Nvidia Caught Between China’s Huawei Pivot and a UK Sovereignty Deal

Jensen Huang’s AI empire is facing a rare two-front challenge. In China, a former Nvidia customer has rebuilt its software stack from scratch to run on Huawei chips. In Britain, a new partnership is locking Nvidia hardware into government-approved infrastructure. The two stories, unfolding simultaneously, tell a tale of a company that remains dominant but is no longer uncontested.

DeepSeek Abandons Nvidia’s CUDA Fortress

The Chinese AI startup DeepSeek has unveiled its V4-Pro and V4-Flash models, and the technology stack is a bombshell. The models run entirely on Huawei’s Ascend 950 chips, orchestrated through the company’s Supernode clustering technology. This is a radical departure from the R1 predecessor, which was trained on Nvidia hardware.

DeepSeek rewrote its core code over recent months to work with Huawei’s CANN framework, effectively abandoning Nvidia’s CUDA ecosystem. That software moat had long been considered a second American lever of control over global AI development. The cracks in that wall are now clearly visible.

The V4-Pro model delivers top-tier performance in programming and mathematics. DeepSeek claims its general knowledge lags only a few months behind Google’s Gemini 3.1-Pro. Morningstar analysts see the release as confirmation that the market has already priced in China’s ability to produce competitive, cheaper AI without Western chips.

Nvidia’s China Shipments Remain Frozen

The timing could hardly be worse for Nvidia. The US government approved sales of the powerful H200 chips to China back in January. Huang restarted production. Yet according to insiders, not a single chip has reached a Chinese buyer. Disagreements over sales terms have blocked deliveries entirely.

The H200 units sit in warehouses. Meanwhile, Beijing is quietly building out its own AI infrastructure with domestic chips, undisturbed by US export policy. Chinese semiconductor stocks such as Huahong Semiconductor surged 15 percent on the DeepSeek news. Nvidia’s stock currently trades at €171.56, holding a modest year-to-date gain, but its roughly 55 percent share of the Chinese AI accelerator market is now under serious threat.

A British Counterweight: BT and Nscale

Across the globe, a very different dynamic is playing out. BT Group and Nvidia-backed Nscale have announced a partnership to build sovereign AI services on British soil. The plan calls for 14 megawatts of capacity spread across three existing BT network sites. Nscale provides GPU cloud services; BT supplies network connectivity and power.

Should investors sell immediately? Or is it worth buying Nvidia?

The goal is to let government agencies and companies run AI workloads that must remain within UK borders for regulatory reasons. Nscale is investing roughly $2.5 billion into the UK market over three years. That includes a 50-megawatt campus in Loughton, Essex, originally scheduled for late 2026 but now pushed to the second quarter of 2027. The delay stems from the need to install Nvidia’s newest Vera Rubin 200 technology. Nscale has secured around 200,000 units of the Blackwell-generation GB300 for its clusters.

Nvidia Stock Finds Support

The BT news arrives during a strong stretch for Nvidia shares. The stock trades at €171.30, roughly 11 percent above its 30-day average and about 9 percent above the 200-day moving average. Over the past twelve months, the share price has gained more than 83 percent.

The broader semiconductor supply chain is also providing tailwinds. TSMC reported a 58 percent profit jump in the first quarter, fueled by sustained demand for AI hardware. ASML raised its 2026 revenue outlook to between €36 billion and €40 billion.

The Sovereignty Trend Accelerates

The BT-Nscale deal reflects a broader shift: governments and corporations want to control AI infrastructure rather than hand it to external providers. Similar projects are emerging across Europe. The SUSE AI Factory in Prague, for example, launched on April 21, 2026, as a turnkey open-source platform running on Nvidia hardware.

Nvidia itself sees the potential as enormous. The company estimates the combined revenue opportunity for its Blackwell and Vera Rubin platforms could hit $1 trillion by 2027. For the first quarter of fiscal 2027, Nvidia is targeting roughly $78 billion in revenue, a figure that beat earlier market expectations. Deals like the one with BT ensure Nvidia’s hardware remains embedded in markets where data sovereignty is not a preference but a legal requirement.

The contrast is stark. In China, Nvidia’s hardware is being systematically replaced. In Europe, it is being deliberately locked in. For Jensen Huang, the challenge is to keep the second story growing fast enough to offset the first.