# AI Buyers Diversify GPU Procurement

AI infrastructure buyers are spreading GPU purchases across hyperscale cloud providers, specialist neoclouds, direct hardware purchases, and in-house data centers as demand outpaces supply and prices rise. Nvidia B200 spot prices have more than doubled since March, and shortages may persist into 2027. For government procurement teams and contractors planning AI workloads, this points to tighter availability and higher costs, making sourcing strategy and capacity planning important. Commercial arrangements—including a SpaceX agreement to supply GPU capacity to Anthropic for $1.25 billion per month through mid-2029—illustrate the scale and duration of commitments in the market, though they are not government awards.

- Buyers should compare availability, flexibility, prepayment exposure, and long-term commitments across cloud, neocloud, direct-purchase, and in-house options rather than relying on a single channel.
- Contractors supporting government AI programs may need to account for GPU price volatility and constrained supply when developing delivery schedules and cost estimates.
- Oracle’s bring-your-own-GPU cloud offering provides another sourcing model for organizations that already own hardware; evaluate it alongside rented capacity where applicable.

**Jurisdictions:** international
**Industries:** Information Technology
**Topics:** Artificial Intelligence, Cloud Services
**Published:** October 12, 2026

### Vendors
- Nvidia (GPU manufacturer)
- AWS (hyperscale cloud GPU capacity provider)
- Nebius (neocloud GPU provider)
- CoreWeave (neocloud GPU provider)
- SpaceX (GPU capacity provider)

### Sources
- [Nvidia GPUs become core AI resource as firms mobilise all channels to secure supply](https://www.digitaltoday.co.kr/en/view/113137/nvidia-gpus-become-core-ai-resource-firms-mobilise-all-channels-to-secure-supply) - 디지털투데이