Do not confuse a relationship story with an NVDA entry signal. The earnings print is the catalyst this week. The SpaceX color is real — exclusive silicon, a named Vera CPU customer, a 2027 orbital rack — and it is still color until it shows up as a disclosed customer percentage in a 10-Q. Partnership headlines are not revenue until they show up in a segment.
What Yahoo Finance reported
Pras Subramanian’s Yahoo Finance piece, “Nvidia’s relationship with SpaceX is deepening,” is the match. Nvidia’s fiscal Q2, which the same article recaps from the company release: $96.2 billion of revenue, up 106% year over year; Data Center $89.0 billion, up 117%. The SpaceX angle, Subramanian wrote, is tied to guidance and build-out plans rather than the reported totals. Nvidia does not break out customer-level revenue.
CFO Colette Kress on the call, as quoted: “We expect Vera to be deployed by every major hyperscaler, neocloud, AI lab, and system OEM, with shipments already underway to our lead partners, including [Oracle], SpaceXAI, and starting this quarter, [Amazon Web Services].” SpaceXAI adopting Vera CPUs to run agentic workloads behind Grok is the product claim. SpaceX and xAI president Mike Nicolls, in a statement Nvidia used: “Vera gives us the CPU performance and memory bandwidth to run enormous amounts of orchestration, code, and data processing while keeping GPUs doing what they do best.”
Deepwater’s Gene Munster, in a post on X that Yahoo embedded: “$SPCX accounted for about 5% of overall revenue. Last quarter, it was more like 3%. Looks like they moved $SPCX revenue from ACIE into Hyperscalers, which makes sense because SPCX is now a hyperscaler with 8GW coming online next year, same as $META and $AMZN.” Applied to $96.2 billion, Yahoo said that implies nearly $5 billion tied to SpaceX. That is Munster’s estimate. It is not Nvidia’s disclosure. Flag it as such. I am not going to upgrade an analyst’s tweet into a recognized-revenue line.
The orbital piece: the two companies confirmed a space-optimized Vera Rubin NVL72 rack-scale system for launch aboard SpaceX’s first-generation Starmind satellite in the fourth quarter of 2027, with larger scale in 2028. The AI1 satellite design carries a 120-kilowatt compute payload, peaking at 150 kilowatts. For Nvidia, SpaceX is a terrestrial customer and a launch vehicle for putting data-center silicon in orbit. For SpaceX, Nvidia is becoming the compute backbone.
What Elon Musk already said — separately
This permalink is the Subramanian earnings-week story, not the August 4 SpaceX earnings-call story. They are related. On SpaceX’s first public earnings call, Musk said the company would build AI services exclusively on Nvidia: “We think the Vera Rubin architecture is the best architecture. We think it’s the best AI computer, and we greatly value our close cooperation and partnership on many levels with Nvidia. So we’re exclusive to Nvidia.” He talked about ending the year with more than 2 gigawatts of compute and close to 10 gigawatts by the end of next year, and about launching Starmind satellites next year. Yahoo’s earlier wrap on that call also had SpaceX AI revenue at $2.6 billion, up 213% quarter over quarter and 247% year over year, plus multibillion-dollar leasing deals with Google and Anthropic. Those figures belong to the SpaceX print. Do not smuggle them into Nvidia’s $96.2 billion as if Nvidia reported them.
A later Yahoo recap said Nvidia had disclosed a $21 billion SpaceX stake (122.8 million Class A shares) via an SEC filing, its second-largest holding behind Intel. That is a separate filing story. I am not folding a $21 billion figure into this column as if it were in Subramanian’s relationship piece. If you care, read the filing.
California angle and the actual trade
Nvidia is Santa Clara. SpaceX is Hawthorne. This is a Northern California / Southern California supply chain that retail accounts already own twice if they hold NVDA and any space-or-Musk sleeve. “Deepens relationship” does not add a third reason to concentrate. California will tax the gain when you chase the press release.
Munster’s 5% is the only customer-mix number on the table, and it is his. Kress putting SpaceXAI on a list with Oracle and AWS is the company confirmation that the logo is real. Starmind in Q4 2027 is a date, not a 2026 revenue line. A 120 kW orbital payload is an engineering spec. It is not a TAM.
Caveats
Exclusive language from Musk is a customer preference, not a take-or-pay contract Nvidia filed. Launch slips. Radiation and thermal engineering for orbital compute are not solved because a press release used “NVL72.” The earnings print — $96.2 billion, $108 billion guide, 70% FY28 growth comment — is the week’s catalyst. This article is a customer vignette. Educational commentary, not investment advice. No entry off a partnership headline.
Source: Nvidia’s relationship with SpaceX is deepening (Pras Subramanian, Yahoo Finance); NVIDIA Q2 FY2027 release.
The Hedge 