SpaceX is traditionally recognized for its dominance in aerospace, but Elon Musk has recently pivoted the company’s focus toward a new frontier: massive-scale AI infrastructure. In a move that has startled the tech industry, Musk announced ambitious plans to deliver unprecedented gigawatt-scale datacenter capacity. By positioning SpaceX as a high-speed provider of AI compute, the company is entering direct competition with traditional hyperscalers like Amazon Web Services and Google, despite having a smaller balance sheet than these tech giants.
What happened
During a recent corporate update, Elon Musk outlined a “conservative” goal for SpaceX to build and deliver between 6 and 8 gigawatts (GW) of incremental compute capacity in 2027 alone. Internal projections suggest the final figure could exceed 10GW. Achieving this scale requires a staggering capital expenditure of approximately $300 billion to $500 billion, assuming a cost of $50 billion per GW.
To finance this massive undertaking, SpaceX is reportedly leveraging a strategic partnership with NVIDIA. By committing exclusively to NVIDIA hardware, SpaceX is expected to secure vendor financing, reducing the immediate cash burden of acquiring high-end GPUs. Furthermore, SpaceX intends to bypass traditional datacenter construction delays by offering lead times as short as three to five months. This speed allows them to command premium rental prices, ranging from $30 million to $50 million per megawatt annually, which could potentially pay back the initial capital investment in under a year.
Context
The primary driver behind this aggressive expansion is the sheer profitability of AI inference. Analysis of modern “tokenomics” suggests that frontier model companies, such as OpenAI and Anthropic, can generate upwards of $100 billion in revenue per GW per year. By running advanced models on cutting-edge clusters, these labs can produce tokens at a rate that far exceeds the cost of hardware and electricity.
Microsoft is also a central player in this shift. Having renegotiated its partnership with OpenAI to eliminate certain revenue-sharing constraints, Microsoft now has a massive incentive to secure as much power as possible. Microsoft has already signed contracts for 10GW of capacity to be delivered through 2028, but a near-term supply gap remains. SpaceX is positioned to fill this gap, offering Microsoft the immediate capacity needed to potentially accelerate Azure’s revenue growth into triple-digit territory.
Why it matters
This development signals a transformation in how global compute capacity is built and sold. By treating datacenters as a rapidly deployable product rather than a multi-year real estate project, SpaceX is disrupting the standard timeline for AI scaling. The use of a 90-day cancellation policy—similar to agreements SpaceX has utilized with other tech leaders—minimizes the long-term financial risk for customers like Microsoft, making it easier for them to commit to massive contracts.
Furthermore, the scale of SpaceX’s ambition highlights the intensifying “arms race” for inference capacity. As AI models move from the training phase to widespread production, the demand for power and silicon is shifting. If SpaceX successfully delivers 10GW by 2027, it will not only solidify its role as a critical infrastructure provider but also redefine the economics of the AI era, proving that speed and specialized financing can overcome the advantages of established hyperscale incumbents.