Nvidia's China Chip Win Backs a $1 Trillion AI Revenue Call

Beijing's approval for Nvidia to sell its H200 AI accelerators in China arrived Wednesday like a structural reset for the company's revenue trajectory, landing in the middle of GTC 2026 just as CEO Jensen Huang put a number on the table that the market has been trying to model ever since Blackwell began shipping: $1 trillion in cumulative revenue from Nvidia's AI architectures through 2027.

NVDA shares climbed in midday trading while much of the tech complex held mixed. Wedbush Securities raised its price target to $500 from $320 — a 56% increase — noting that the China approval changes the addressable market calculation in ways that weren't priced into any previous model. RBC Capital maintained its outperform rating with a $525 target, pointing to accelerating high-bandwidth memory orders as evidence that Nvidia's pipeline is converting into deployments at scale.

The approval is significant not simply because of the market size but because of what it signals about the negotiated framework behind it. U.S. export restrictions had progressively tightened from 2022 through 2023, effectively making Nvidia's most advanced chips unavailable in China. The H200 approval suggests a middle-ground arrangement has been reached — one that opens China for Nvidia's current-generation chip while the company is already moving on to Vera Rubin.

Monday's keynote unveiled the Vera Rubin platform — GPUs and CPUs in a tightly integrated architecture designed for the inference-heavy workloads that now dominate enterprise AI spending. The NVL72 Supercomputer, built from 18 DGX Rubin NVL72 racks with 1,296 GPUs and 640 terabytes of HBM4 memory, delivers 3.6 exaflops of inference performance. Nvidia is building this for Anthropic, OpenAI, Meta, and their equivalents globally — the China approval means the list now includes Alibaba, Baidu, and ByteDance again.

Also announced at GTC was the Groq-3 LPU and a Groq architecture variant developed specifically for the Chinese market. The strategic logic: if China's AI buildout proceeds regardless of U.S. policy, Nvidia wants to own the infrastructure layer it runs on.

Competitive implications are immediate. AMD shares dipped about 1.2% even as the Nasdaq held gains — the market treating Nvidia's China return as direct market share recapture from alternatives AMD had been selling into the gap. Chinese AI developers weren't loyal to AMD's MI300X; they were buying it because it was the best available option. The H200 is a better option, and the CUDA software ecosystem makes switching costs prohibitively high. Intel's Gaudi series faces the same headwind in a geography where it had recently found footing.

The $1 trillion projection Huang put on the table spans 2025 through 2027 and requires AI infrastructure investment to continue accelerating without meaningful interruption. The Iran war and its oil shock have stalled parts of the economy, but enterprise AI capital expenditure programs — committed to multi-year spending plans — have shown no sign of deceleration in any hyperscaler earnings report this quarter. The next hard catalyst for NVDA is Friday's analyst session, where the company is expected to provide formal Vera Rubin shipment timelines and initial China H200 order pipeline guidance. Any quantification of China revenue will move the stock immediately.