A single regulatory decision could unlock upside for Nvidia Corp (NASDAQ:NVDA, XETRA:NVD). But the more important signal from this note is how tight supply, power chips and geopolitics are reshaping the entire semiconductor cycle.
Wedbush’s short daily note looks narrow on the surface. A delayed approval here, a better-than-feared outlook there. Look closer and it captures something much bigger about where global tech demand really sits in early 2026.
Nvidia is waiting on Washington, not demand
The immediate headline is familiar. Sales of Nvidia’s H200 AI accelerators to China are still awaiting final US government approval, according to the Financial Times. The State Department is reportedly considering additional guardrails to ensure shipments do not breach national security rules.
This is not new risk, but it is unresolved risk. And markets hate unresolved risk more than bad news.
Wedbush’s view is pragmatic. Deals that relax restrictions under the current administration tend to take longer than investors expect. The Middle East precedent looms large. But the endgame is still likely approval, not prohibition. China needs the product. Domestic alternatives remain constrained. Existing AI infrastructure is already built around Nvidia’s ecosystem.
Crucially, Wedbush has no China revenue in its Nvidia estimates. That leaves meaningful upside optionality if approvals come through, rather than downside risk if they do not.
The message is clear. This is a timing issue, not a demand issue.
Supply, not competition, is Nvidia’s real edge
Buried in the note is the more durable bull case. Nvidia has locked down component and materials supply through 2026 better than almost anyone else in the industry.
That matters because the next constraint in AI hardware is not GPUs. It is memory, power management and advanced packaging. Tight availability is already driving higher pricing, and companies with secured supply chains will capture disproportionate margin and share.
This is why Nvidia’s advantage increasingly looks structural rather than cyclical. Even if geopolitics slows one region, global demand can simply be rerouted.
Power management chips are the quiet bottleneck
The second half of the note shifts attention to mature foundries, and it is arguably the more revealing section.
VIS and UMC are seeing better-than-seasonal utilisation and gross margin trends, driven by demand for power management silicon. This is not consumer electronics froth. It is data centre infrastructure.
AI workloads are power-hungry. Every new rack needs layers of power regulation and control. That demand flows down the stack, benefiting older-node foundries that investors have largely written off as ex-growth.
Wedbush remains cautious near term, citing headwinds in consumer electronics and the risk that higher memory costs feed through into device prices and demand destruction. But the direction of travel is clear. Power is now a growth vector.
The next semiconductor cycle will look different
The note hints at a deeper shift. The next upcycle will not be led by smartphones or PCs. It will be driven by edge AI, data centres, physical AI systems and a delayed but powerful consumer refresh once costs stabilise.
That has implications for valuation. Mature foundries are no longer just cyclical laggards. They are becoming critical enablers of the AI buildout, even if the timing remains uneven.
What to take away from this
For Nvidia, the story is not about whether AI demand is real. That question is settled. It is about how quickly politics allows revenue to catch up with reality.
For the broader semiconductor space, this note reinforces an uncomfortable truth. The AI boom is now constrained by physics, power and supply chains, not enthusiasm or capital.
Investors looking only at headline GPU shipments risk missing where the real bottlenecks, and therefore the real pricing power, are forming.
In 2026, semiconductors are less about who designs the smartest chip, and more about who can actually ship the full system.