SpaceX's Google AI Pact Is Worth Up to $29 Billion, but Investors Shouldn't Bank on It

Alphabet is paying SpaceX $920 million per month under a 32-month compute-capacity contract that could total up to $29 billion, while Alphabet describes third-party capacity as a bridge during its data-center build-out. Either party can terminate the agreement on 90 days’ notice after the ramp-up period, and the article argues SpaceX’s current AI-compute demand and pricing—estimated by Elon Musk at $30–$50 per watt—may weaken as supply expands. It flags risks to SpaceX’s revenue durability and returns on data-center investment, while noting Alphabet holds a significant stake in SpaceX.

Analysis

The key asymmetry is that Alphabet can use rented compute to protect customer relationships and cloud-contract conversion while shifting utilization and asset-obsolescence risk to providers. That makes near-term capacity leasing potentially rational for Alphabet even if it is uneconomic for SpaceX once supply catches up. In the medium term, the critical variable is not headline contract value but realized usage, renewal behavior, and pricing per unit of capacity; a 90-day termination right makes booked demand a weak proxy for durable cash flow.

This is also a potential capacity-cycle amplifier. If hyperscalers bring more workloads in-house, neocloud utilization and rental pricing could weaken together, extending payback periods and reducing appetite for additional GPU deployments. That would eventually pressure infrastructure suppliers as well, though the timing depends on whether power, networking, and data-center delivery—not GPUs alone—remain bottlenecks. Oracle and other third-party capacity providers may gain from today’s shortage, but face the same customer insourcing risk.

Alphabet’s SpaceX stake may soften the perceived economics of the arrangement, but should not be treated as a dollar-for-dollar hedge against compute expense: verify how the investment is accounted for and whether any value change flows through reported earnings. Over 6–18 months, the thesis turns on Alphabet’s internal capacity coming online and SpaceX demonstrating a cost or performance advantage beyond scarcity. The orbital-compute case remains an execution option, not evidence of current unit-economics durability.

Key Decisions for Investors

• Consider a conditional relative-value position: long GOOG versus short SPCX, only after checking SPCX borrow availability, liquidity, valuation, and the market value of Alphabet’s SpaceX stake. The thesis is stronger durability of Alphabet’s cloud franchise versus greater sensitivity of SpaceX’s compute economics to utilization and pricing; near-term contract revenue is a material risk to the short.
• For the next 1–3 months, monitor Alphabet Cloud backlog conversion and management commentary on third-party capacity, alongside any SpaceX disclosure of customer concentration, actual utilization, realized pricing, and contract renewals. Do not value the stated contract ceiling as committed revenue without usage and termination detail.
• Falsify the relative-value thesis if Alphabet delays internal capacity materially or continues expanding external leases while cloud demand remains strong, or if SpaceX demonstrates repeatable cost/performance advantages and durable customer renewals as supply improves. Conversely, falling rental prices, lower utilization, or longer compute-capex paybacks would strengthen the SPCX-underweight case.