Wolfe Research: Nvidia’s $1T Orders Guidance Is a Floor, Not a Ceiling

NVIDIA (NASDAQ:NVDA | NVDA Price Prediction) CEO Jensen Huang walked onto the GTC 2026 stage and dropped a number that stopped the room: purchase orders for the Blackwell and Vera Rubin platforms are expected to hit $1 trillion. That alone would be a headline for any company on earth. But Wolfe Research Senior Analyst Chris Caso thinks the Street is still not thinking big enough.

NVIDIA (NVDA)

“We think that represents more of a floor than a ceiling.”

That is the sharpest analyst read to come out of GTC 2026, and it deserves unpacking.

The Timeline Is the Tell

In October, Nvidia disclosed $500 billion in orders covering 2025 and 2026. The new $1 trillion figure expands that window to cover 2025, 2026, and 2027. Here is what matters most: it is only March 2026. The 2027 order book is nowhere near closed. Hyperscalers, sovereign AI projects, and enterprises are still signing deals. The $1 trillion is what Nvidia can see today, not the final tally.

Caso’s argument is that the $1 trillion figure already provides “pretty healthy upside” to where consensus estimates currently sit for 2027. Sell-side models likely need to move higher, and investors pricing Nvidia off current Street numbers may be anchoring to a figure that is already stale.

What the Numbers Already Show

The foundation under this thesis is not speculative. Nvidia just reported Q4 FY2026 revenue of $68.13 billion, up 73.2% year over year, with Data Center revenue of $62.31 billion representing 91% of the total. Data Center Networking alone grew 263% year over year to $10.98 billion, signaling customers are buying the full NVLink stack, not just GPUs. Full-year free cash flow came in at $96.58 billion.

Q1 FY2027 guidance calls for ~$78 billion in revenue, explicitly excluding any Data Center compute revenue from China. The floor Caso describes already has a missing brick, and the building is still going up.

Huang framed demand clearly at GTC: “Enterprise adoption of agents is skyrocketing. Our customers are racing to invest in AI compute.” The partnerships backing that claim are concrete: OpenAI deploying at least 10 gigawatts of Nvidia systems, Anthropic committing to 1 gigawatt, and CoreWeave targeting more than 5 gigawatts of AI factories by 2030.

The Edge AI Counterargument

One question raised during Caso’s interview touched on whether Apple’s M5 chips and the broader shift toward on-device edge inference could reduce demand for centralized GPU compute. Attributing a position to him on that point would not be accurate, as the transcript did not capture his full response. It is a legitimate debate, but hyperscaler capex commitments already on the books suggest the data center build-out is not waiting for that question to resolve.

If Caso is right and $1 trillion is a starting point rather than a finish line, analysts still modeling 2027 off pre-GTC numbers are working with an outdated map. The order book is open, the products are shipping, and the customers are not slowing down.