This note separates what the company reported from what the evidence can support today, then defines the operating signals that should confirm or challenge the thesis.
Thesis status
Active investigation · Preliminary view · Rating not assigned
SpaceX and xAI have disclosed a rapid expansion of NVIDIA-based compute capacity. That makes them a potentially material incremental demand channel for NVIDIA, but not yet a quantifiable revenue contribution. The decisive question is whether disclosed processor deployments and planned capacity translate into a time-bounded stream of NVIDIA systems revenue before internally designed silicon takes a meaningful share of future workloads.
The direction is supported by primary evidence. The size and timing are not. This memo therefore treats SpaceX and xAI as a catalyst to measure, not as revenue already earned or contracted.
The central question
Can SpaceX and xAI's AI-capital-spending acceleration create a measurable, incremental NVIDIA revenue stream large enough to extend NVIDIA's growth cycle before SpaceX and Tesla shift meaningful workloads to internally designed processors?
Answering that question requires a bridge from product deployment to recognized revenue:
processor quantities by product family × defensible NVIDIA content per processor or system × delivery and revenue-recognition timing, plus separately supported NVIDIA CPU and networking content
Land, buildings, power, cooling, storage, construction and non-NVIDIA equipment must remain outside that bridge. Total AI infrastructure spending is not NVIDIA revenue.
Why the question is material
NVIDIA reported fiscal Q1 2027 revenue of $81.615 billion, up 20% sequentially and 85% year over year. Data Center revenue reached $75.246 billion, or approximately 92.2% of total revenue by author calculation. Data Center grew 21% sequentially and 92% year over year.
At that scale, a new customer must represent more than a large headline capital budget to move the thesis. It must add systems demand that is incremental to current expectations, delivered on a defined schedule, and large enough to matter relative to NVIDIA's existing Data Center base.
SpaceX's June 5, 2026 prospectus supplies evidence that this possibility is real. It reports the following deployed and planned capacity:
| Facility or phase | Company-reported processor count | Compute power | Status |
|---|---|---|---|
| COLOSSUS first cluster | Approximately 100,000 H100 processors | Approximately 130 MW | Online |
| COLOSSUS II first cluster | Approximately 110,000 GB200 processors | Approximately 210 MW | Online |
| COLOSSUS II second cluster | Approximately 110,000 GB300 processors | Approximately 220 MW | Online |
| COLOSSUS II next phase | At least 220,000 additional GB300 processors | More than 400 additional MW | Expected once fully operational |
SpaceX states that COLOSSUS and COLOSSUS II collectively provide approximately 1.0 GW of compute power. The named H100, GB200 and GB300 processors establish direct NVIDIA product use. The next phase, if completed as described, would double the disclosed GB300 count at COLOSSUS II.
This is meaningful product evidence. It is not an order ledger. The prospectus does not disclose NVIDIA's realized revenue, purchase prices, order dates, delivery schedule, gross margin, complete networking content or the share of future workloads that could move to internal silicon.
Capital spending confirms acceleration, not supplier revenue
SpaceX reports the following capital expenditures:
| Period | AI capital expenditures | Total capital expenditures | AI share, author calculation |
|---|---|---|---|
| Q1 2026 | $7.723B | $10.107B | 76.4% |
| FY2025 | $12.727B | $20.737B | 61.4% |
| FY2024 | $5.633B | $11.163B | 50.5% |
Q1 2026 AI capital expenditures increased $5.156 billion, or approximately 201%, from $2.567 billion in Q1 2025. FY2025 AI capital expenditures increased $7.094 billion, or approximately 126%, from FY2024.
The prospectus says AI capital expenditures supported terrestrial compute infrastructure including data centers, processors, networking equipment and power infrastructure. Those categories explain why the total cannot be assigned to NVIDIA. The disclosed acceleration strengthens the demand question while leaving the supplier-revenue bridge unresolved.
NVIDIA's operating base
NVIDIA's Q1 FY2027 results show that the company enters this investigation from a position of exceptional growth and profitability:
| GAAP metric | Q1 FY2027 | Sequential change | Year-over-year change |
|---|---|---|---|
| Revenue | $81.615B | 20% | 85% |
| Gross margin | 74.9% | down 0.1 percentage points | up 14.4 percentage points |
| Operating expenses | $7.621B | 12% | 52% |
| Operating income | $53.536B | 21% | 147% |
| Net income | $58.321B | 36% | 211% |
| Diluted EPS | $2.39 | 36% | 214% |
Net income was unusually elevated by investment gains. NVIDIA's Form 10-Q reports $16.367 billion of total other income, net, driven primarily by $13.4 billion of unrealized gains on publicly held equity securities and $2.6 billion of unrealized gains on non-marketable equity securities. The resulting net margin should not be treated as a normalized operating margin.
