Nvidia vs. AMD: Elon Musk Picked a Side on the SpaceX Earnings Call

Elon Musk said SpaceX will build exclusively on Nvidia's Vera Rubin architecture, reinforcing Nvidia's position in AI infrastructure versus AMD. The article states SpaceX has monetized Nvidia-based compute through large customer commitments, including Anthropic at $1.25 billion per month through 2029 and Google Cloud at $920 million per month through mid-2029. Nvidia's Q2 data-center revenue reportedly reached $89 billion, up 117% year over year, versus AMD's $6.7 billion, up 107%; Nvidia also trades at a lower forward P/E of about 23x versus AMD's roughly 67x.

Analysis

The incremental signal is less about one customer's GPU preference than about the economics of vertically integrated AI compute: a proprietary compute owner that can monetize capacity through long-duration external contracts raises the value of Nvidia's full-system moat—accelerators, networking, software and deployment support—rather than merely its silicon share. If independently verified, this strengthens NVDA's pricing power in rack-scale deployments and is incrementally negative for AMD's ability to close the software- and systems-integration gap. The second-order beneficiaries are NVDA supply-chain bottlenecks, notably SK Hynix (HBM), AVGO (custom networking/optics exposure) and VRT (data-center power and cooling), where demand is constrained by physical deployment capacity rather than chip availability alone.

The article's commercial claims require verification before being traded: the described SpaceX public earnings call, contract terms, and GPU fleet figures are not sufficient evidence on their own. Even if real, customer concentration cuts both ways—large lease commitments can create an AI-infrastructure credit/liquidity risk if downstream model providers reduce usage or if GPU utilization disappoints. Over the next 1-3 months, confirmation through filings, counterparty disclosures, supplier commentary, or NVDA backlog language would be a catalyst; over 6-18 months, the key question is whether external compute leasing earns returns above power, depreciation and financing costs, rather than simply pulling forward GPU purchases.

Consensus likely overweights the headline as a direct AMD read-through. Hyperscaler and sovereign demand, not a single Musk-linked ecosystem, remains the determinant of AMD's accelerator trajectory; MI-series share gains can still matter materially from a small base. The cleaner expression is therefore relative: NVDA can outperform AMD if validated evidence shows enterprise buyers increasingly standardizing on integrated rack-scale architectures, but the pair fails if AMD demonstrates comparable cluster economics, materially improves software portability, or wins large disclosed deployments outside the hyperscalers.

Key Decisions for Investors

• Do not trade the SpaceX-specific claim until it is corroborated by SEC filings, counterparty disclosures, or management commentary; set an event-driven alert for NVDA supplier/backlog confirmation rather than treating promotional reporting as a catalyst.
• Conditional 1-3 month pair: long NVDA / short AMD in equal beta-adjusted dollars only after verification. Target 10-15% relative outperformance; exit if AMD reports accelerator revenue or MI400 bookings materially above consensus, or if NVDA's next data-center gross-margin guide implies system-level pricing pressure.
• For broader AI infrastructure exposure, accumulate VRT on deployment-led pullbacks rather than chase NVDA beta. The thesis is power-density and cooling scarcity; reassess if data-center order growth decelerates for two consecutive quarters or major customers defer capacity additions.
• Avoid treating TSLA as a direct beneficiary. Incremental third-party GPU procurement can support autonomy training capacity, but it also reinforces dependence on external compute; require evidence of improving robotaxi utilization or autonomy economics before assigning multiple expansion to the AI narrative.