Short-Term AI Compute Contracts Price Near Double Long-Term Deals
Nebius priced short-duration capacity at $40 billion to $50 billion per gigawatt annually, compared with $20 billion to $25 billion for longer agreements.

Short-term AI-compute contracts are priced at roughly twice the level of longer agreements, highlighting the premium for immediate capacity as companies expand power, facilities and GPU availability.
Nebius disclosed Aug. 12, 2026, that short-duration capacity was priced at approximately $40 million to $50 million per megawatt of annualized revenue. That equals $40 billion to $50 billion per gigawatt annually. These arrangements typically cover periods ranging from three to six months.
Longer-term agreements were priced at approximately $20 million to $25 million per megawatt annually, or $20 billion to $25 billion per gigawatt. The lower pricing provides suppliers with more predictable revenue and can support financing for expansion.
Nebius also indicated that it could sell its entire 2027 capacity at the short-term pricing level. The comparison shows how customers seeking immediate access to constrained infrastructure may pay more than those committing to longer arrangements.
SpaceX provides a separate measure of the scale involved. The company had 1.4 gigawatts of nameplate compute capacity as of June 30, 2026. Its AI segment generated $2.561 billion in second-quarter revenue and recorded $15.828 billion in capital expenditures during the quarter.
A Google Cloud Services Agreement signed June 5, 2026, covers approximately 110,000 NVIDIA GPUs. The agreement calls for payments of $920 million per month from October 2026 through June 2029.
After Dec. 31, 2026, either side can end the agreement by providing 90 days' notice. That provision means the stated payment schedule does not necessarily guarantee revenue through the full term.
The pricing gap underscores the trade-off between urgency and commitment in AI infrastructure. Shorter contracts can command higher annualized pricing when capacity is scarce, while longer agreements can provide the revenue visibility needed to finance new capacity. Expansion by major AI-compute buyers will determine how much pressure develops on the premium.