Partnership
Crusoe x Thinking Machines Lab (Managed Inference)
Sep 23 2026 company announcement: Thinking Machines Lab signed a $65 million annual agreement to run production open-model inference on Crusoe Managed Inference via a dedicated NVIDIA HGX B200 Tailored Deployment (Quantum-2 InfiniBand). Crusoe says the deal takes Managed Inference past $100M contracted ARR. Company press; commercial terms otherwise undisclosed.
technology · Active · just announced
Machine-readable surfaces
- Markdown mirror: /partnerships/crusoe-x-thinking-machines-lab-managed-inference.md
- JSON-LD: embedded in this page’s head
- REST API: /v1/partnerships/crusoe-x-thinking-machines-lab-managed-inference
- Revision history: /partnerships/crusoe-x-thinking-machines-lab-managed-inference/history
- Data documentation: /data
- Query this programmatically: Deploy MCP
a25aeece-47e5-4ae7-989d-1db02799d52b- Type
- technology
- Status
- Activejust announced
- Announced
- 2026-09-232 months ago
- Active for
- 2 months
Who’s involved
Crusoe started in 2018 as a stranded-energy play, running Bitcoin miners on flare gas at oil wellheads in the Bakken and Permian, then pivoted hard into AI compute as the training market took off. In early 2025 it divested the Bitcoin mining business to NYDIG and rebranded around a single thesis: an "energy-first AI factory company" that builds and operates its own campuses rather than reselling colocation. The result is a vertically integrated stack, from power procurement and site development through data center construction to a managed GPU cloud and, most recently, a serverless fine-tuning and inference layer called Crusoe Intelligence Foundry. The company's flagship site is the Abilene, Texas campus, originally announced as the anchor build for the Oracle-led Stargate project supporting OpenAI training workloads, and now serving Microsoft AI capacity as well. Crusoe says its contracted AI infrastructure pipeline is approaching five gigawatts across sites including Abilene, a newly announced 1.0 GW campus in Childress with Lancium, and international expansion in Iceland (via atNorth), Norway (Polar) and Argentina. Announced power deals with Bergen Engines, ON.energy, Form Energy and a nuclear partnership with Aalo Atomics signal a strategy of locking in low-cost, purpose-built energy well ahead of GPU delivery. Compared with the rest of the neocloud pack, Crusoe is distinctive on two axes: it is a builder-operator, not primarily a lessor of hyperscale colocation, and its go-to-market spans both hyperscaler-scale anchor tenants (Microsoft, and via Stargate the OpenAI workload) and a long tail of AI-native startups (Cursor, Fireworks, Together AI, Modal, Baseten, Luma, HeyGen, Figure, Cognition). That mix places it in the same conversation as CoreWeave and Lambda for training capacity, while its energy-development posture more closely resembles Crusoe's own Bitcoin-era competitors and traditional data center developers than pure GPU-cloud peers.
Most-frequent partners
Thinking Machines Lab is an artificial intelligence research and product company founded by Mira Murati (former OpenAI CTO). Tracked here for open-model inference and Physical AI-adjacent infrastructure partnerships, including a Sep 2026 Crusoe Managed Inference agreement.
Most-frequent partners
Related partnerships
Other partnerships of Crusoe
- deployment2026-09-28just announced
- deployment2026-03-27operating
Recent partnerships of the same kind
- technology—just announced
- technology—just announced
- General Mills x Google CloudGeneral Mills + Google Cloudtechnology—operating
- loram x One Big CircleLoram + One Big Circletechnology—operating
- technology—operating
- technology—just announced
Sources (2)
Deploy Watch
Track this partnership.
We notify you when the status changes against primary-source evidence, expansion is verified, or the partnership is terminated.