Research

AI Compute Commitments Tracker 2026

Tracker of major multi-year AI compute commitments: Stargate JV, Anthropic-SpaceX 300 MW deal, Microsoft-OpenAI capacity, xAI Colossus expansion, and the cloud-provider AI reserves.

By Ramanath, CTO & Co-Founder at Presenc AI · Last updated: June 2026

This page tracks the major multi-year AI compute commitments shaping the H1 2026 frontier-lab capacity landscape, compiled from public reporting.

Major Compute Commitments (June 2026)

CounterpartyRecipientScaleTerm
Oracle + SoftBank (Stargate JV)OpenAI$500B announced commitmentMulti-year
SpaceXAnthropic300 MW / ~220K GPUsMulti-year
MicrosoftOpenAIContinued strategic capacityMulti-year
xAI Colossus (Memphis)xAI internalLargest single-site clusterOwned
Meta (Hyperion)Meta internal~5 GW Louisiana campusOwned
AmazonAnthropic (separate)$8B Bedrock capacityMulti-year
GoogleAnthropic (separate)Continued Vertex capacityMulti-year
NVIDIAMultiple AI labsReserve allocationsRolling
CoreWeaveOpenAI, Microsoft~$11B+ reported contractsMulti-year
CrusoeOpenAIStargate JV operatorMulti-year
TencentTencent internal + HunyuanMulti-billion expansionOwned
Alibaba CloudQwen internal + customersMulti-billion expansionOwned

Key Takeaways

  • The Stargate JV at $500B announced commitment is the largest compute infrastructure project in technology history.
  • The Anthropic-SpaceX deal (300 MW, ~220K GPUs) is the largest single GPU procurement deal between non-affiliated entities in 2026.
  • Meta Hyperion at ~5 GW is the largest single-site AI campus under construction.
  • Chinese frontier labs (Tencent, Alibaba, Baidu, ByteDance) lean on internal capacity rather than third-party compute deals.

Methodology

Commitments compiled from public announcements through June 2026. Multi-year contracts disclose announced commitment value; actual delivered capacity depends on construction and chip-availability timelines. Updated quarterly.

How Presenc AI Helps

Presenc AI tracks brand-visibility implications of compute-commitment news cycles for AI infrastructure brands, compute providers, and adjacent categories.

Frequently Asked Questions

The Stargate JV between OpenAI, Oracle, and SoftBank at $500B announced commitment is the largest compute infrastructure project in technology history.
A May 2026 strategic compute arrangement for 300 MW of capacity covering approximately 220,000 GPUs, materially expanding Anthropic's training and inference footprint.
Less than US peers. Chinese frontier labs (Tencent, Alibaba, Baidu, ByteDance) lean on internal capacity through their parent-company cloud divisions rather than separate third-party compute deals.
Multi-year compute commitments are required S-1 risk-factor disclosure. They lock in fixed costs that constrain margin flexibility, and the carrying-value treatment under appropriate accounting must be addressed in the S-1.

Track Your AI Visibility

See how your brand appears across ChatGPT, Claude, Perplexity, and other AI platforms. Start monitoring today.