The narrative around Nvidia has long been pinned to the cap-ex whims of a handful of US hyperscalers — Microsoft, Amazon, Alphabet, Meta. If they turned off the spending tap, the argument went, the GPU king would be left high and dry. That thesis is being rewritten, quietly, by an entirely new class of buyer: sovereign states.
Governments in Britain, Japan, and Saudi Arabia are now treating artificial intelligence compute capacity as a strategic national asset, akin to energy grids or water systems. And they are putting real money behind the idea. In fiscal 2026, Nvidia’s revenue from sovereign AI hit over €30 billion — a threefold jump from the prior year and roughly 14% of total sales. The company’s CFO has signaled the segment will grow at least as fast as the broader AI infrastructure market.
The structural significance here isn’t just the volume. It is the buyer profile. Corporations invest with an eye on quarterly returns; governments invest for long-term survival. A sovereign mandate that spans decades is far harder to cancel than a corporate budget line. The classic bear case — that a disappointing earnings report from Microsoft could crater Nvidia’s order book — loses some of its sting when a growing share of orders come from ministries with multi-year infrastructure commitments.
The physical footprint is expanding fast. Nvidia is building 20 AI factories across Europe, five of them at gigafactory scale. In Germany, the world’s first industrial AI cloud — powered by 10,000 Blackwell GPUs — has gone live, operated by Deutsche Telekom. In France, Mistral AI secured €830 million in institutional loans to buy roughly 13,800 Nvidia chips and construct a major data center near Paris, the largest private infrastructure bet on sovereign AI in Europe to date. Asia is no different: Indian infrastructure giant Larsen & Toubro is working with Nvidia on a gigawatt-scale network of AI data centers, explicitly designed to keep sensitive data and model training on home soil.
Industry demand is equally robust beyond the state sector. Roche, the Swiss pharmaceutical group, recently brought its own AI factory online, powered by more than 3,500 Blackwell GPUs in a facility that cost an estimated half a billion dollars. The payoff has been immediate: the Roche subsidiary Genentech now designs oncology molecules 25% faster, and the company simulates drug production digitally using Nvidia’s Omniverse platform.
Should investors sell immediately? Or is it worth buying Nvidia?
Yet none of this spared the stock from a brutal week. On Friday, a wave of inflation anxiety and cautious competitor forecasts triggered a $1.2 trillion rout across the entire AI sector. Nvidia’s shares sank to €168.80, a weekly loss of more than 7%. The stock now sits nearly 17% below its all-time high set in mid-May — even as the underlying business churns out record after record.
Operationally, Nvidia is flexing muscle in unexpected places. In the first quarter of 2026, the company grabbed a 21.5% share of the data-center Ethernet switch market, a once-traditional networking stronghold. That segment alone generated $2.1 billion. Total networking revenue tripled to $15 billion, supercharged by 800-gigabit switches. The move into broader Ethernet compatibility opens a vast new addressable market beyond GPUs.
To feed the insatiable appetite for high-margin AI chips, management has made hard choices. Production of gaming graphics cards has been cut by as much as 30% as supply chains remain stretched. That strategic pivot underscores where the real money lies — and where the risks sit.
Sovereign AI has its own weak points. Government procurement cycles are slow. Infrastructure projects face political budgets, and a change of administration or a fiscal squeeze can delay or even kill commitments that a corporate buyer would treat as firm. If the wave of sovereign AI projects fails to produce tangible economic returns within two to three years, a pullback in state spending is a real possibility.
For now, the technical picture offers some hope. The RSI has dipped to 38.2, signaling oversold conditions. The 200-day moving average sits at €163.66, a line in the sand that, if it holds, could allow the long-term uptrend to reassert itself. Whether that support stands depends on whether sovereign AI matures from a compelling narrative into a durable, decade-long procurement cycle — one no single hyperscaler can derail with a weak quarterly report.
Ad
Nvidia Stock: Buy or Sell?! New Nvidia Analysis from June 29 delivers the answer:
The latest Nvidia figures speak for themselves: Urgent action needed for Nvidia investors. Is it worth buying or should you sell? Find out what to do now in the current free analysis from June 29.
Nvidia: Buy or sell? Read more here...
