Nvidia’s stock is caught in a tug-of-war between two powerful and opposing forces. On one side, the company has just booked over $30 billion in revenue from sovereign AI projects — state-built compute infrastructure that locks in customers for years. On the other, a 31% plunge in GPU rental rates over three weeks suggests the supply of AI computing power is racing ahead of demand. The result is a stock trading 15% below its 52-week high, even as the company posts record numbers.
The market’s indecision is written into the technicals. At €171.18, the share sits about 6% below its 50-day moving average with an RSI around 41 — not panic territory, but clearly not conviction either. Analysts see a 53% upside to their consensus target of €262.18, yet price action since Nvidia’s May 20 earnings report has been stubbornly negative. That day, despite a blowout quarter, the stock slipped 1.77% and has not regained its pre-report level.
The sovereign AI engine
The structural shift in Nvidia’s customer base is hard to overstate. In its 2026 fiscal year, revenue from sovereign AI — governments building their own national AI infrastructure — surpassed $30 billion, accounting for roughly 14% of total sales and tripling year-over-year. Nvidia is no longer just selling GPUs to hyperscalers; it delivers pre-configured “AI factory” reference architectures that can go live in about 90 days. Europe alone has 20 such projects underway: 18,000 Grace-Blackwell systems in France, £1 billion earmarked for UK research capacity by 2030, and Germany’s first industrial AI cloud built with Deutsche Telekom.
The switching costs are prohibitive. Nvidia’s CUDA software stack includes more than 400 optimized libraries, and any government that builds its national AI systems on top of that foundation would face a massive rewrite to shift to AMD or Intel. This is by design — and it gives Nvidia a durable annuity from state customers who, unlike hyperscalers, have no interest in switching for marginal cost savings.
The GPU rental signal
Yet while sovereign AI orders boom, the market for on-demand GPU compute is softening dramatically. The hourly rental rate for Nvidia’s B200 GPU fell from $6.11 on May 30 to $4.22 by June 21 — a 31% slide in three weeks. That suggests AI compute supply is flooding in faster than new workloads are consuming it. When rental prices drop, Nvidia’s pricing power erodes, and the assumption of endless demand is tested. Analysts at D.A. Davidson have noted that Nvidia and memory maker Micron are trading as if the AI cycle has already peaked.
The company’s reported first-quarter numbers for fiscal 2027 were superb by any measure: revenue of $81.6 billion (up 85% year over year), datacenter revenue of $75.2 billion (beating the $73.5 billion consensus), and a second-quarter guidance of $91 billion that also topped expectations. Management slapped an $80 billion stock buyback on top and raised the quarterly dividend to $0.25 per share. None of that was enough to push the stock higher.
Should investors sell immediately? Or is it worth buying Nvidia?
China’s structural void
The other persistent drag is China. The US Commerce Department closed another loophole in early June 2026, blocking Nvidia’s Rubin and Blackwell chips from reaching Chinese AI firms via offshore subsidiaries. While less dramatic than the H20 restrictions of 2025, the direction is unmistakable. Jensen Huang acknowledged on May 21 that Nvidia has largely ceded the advanced AI chip market in China to Huawei. That market once accounted for at least a fifth of Nvidia’s datacenter revenue; JPMorgan and Bernstein put the cumulative revenue hit from export controls and Chinese competition at $5.5 billion to $16 billion in the current fiscal year. China’s share is now essentially zero, and Nvidia’s Q1 guidance for fiscal 2027 explicitly excludes any return.
Balancing the scales
The question investors are grappling with is whether sovereign AI can fill the China-sized hole. So far, the evidence is mixed. The state customers aren’t buying cut-down versions — they are purchasing the same GB200-NVL72 infrastructure that hyperscalers use, which supports both margins and average selling prices. The customer base is now roughly half hyperscaler, half government and enterprise, a notably more diversified and sticky mix than a year ago.
Still, the market is pricing in uncertainty. The 53% gap to the analyst consensus shows that the full sovereign AI scenario is not yet reflected in the share price. With a market capitalisation of around €4.1 trillion, Nvidia trades at a hefty premium that leaves little room for disappointment.
The next catalyst is Vera Rubin, Nvidia’s next-generation chip architecture, due to enter production in the second half of 2026. AWS, Google Cloud, Microsoft Azure, and CoreWeave are already lined up as early customers, and a new multi-year memory partnership with SK hynix will support the broader AI ecosystem. The analyst community remains overwhelmingly bullish: 48 of 61 analysts rate the stock a “buy”, ten say “outperform”, and only one recommends selling. Their mean price target has risen from roughly $269 in April to about $300 in June.
For the stock to regain momentum, two things need to happen: the B200 rental rate must stabilise, and Vera Rubin must launch on schedule. If both fall into place, the current weakness may look like a buying opportunity. If not, the gap between record revenue and a hesitant stock price could widen further.
Ad
Nvidia Stock: Buy or Sell?! New Nvidia Analysis from June 30 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 30.
Nvidia: Buy or sell? Read more here...
