Nvidia's Great Rebalancing: China's Exit, Sovereign Surge, and the Vera Gambit

Despite losing $4.6 billion in China sales, Nvidia finds a new growth engine in sovereign AI infrastructure, driving a 42% share price rise over the past year.

The stock sits at €177.96, practically glued to its 50-day moving average of €177.32. The RSI reads a neutral 46.3 — no panic, no euphoria. On the surface, it looks like a textbook consolidation. But beneath that flat line, Nvidia is undergoing a structural transformation that rewires its entire demand base.

The $4.6 Billion Hole That Didn't Sink the Ship

The numbers are brutal. In the latest quarter, Nvidia shipped precisely zero Data Center Hopper products to China, versus $4.6 billion in the same period a year earlier. That is not a decline — it is a complete evaporation. CEO Jensen Huang acknowledged that Nvidia's market share in China has effectively fallen to zero. US export controls and Beijing's push for technological self-reliance have converged to seal that market.

Even where licenses exist, not a single H200 chip has been delivered to a Chinese customer. And in early June, Washington tightened the screws further: the US Commerce Department clarified that licensing requirements for AI chips now apply to any company with a Chinese parent, regardless of where its subsidiaries operate. One industry source estimates that before that clarification, hundreds of thousands of chips may have flowed to China-linked entities through third countries like Malaysia.

Yet despite losing one of the world's largest single markets, Nvidia's shares are up 42% over the past twelve months. That paradox is the key to understanding the company's current trajectory.

The Sovereign Fill

What fills the void is not another country — it is the concept of the nation-state itself. Sovereign AI — infrastructure that lets countries build and control their own artificial intelligence — generated $30 billion in revenue in fiscal year 2026, more than triple the prior year. Growth exceeded 80% year-over-year, spread across nearly 40 countries including Canada, France, the Netherlands, Singapore and the United Kingdom.

These economies represent roughly $50 trillion in combined GDP. The logic is straightforward: any government that does not want to hand its digital future to American cloud giants needs its own AI stack — and that stack runs on Nvidia hardware. In South Korea, NAVER is expanding its sovereign AI infrastructure to an initial 55 megawatts with ambitions to reach gigawatt scale. The UK is targeting operational AI factories by the end of 2026 under national sovereignty goals.

Nvidia's CFO expects the segment to grow at least in proportion to each country's GDP — structurally, not cyclically. The company is so confident in this shift that it has reorganized its reporting structure. Going forward, Data Center will be split into Hyperscale (which still accounts for roughly half of data center revenue) and a new category called ACIE — AI Clouds, Industry and Enterprise. One year ago, hyperscalers dominated. Today, the other half comes from a diversified mix including state-backed buyers.

The Vera Counter-Move

While governments fill the revenue gap, Nvidia is also trying to re-enter China through a different product door. The company has informed Chinese customers that its new "Vera" central processor for AI data centers could be available as early as August, and orders are welcome. At least one major Chinese cloud provider is planning to order more than 300 servers, each equipped with two Vera CPUs — though the final order depends on test results.

Vera is not a workaround chip. It marks Nvidia's formal entry into the general-purpose processor market, sitting alongside its own Rubin GPU platform. It orchestrates and manages workflows in agentic AI systems. Globally, the product is already gaining traction: Nvidia reported nearly $20 billion in revenue visibility from the CPU segment alone for the current fiscal year.

But China's government is actively pushing local providers and advising against purchases of Nvidia data center products — even those designed specifically for the Chinese market. Huang still counts China inside his long-term $200 billion CPU opportunity, a signal that he sees structural demand there despite the headwinds. The real verdict will come in the second half of 2026.

Record Earnings and a Waiting Market

None of this geopolitical maneuvering obscures the core business. Quarterly revenue hit an all-time high of $81.6 billion, up 85% year-over-year. Data center revenue alone reached $75.2 billion, a 92% surge driven by the ramp of the Blackwell-300 product line. Hyperscalers still account for roughly half of that, with the rest coming from a growing mix of AI cloud providers, industrial customers, enterprises and governments.

Jensen Huang projected on the last earnings call that hyperscaler AI capital expenditure alone — from Alphabet, Amazon and others — could reach $3–4 trillion. That does not include neoclouds or other segments.

The stock is trading about 12% below its 52-week high of €202.50 set in May. The RSI at 46.3 signals no clear direction. Analyst consensus still sees a target of €258.74, implying roughly 45% upside from current levels. The 200-day moving average sits 9% below the current price, while the stock has pulled back about 8% over the past 30 days. Annualized volatility of nearly 42% is a reminder that this is no steady hand.

Nvidia is caught between a lost market and a newly drawn map of demand. The pause on the 50-day line may be exactly what it looks like: a market catching its breath before pricing in a world where every sovereign state is a potential customer — and China is an open question.