Three years into the AI boom, Nvidia Corp (NASDAQ:NVDA) is still not talking about moderation. It’s talking about magnitude. That was the tone of a recent JPMorgan fireside chat with Nvidia’s CFO, Colette Kress, in which management described the compute-demand environment in one word: “tremendous.” Not cautiously optimistic. Not stable. Tremendous.
- Track NVDA stock here.
Nvidia’s AI Orders Are Still Building, Not Plateauing
Nvidia previously disclosed an over $500 billion backlog through the end of 2026, which is no longer a static number. While management didn’t update the headline figure, it confirmed that additional customer agreements — including OpenAI and Anthropic — were signed after that disclosure. JPMorgan estimates the existing backlog alone could translate into roughly $330 billion of data center revenue, suggesting visibility continues to improve, not fade.
What stood out was how far in advance customers are planning. Long lead times for AI data center buildouts mean Nvidia is already engaged with customers well ahead of deployment, and even for 2026, buyers are pushing to add to existing orders — a sign demand remains ahead of supply.
Networking Becomes A Lock-In Lever
Another quiet shift is happening alongside compute. Nvidia now sees a 90% attach rate for its networking products with AI deployments, up from around 80% last year.
That means nearly every new Nvidia-powered data center is buying into the full stack, reinforcing Nvidia’s position beyond just GPUs.
Physical AI Enters The Conversation
Beyond data centers, Nvidia is leaning harder into physical AI — systems that interact with the real world. Automotive remains a visible entry point, but management highlighted growing collaboration across multiple industrial verticals.
JPMorgan views this as a largely untapped opportunity where Nvidia’s end-to-end platform could again matter.
Risks Remain, But The Trend Is Clear
Licensing approval for H200 shipments to China remains pending, and any approval would be upside rather than base case. Meanwhile, Nvidia is preparing to juggle multiple platforms while transitioning to Rubin in late 2026.
Nvidia isn’t seeing demand cool — it’s seeing it stack. When orders keep building and attach rates keep rising, the AI story looks less crowded and more entrenched.
Photo: JRdes / Shutterstock
Recent Comments