Nvidia's Two-Speed Market: A SpaceX Endorsement Meets Goldman's Caution

Nvidia secures SpaceX AI deal and ramps Vera Rubin, but Goldman warns of high expectations ahead of earnings. Key updates on demand, supply, and financing.

The countdown to August 26 has taken on an unusual texture for Nvidia shareholders. On one side sits a freshly announced exclusive partnership with SpaceX that reads like a technologist's dream; on the other, a Goldman Sachs warning that the stock's recent run has left virtually no room for error when earnings land.

SpaceX's decision to build its future AI compute infrastructure exclusively on Nvidia's Vera Rubin platform — targeting 6 to 10 gigawatts of capacity by the end of 2027 — carries weight well beyond the headline number. A company with the engineering pedigree and financial firepower of SpaceX could easily have pursued a multi-vendor strategy or developed in-house silicon. It chose not to. That single decision undercuts the narrative that customer-side diversification is eroding Nvidia's dominance anytime soon.

The operational signals around Vera Rubin are equally telling. Foxconn has confirmed the architecture has reached full production status, with shipments slated to begin in the fourth quarter of 2026. The next-generation platform is no longer a roadmap slide; it is moving through the supply chain. Meanwhile, Nvidia is accelerating development of the subsequent Feynman generation, targeted for late 2028, and confirmed the Rubin Ultra successor will carry 768 gigabytes of HBM4E memory — trimmed from the originally planned terabyte due to high-bandwidth memory supply constraints. The associated Kyber platform remains on schedule for the second half of 2027, with the company pushing back against speculation of a delay to 2028.

The demand picture extends well beyond marquee names. IBM's $240 million agreement with Together AI to deploy Nvidia Blackwell systems underscores that the customer base reaches deep into established enterprise IT, not just hyperscalers and aerospace pioneers. JPMorgan calculates that Alphabet, Amazon, Meta, Microsoft, and Oracle will collectively spend roughly $733 billion on capital expenditures this year, with Nvidia expected to capture about 26 percent of that outlay.

Geographic expansion adds another layer. A partnership with Firebird, Dell, and CoreWeave is building what would be the largest AI factory in the CIS region in Armenia, planned to house more than 70,000 Rubin and Blackwell GPUs with 300 megawatts of capacity by the end of 2027. The company has also begun series production of its Spectrum-X Ethernet photonics networking technology, which it says delivers five times the energy efficiency of prior approaches.

The financing architecture announced earlier this week — the $500 billion funding structure involving Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs, and KKR — has drawn its share of scrutiny. CEO Jensen Huang has clarified that Nvidia's direct financial commitment is capped at 25 percent per individual project, a self-imposed limit designed to address concerns about circular financing, where Nvidia would effectively be funding its own demand. Whether the capital partners ultimately mobilize at scale remains an open question, but the structure appears more thoughtfully constructed than a mere marketing exercise.

The institutional investor picture is decidedly mixed. Baillie Gifford significantly expanded its position in the second quarter of 2026, with Nvidia remaining its second-largest holding at 8.26 percent of the portfolio. Sands Capital Management trimmed by roughly 22 percent over the same period, though the stock still represents its largest single position at 10.45 percent. Hilltop Holdings cut its stake by more than half in the first half of the year. Cathie Wood's Ark Invest, by contrast, bought $59.9 million worth of Nvidia shares across five ETFs over three trading days in July and August, boosting its total position by 24 percent to over $303.6 million.

These moves look less like a verdict on fundamentals than portfolio management at a stock that has climbed sharply. The shares closed Thursday at 195.34 euros, roughly 3.5 percent below the 52-week high of 202.50 euros reached in May. The distance to the 200-day moving average of 168.02 euros stands at 16 percent, suggesting the longer-term uptrend remains intact even if near-term valuation has cooled. The 6.6 percent rally over the past five trading days has pushed the stock 8.8 percent above its 50-day average of 179.46 euros — precisely the kind of momentum that makes Goldman Sachs cautious.

The broker's concern is straightforward: with the bar set so high, even solid results could trigger a sell-the-news response. The consensus calls for second-quarter revenue of roughly $92 billion, up about 96 percent year over year, with earnings per share of $2.08. Nvidia's own guidance was $91 billion. The first quarter delivered revenue of $81.615 billion, a gain of 85 percent, with the data center business growing 92 percent to $75 billion.

The real test arrives on August 26. Between now and then, the SpaceX partnership stands as the strongest argument for the bulls — evidence that technological leadership remains intact where it matters most. But Goldman's caution serves as a reminder that in a market this extended, the gap between meeting expectations and exceeding them can be the difference between a new high and a sharp pullback.