Nvidia Streamlines Rubin Design While China Revenue Sinks into $8 Billion Hole

Nvidia redesigns its flagship GPU mid-production, faces an $8 billion China revenue hole from export bans, yet datacenter demand and price hikes support margins. Stock trades 10% below last month's high.

Investors in Nvidia are having to digest two very different stories at once. On one hand, the company is reshaping its next-generation flagship GPU on the fly to sidestep manufacturing snags. On the other, its once-promising China business is collapsing under the weight of geopolitics, with a potential $8 billion revenue hole opening up this quarter. The shares are caught in the crossfire, trading near €174 and roughly 10% below last month’s level.

The change to the Rubin Ultra architecture is a radical mid-production pivot. Originally planned as a four-chip design, Nvidia is now moving to a dual-die configuration because of difficulties with advanced packaging technology. The new version uses eight memory modules instead of sixteen, halving per-chip performance. The company intends to compensate by arranging the chips in special layouts on server motherboards. Despite the setback, SemiAnalysis analysts believe datacenter revenue could beat market expectations by 20% in the second half of fiscal 2027, as earlier memory-module shortages have been resolved. Mass production of the Vera-Rubin platform began in early June, with eight major cloud providers receiving first shipments this autumn.

That supply tightness is feeding through to pricing. Amazon Web Services is hiking tariffs for certain compute capacity by 20% starting July 1, while rental rates for Nvidia’s H100 chips have climbed roughly 40% since last October — a boon for Nvidia’s margins even as its own product design gets reworked.

The China picture is far darker. Nvidia’s H20 chip, already subject to US export licensing since April 2025, is now effectively unsalable. Washington demands 25% of proceeds, and Beijing has directed local firms to steer clear. The result: a $4.5 billion charge in the first quarter of fiscal 2026 for excess H20 inventory, and management forecasts zero H20 shipments to China in the current quarter — an $8 billion revenue hit. In fiscal 2025, China still accounted for 13% of total sales, or $17.11 billion. That share is now evaporating. Rivals such as Huawei, with clients like DeepSeek, are filling the vacuum, embedding their own hardware into Chinese data centers while Nvidia is locked out.

Nvidia CEO Jensen Huang has drawn a hard line: no support or repairs for chips smuggled into sanctioned markets. It’s a principled stance, but it underscores the company’s powerlessness in the US-China standoff.

Elsewhere, Nvidia is expanding its global footprint. A new partnership with Firmus Technologies will build an AI factory in Indonesia, with up to 170,000 Nvidia accelerators expected to go online by 2028. Firmus projects customer contracts worth as much as $30 billion over the first six years — a sign that sovereign AI demand remains robust.

On the technical side, the stock is holding above its 200-day moving average at €163.91, a key support level. However, it slipped below the 50-day line in late June, and the distance from the May record high of €202.50 is widening. Analysts’ average price target sits at €262.17, implying substantial upside if the China stalemate eases. Nvidia is expected to generate over $96 billion in free cash flow this year, half of which will be returned to shareholders via buybacks and dividends. The datacenter segment nearly doubled revenue to $75.2 billion in the first quarter of fiscal 2027, and upcoming cloud price increases should further underpin margins.

For now, the China trap is a drag on momentum rather than an existential threat. But with both Washington and Beijing showing no urgency, investors may need patience before the next leg higher materializes.