as a key tool. Meanwhile, China remains a drag: the US approved H200 exports, but NVIDIA has booked no revenue and excludes any China data center compute from its outlook.
Financial Momentum
Revenue grew 85% year-over-year to $81.6 billion, with operating income up 219% and record free cash flow of $49 billion. Gross margins held at 75%, guided to stay in the mid-70s for the year. “We generated record free cash flow of $49 billion” — Colette Kress, Executive Vice President and Chief Financial Officer · 2026-05-20 – a stunning number that supports the capital returns. Revenue has tripled in the last eight quarters. The market's 9% drawdown from the May high suggests investors are still pricing in execution risk, but with a $200 billion TAM opening up and a broader platform story, the thesis is expanding.
Agentic AI has arrived. AI can now do productive and valuable work.
The bottom line: NVIDIA is no longer just riding the GPU wave; it is building the compute substrate for the agentic era. The new segmentation, the standalone CPU, and the capital returns all point to a company that believes its most valuable work is ahead. Whether the market rewards that conviction remains to be seen, but the strategic shift is real and unambiguous.