Nvidia's $21B SpaceX Stake: What the 13F Reveals

Nvidia’s investment portfolio is starting to look as consequential as its chip business. In a 13F filing submitted to the SEC on August 14, 2026, the company disclosed a public-equity book worth roughly $63.4 billion as of June 30 — and at the center of it are two positions that double as its largest customers. The filing revealed a new, concentrated $21 billion stake in SpaceX and confirmed that an old bet on Intel has swelled to about $30 billion. Together those two holdings account for more than 80% of Nvidia’s disclosed public-equity portfolio.

The numbers are striking on their own, but the strategic pattern behind them is the real story: Nvidia is increasingly an investor in the companies that buy its accelerators, binding its balance sheet to the demand it is trying to create.

The SpaceX position, and how it got there

According to the filing, Nvidia held about 122.8 million shares of SpaceX Class A stock, valued at roughly $21 billion at quarter-end. That makes Elon Musk’s rocket company Nvidia’s second-largest disclosed equity holding, behind only Intel.

The stake did not come from a direct check written to SpaceX. It traces back to Nvidia’s $10 billion investment in xAI, Musk’s AI startup. When SpaceX acquired xAI earlier in 2026, Nvidia’s xAI position converted into SpaceX shares — turning a bet on a frontier AI lab into a large slug of one of the most valuable private companies in the world. In accounting terms, a roughly $10 billion cost basis is now marked at about $21 billion, a paper gain of more than 100% in a matter of months.

There is a hardware quid pro quo underneath the equity. SpaceX has committed to Nvidia’s Vera Rubin architecture for the data centers powering xAI’s compute buildout, and is expected to receive a significant allocation of those chips next year. The same platform that anchors Nvidia’s Rubin roadmap is the one its portfolio company will be buying — a tight loop between supplier and customer.

The Intel windfall

The Intel line is a case study in patience paying off. Nvidia’s roughly $5 billion stake in Intel — taken as part of a strategic partnership — is now worth about $30 billion, holding around 214.8 million shares and representing the single largest slice of the portfolio at roughly 44%. That is a sixfold increase on the original investment.

Intel’s recovery has been the surprise re-rating of the year, driven by renewed confidence in its foundry ambitions and its role as a domestic manufacturing hedge. For Nvidia, the position is both a financial win and a strategic signal: Intel is positioned as an exclusive chip-manufacturing partner in parts of Nvidia’s supply chain, and the equity stake aligns the two companies’ interests as U.S. policy pushes for onshore fabrication capacity. The dynamic rhymes with the $250 billion data-center backstop arrangements Nvidia has struck elsewhere in the AI stack.

The rest of the book

Beyond the two megapositions, the filing showed a portfolio built almost entirely around AI infrastructure:

• CoreWeave — about $4.7 billion, roughly 6.9% of the portfolio. The GPU-cloud specialist is both a major Nvidia customer and a vehicle for renting out Nvidia silicon.
• Nebius Group — about $4.69 billion, another AI-cloud provider tied to Nvidia hardware.
• Coherent — roughly $3.07 billion, an optical-networking and photonics supplier central to data-center interconnects.
• Nokia — about $2.21 billion.
• Synopsys — about $2.15 billion, a pillar of the chip-design software stack.

One notable subtraction: Nvidia exited its Arm Holdings position entirely during the quarter, closing out a stake in the chip-design licensor it once tried to acquire outright. And the portfolio’s overall size tells its own story — it jumped from roughly $18.4 billion at the end of the prior quarter to $63.4 billion, a leap driven mostly by the SpaceX conversion and Intel’s appreciation rather than a wave of fresh buying.

A portfolio that mirrors the customer list

What ties the whole book together is that Nvidia’s biggest holdings are, to varying degrees, its own customers or suppliers. SpaceX buys Rubin chips. CoreWeave and Nebius rent Nvidia GPUs to their own customers. Coherent and Synopsys sit in Nvidia’s supply and design chain. Intel is a manufacturing partner. This is not a diversified treasury operation seeking uncorrelated returns; it is a strategic-investment arm engineered to reinforce demand for Nvidia’s core product.

That approach has echoes across the AI economy, where the largest players increasingly finance the infrastructure that consumes their output. It is also why some analysts have grown cautious about the circularity of AI spending — capital flowing from chipmakers into the very companies that turn around and buy chips, inflating both revenue and equity marks at the same time.

What it means

Nvidia’s 13F is a snapshot of how thoroughly one company now sits at the center of the AI buildout — not just as a vendor, but as a shareholder in the demand it supplies.

Why the concentration matters. A portfolio where two names make up 80% of the value, and where nearly every holding is a customer or supplier, is not a hedge against Nvidia’s business risk — it is a leveraged bet on the same theme. If AI capital spending keeps compounding, these marks rise alongside Nvidia’s core revenue. If the capex cycle cools, the portfolio and the product line would soften together. The equity book amplifies Nvidia’s exposure to the AI trade rather than diversifying it.

Who wins. SpaceX and xAI gain a chip supplier with a direct financial incentive to see them succeed and a guaranteed allocation of scarce accelerators. Intel wins a validation stamp and a strategic partner as it fights to prove its foundry model. Nvidia wins on paper — a roughly $10 billion xAI stake now marked at $21 billion, a $5 billion Intel bet at $30 billion — and wins strategically by locking in demand.

What to watch. First, whether these gains are ever realized: 13F values are mark-to-market snapshots, and private-company marks like SpaceX depend on the next funding round rather than a liquid market. Second, whether regulators or Nvidia’s own auditors scrutinize the circularity of investing in customers who then buy chips — a pattern that flatters both sides of the ledger. Third, the SpaceX position specifically, which will move with any secondary-market repricing or eventual IPO. For now, the filing confirms the obvious: in 2026, betting on Nvidia and betting on the AI infrastructure boom have become nearly the same trade — and Nvidia has arranged its own portfolio to prove it.