Nvidia’s New AI Factory Model and Rubin-Ultra Timeline Propel Stock Higher

Nvidia pivots from hyperscaler chip sales to financing smaller cloud providers, confirms Rubin-Ultra for 2027, and pushes into CPUs amid a customer diversification drive.

Nvidia is rewriting its playbook on two fronts simultaneously. While CEO Jensen Huang used a Morgan Stanley investor event over the weekend to put to rest rumors of a delay for the next-generation Rubin-Ultra platform — confirming a 2027 launch — the company is also quietly overhauling how it makes money. Since July 1, 2026, Nvidia has been shifting away from selling chips to a small group of hyperscaler customers and toward a model the company calls “AI Factories,” in which it finances smaller cloud providers directly, extends credit lines, and takes a share of future revenue. The result is a broader, more durable revenue base that analysts are already betting heavily on.

The strategy comes as the customer mix undergoes a structural change. A Citi analysis from mid-July notes that the strongest growth impulses are now coming from specialized AI labs, government programs, and enterprises building their own data centers — not the large cloud vendors that once drove Nvidia’s trajectory. The rise of “sovereign AI,” where nations invest in domestic computing infrastructure independent of U.S. cloud giants, is acting as a buffer against cyclical downturns. That diversification is timely: custom chips from competitors are expected to lift their market share from 20.9% in 2025 to 27.8% in 2026. By locking in recurring demand from a wider set of partners, Nvidia can better absorb that competitive pressure.

Product development remains on track. Huang reiterated that the Rubin-Ultra platform will hit the market in 2027, while the intervening Vera-Rubin architecture — a successor to Blackwell — is slated to start production in the second half of 2026. The shift in computing focus is also tangible: Nvidia calls this the “year of inference,” as real-time data generation now accounts for an estimated 80% to 90% of data-center workloads. The new chips are optimized for efficiency per watt, not brute force, and feature closed-loop water cooling that all but eliminates on-site water consumption — a direct response to the energy and water strains of massive AI facilities.

Beyond GPUs, Nvidia is pushing into the CPU market with its Vera platform, targeting roughly $20 billion in standalone CPU revenue for the current fiscal year. Wedbush analysts see this as more than a side product, arguing it opens up a market well beyond graphics chips. According to the company, Vera delivers 1.8 times the performance of comparable x86 processors. Meanwhile, Huang announced a sharp increase in Taiwan investment, lifting annual spending in the country from $100 billion to $150 billion, partly to fund a “Constellation” campus in Taipei that will house up to 4,000 employees by 2030.

The stock has caught the updraft. Shares closed Friday at €184.60, gaining nearly 8% over the previous week and recovering roughly $393 billion in market capitalization in just two trading days. Nvidia’s market cap now stands at around €4.475 trillion after the rally, compared with roughly €4.3 trillion earlier in the week. While the price remains 8.84% below its 52-week high of €202.50 from mid-May, it sits comfortably above both its 50-day moving average of €181.22 and its 200-day average of €164.78. The relative strength index of 58.6 suggests room to run before entering overbought territory.

Valuation remains reasonable given the earnings momentum. With a price-to-earnings ratio estimated between 18 and 23 — below the S&P 500’s current P/E of over 20 — the business fundamentals provide a solid floor. Revenue surged 85% year-over-year to $81.6 billion in the first quarter, and the company expects around $91 billion in the second quarter. An $80 billion buyback program has been approved, and the quarterly dividend was recently raised to $0.25 per share.

Analysts see more upside. The average price target stands at €264.16, implying a potential gain of 43.1% from current levels. That optimism is pinned not just on new chips but on the belief that sovereign AI projects in India, Brazil, and the U.S. will drive the next growth wave. Yet the new financing model carries its own risk: Nvidia is now exposed to the creditworthiness of its cloud partners. If some stumble, the company bears that risk directly. The second half of 2026, with the Rubin ramp and the deepening reliance on inference workloads, will test whether this twin transformation — financial engineering and a confirmed product roadmap — can sustain the rally.