The 10-Q also shows concentration risk that the earnings release alone does not capture. Three direct customers represented 21%, 17% and 16% of Q1 FY2027 revenue, respectively. NVIDIA additionally estimated that one AI research and deployment company contributed a meaningful amount of revenue by purchasing cloud services from NVIDIA customers. A large SpaceX or xAI build could be incremental while still increasing dependence on a small group of direct and indirect demand sources.
Management guided Q2 FY2027 revenue to $91.0 billion, plus or minus 2%, while assuming no Data Center compute revenue from China. That is a material geographic and policy constraint, but the available disclosure does not quantify the revenue that might otherwise have been recognized from China.
Balance sheet and cash generation
As of April 26, 2026, NVIDIA reported $50.335 billion of cash, cash equivalents and marketable debt securities, compared with $49.670 billion at the prior fiscal year-end. It also held $30.2 billion of marketable equity securities. Net cash provided by operating activities was $50.344 billion in the quarter.
NVIDIA repurchased 108 million shares for $20.2 billion during Q1 FY2027 and paid $243 million in cash dividends. Its board subsequently approved an additional $80.0 billion repurchase authorization and increased the quarterly dividend from $0.01 to $0.25 per share.
These figures support NVIDIA's capacity to fund product development and capital returns. They do not reduce the need to determine whether a SpaceX and xAI demand channel is incremental, durable and margin-accretive.
What would confirm the thesis
The thesis strengthens if primary disclosures establish several of the following:
- The planned 220,000-plus additional GB300 processors enter service on a disclosed schedule.
- NVIDIA or SpaceX discloses system orders, delivery timing or architecture mix sufficient to build a time-bounded revenue estimate.
- NVIDIA content extends beyond accelerators into separately supported CPU, NVLink or networking revenue.
- SpaceX and xAI capacity continues expanding after the current COLOSSUS II phase instead of representing a one-time build.
- NVIDIA's Data Center growth remains strong without material deterioration in gross margin as Rubin deployments scale.
What would weaken or break the thesis
The thesis weakens if:
- The planned COLOSSUS II expansion is delayed, reduced or changes architecture materially.
- SpaceX or Tesla processors take a meaningful share of the workloads assumed to use NVIDIA systems.
- Disclosed NVIDIA purchases are primarily replacements or demand already embedded in existing customer expectations rather than incremental orders.
- The deployment spans enough years that its annual NVIDIA contribution is immaterial relative to the Data Center revenue base.
- Greater customer concentration or unfavorable product mix weakens revenue durability or gross-margin economics.
The thesis breaks as a material near-term NVIDIA catalyst if primary evidence shows that the planned expansion will not use NVIDIA architecture at meaningful scale, or if a defensible delivery schedule demonstrates that the annualized contribution is immaterial to NVIDIA's operating base.
Current assessment
The evidence supports a stronger conclusion than “SpaceX is interested in AI” but a narrower one than “SpaceX spending will transform NVIDIA's forecasts.” SpaceX and xAI have disclosed large, rapidly deployed clusters using NVIDIA H100, GB200 and GB300 processors, and SpaceX expects a further expansion of at least 220,000 GB300 processors. AI capital spending is also accelerating sharply.
That establishes product use, direction and potential materiality. It does not establish order value, recognized NVIDIA revenue, delivery timing, gross margin or durable architecture share. The thesis remains an active investigation with potentially high materiality and low current quantification confidence.
The next decisive evidence is not another estimate of total data-center construction cost. It is a primary-source disclosure that connects processor quantities and system content to a delivery period. Until that bridge exists, no SpaceX or xAI revenue estimate belongs in the base case.
Sources
- NVIDIA Form 10-Q for the quarter ended April 26, 2026
- NVIDIA fiscal Q1 2027 results, May 20, 2026
- SpaceX prospectus, June 5, 2026
Disclosure
This publication is independent, impersonal research for informational and educational purposes. It is not personalized investment advice, a solicitation, a guarantee, or a recommendation to buy or sell any security. No rating or price target is assigned. Company statements, guidance and planned deployments may change. Author calculations are identified and may differ slightly because of rounding.