Nvidia Just Reshuffled Its $99 Billion Portfolio, and 1 Stock Moved Up the Ranks

Nvidia reported a roughly $21 billion SpaceX position at the end of fiscal Q2, making it its second-largest holding behind approximately $30 billion in Intel shares; its total investment portfolio was about $99 billion. The stake originated from Nvidia's initial $10 billion xAI investment, following xAI's merger with SpaceX, and reinforces a strategic partnership under which SpaceX plans to use Nvidia GPUs as the primary hardware for its data centers. The article argues that SpaceX's AI-focused data-center expansion could make it a significant buyer of Nvidia's next-generation GPU and CPU products, while limiting AMD's opportunity in this account.

Analysis

The investable signal is weaker than the headline implies: the claimed ownership chain, filing mechanics, and public/private status descriptions require primary-source verification before assigning any incremental earnings value to NVDA. Even if the commercial relationship is real, a customer hardware preference is not a binding multi-year purchase commitment; the relevant underwriting variables are rack-scale order volume, deployment timing, networking content, and payment terms. Treat this as narrative support for NVDA's AI-infrastructure moat rather than a near-term revenue catalyst until a 13F, 10-Q, or customer capex disclosure reconciles the position.

If verified, the more important second-order effect is supplier concentration. A large Musk-linked compute buildout would favor NVDA's full-stack attach rate—GPU, networking and potentially Grace CPU systems—while reducing the probability that AMD gains a meaningful foothold at one of the few customers capable of placing multibillion-dollar accelerator orders. CRWV is a mixed read: additional ecosystem demand validates external GPU-cloud capacity, but vertically integrated customer deployments can displace rented capacity and pressure CRWV utilization or pricing over the next 6-18 months.

Consensus may overstate the strategic value of NVDA holding a mark-to-market stake in a highly valued platform company. Such a stake can increase non-operating earnings volatility and creates perceived related-party/customer-concentration risk, while it does not improve NVDA's core supply position if advanced packaging, HBM, power delivery, or grid interconnection become the binding constraints. The thesis is falsified if verified customer orders fail to appear in NVDA backlog/commentary over the next two earnings cycles, or if AMD discloses a material accelerator win at Musk-affiliated entities.

Key Decisions for Investors

• Do not trade SPCX on this article alone; require confirmation of public float, SEC filings, and the reported NVDA holding before treating it as an investable security. If those records are unavailable or inconsistent, classify the event as misinformation risk rather than a catalyst.
• Maintain or initiate a 1-3 month long NVDA / short AMD pair, sized modestly, only after primary-source confirmation of exclusive or preferred procurement. Use relative performance of +8% in NVDA versus AMD as a profit-taking review point; exit if AMD reports a named hyperscaler or Musk-entity accelerator deployment, or NVDA guides data-center revenue below consensus.
• For a 6-18 month expression, prefer NVDA over CRWV where valuation permits: NVDA captures silicon and networking economics at deployment, while CRWV remains exposed to utilization, financing costs, and customer insourcing. Reassess if CRWV reports sustained utilization above 90% alongside contracted backlog growth sufficient to offset potential customer displacement.
• Monitor NVDA's next two quarterly disclosures for investment-mark changes versus operating revenue and for receivables/customer-concentration commentary. A material unrealized-gain contribution without corresponding hardware revenue is a reason to avoid extrapolating the strategic narrative into EPS estimates.