Enterprise AI Compute Cost
- Value
- $8.41 / H100e-hourH100e-hour means an H100-equivalent GPU hour, a normalized unit used to compare different GPUs and providers.
- Daily move
- No Change
Morning Brief
Morning Brief: 2026-05-22
Start-of-day AI compute market report.
Published each morning with the prior day’s source-backed signals across GPU pricing, cloud capacity, power constraints, data centers, and AI infrastructure. Track the prior day’s source-backed news and changes shaping GPU supply, cloud availability, power constraints, data centers, and compute market pressure.
Morning Brief
Morning Tape
- Report date
- Friday, May 22, 2026
- Signal window
- Thursday, May 21, 2026
- SpaceX IPO filing reveals Anthropic set to pay Musk's firm $1.25bn a month to rent xAI data center space
- AMD Announces More Than $10 Billion in Taiwan Ecosystem Investments to Accelerate AI Infrastructure
- OpenAI launches "Guaranteed Capacity" offering, giving customers ability to secure long-term access to compute
Compute Cost Benchmarks
Compute Cost Benchmarks
Spot Compute Price
- Value
- $2.73 / H100e-hourH100e-hour means an H100-equivalent GPU hour, a normalized unit used to compare different GPUs and providers.
- Daily move
- No Change
AI Compute Market Rate
- Value
- $6.60 / H100e-hourH100e-hour means an H100-equivalent GPU hour, a normalized unit used to compare different GPUs and providers.
- Daily move
- No Change
Enterprise 7B Training Cost
- Value
- $73,235 / reference training run
- Daily move
- Daily n/a
Model Serving Cost
- Value
- $3.40 / 1M tokens
- Daily move
- No Change
GPU Availability
- Value
- 70 / 100 score
- Daily move
- No Change
Signals
Top market signals
These are the highest-impact source-backed items from the signal window used for this morning’s brief.
May 21, 2026
SpaceX IPO filing reveals Anthropic set to pay Musk's firm $1.25bn a month to rent xAI data center space
Watch item for AI compute supply, demand, capacity, infrastructure, or market structure.
May 21, 2026
AMD Announces More Than $10 Billion in Taiwan Ecosystem Investments to Accelerate AI Infrastructure
this may affect financing supply, demand, capacity, bottlenecks, or financing.
May 21, 2026
OpenAI launches "Guaranteed Capacity" offering, giving customers ability to secure long-term access to compute
Watch item for AI compute supply, demand, capacity, infrastructure, or market structure.
May 21, 2026
Global data center capacity under construction reached 31.7GW in 2025, more than double the year before - report
Watch item for AI compute supply, demand, capacity, infrastructure, or market structure.
May 21, 2026
Singapore's Sembcorp submits planning application for 280MW data center in Teesside, UK
Watch item for AI compute supply, demand, capacity, infrastructure, or market structure.
Learn the Market
Term of the Day: GPU-hour
A GPU-hour is one graphics processor running for one hour. It is one of the simplest units for comparing AI compute because it turns a messy market into a measurable rental unit. If a provider charges for one H100 running for one hour, that is an H100 GPU-hour. ComputeTape can use GPU-hours to compare on-demand cloud, reserved capacity, neocloud rentals, and spot compute.
Why it matters today: GPU-hour pricing is the base layer for understanding training cost, model serving cost, and compute availability.
Read more: GPU-hourCapacity Watch
Capacity Watch
- AMD Announces More Than $10 Billion in Taiwan Ecosystem Investments to Accelerate AI Infrastructure (AMD Investor Relations).
Power & Data Center Watch
Power & Data Center Watch
- Financing for California’s signature virtual power plant remains uncertain (Utility Dive).
- Fleet breaks ground on data center outside Reno, Nevada (Data Center Dynamics).
- Global data center capacity under construction reached 31.7GW in 2025, more than double the year before - report (Data Center Dynamics).
Watch next
Signals for the next issue
- AMD and Meta Announce Expanded Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs (AMD Investor Relations).
- AMD and OpenAI Announce Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs (AMD Investor Relations).
- Applied Materials to Pay $252 Million Penalty to BIS for Illegally Exporting Semiconductor Manufacturing Equipment (U.S. Bureau of Industry and Security).
- NVIDIA and IREN Announce Strategic Partnership to Accelerate Deployment of up to 5 Gigawatts of AI Infrastructure (NVIDIA Newsroom).
- Lambda Partners with Hudson River Trading to Power Quantitative Research and Development (HPCwire).
Recent Briefs
Last 10 Briefs
Saturday, May 23, 2026
Morning Brief: 2026-05-23
Signal window: Friday, May 22, 2026. Powering multi-cluster workloads with seamless cross‑cluster networking for Azure Kubernetes Fleet Manager
Friday, May 22, 2026
Morning Brief: 2026-05-22
Signal window: Thursday, May 21, 2026. SpaceX IPO filing reveals Anthropic set to pay Musk's firm $1.25bn a month to rent xAI data center space
Thursday, May 21, 2026
Morning Brief: 2026-05-21
Signal window: Wednesday, May 20, 2026. Lambda Partners with Hudson River Trading to Power Quantitative Research and Development
Wednesday, May 20, 2026
Morning Brief: 2026-05-20
Signal window: Tuesday, May 19, 2026. NVIDIA and Google Cloud Empower the Next Wave of AI Builders
Tuesday, May 19, 2026
Morning Brief: 2026-05-19
Signal window: Monday, May 18, 2026. AMD and Meta Announce Expanded Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs
Monday, May 18, 2026
Morning Brief: 2026-05-18
Signal window: Friday, May 15, 2026 - Sunday, May 17, 2026. AMD and Meta Announce Expanded Strategic Partnership to Deploy 6 Gigawatts of AMD GPUs