Thinking Machines Lab signs $65 million annual inference agreement with Crusoe
- Amount
- $65M
- Place in VP50
- #10
- Source
- 1
- Claims verified
- 5 of 5
What happened
Crusoe announced on September 23, 2026 that Thinking Machines Lab signed a $65 million annual agreement to power inference for the lab's open models on Crusoe Cloud. The release, datelined San Francisco, described a dedicated Tailored Deployment of NVIDIA HGX B200 systems connected with NVIDIA Quantum-2 InfiniBand networking. Crusoe said it would run and support the cluster directly as a benchmarked, SLA-backed endpoint, so the lab could scale without managing its own inference stack.
According to Crusoe, Thinking Machines Lab runs production workloads through Crusoe Managed Inference, including its Inkling models, GLM 5.2 and 5.3, and its own fine-tuned variants. The two companies said they were also looking to expand into batch inference for large-scale synthetic data generation. In the same release Crusoe said its customer roster had taken Managed Inference past $100 million in contracted ARR less than a year from launch. The release stated the amount as annual, gave no contract term and named no data center site for the cluster.
Key points
- Thinking Machines Lab signed a $65 million annual agreement for inference on Crusoe Cloud.1
- The deployment uses dedicated NVIDIA HGX B200 systems with NVIDIA Quantum-2 InfiniBand networking, run by Crusoe.1
- Workloads include Inkling models, GLM 5.2 and 5.3, and Thinking Machines' own fine-tuned variants.1
- Crusoe said Managed Inference had passed $100 million in contracted ARR less than a year from launch.1
- The companies said they were looking to expand into batch inference for synthetic data generation.1
Why it matters
The $65 million annual figure is a stated amount for a named customer, a concrete revenue datapoint for Crusoe's inference business.
Crusoe reported the agreement alongside contracted ARR above $100 million for Managed Inference. The release does not say how much of that figure the agreement represents.
The release states the amount as annual and gives no term length, so how long the commitment runs is not disclosed.
In context
| Round | Date | Post-money valuation | Raised | Led by |
|---|---|---|---|---|
| Series C | Apr 21, 2022 | $1.8B | $350M | G2 Venture Partners |
| Series D | Dec 12, 2024 | $2.8B | $600M | Founders Fund |
| Series F | Sep 17, 2026 | $31B | $3.9B | Atreides Management, Mubadala Capital, and Valor Equity Partners |
Updated
Fact-check
- Thinking Machines Lab signed a $65 million annual agreement for inference on Crusoe Cloud.Confirmed by [1] Crusoe: "a $65 million annual agreement to power inference for the lab's open models on Crusoe Cloud."Verified
- The announcement is dated 2026-09-23.Confirmed by [1] Crusoe: dateline "SAN FRANCISCO — September 23, 2026"; the page also shows September 22, 2026 above the header image.Verified
- The deployment uses dedicated NVIDIA HGX B200 systems with Quantum-2 InfiniBand, run and supported by Crusoe as an SLA-backed endpoint.Confirmed by [1] Crusoe: a dedicated Tailored Deployment of NVIDIA HGX B200 systems with NVIDIA Quantum-2 InfiniBand; Crusoe will run and support the cluster directly as a benchmarked, SLA-backed endpoint.Verified
- Workloads include Inkling models, GLM 5.2 and 5.3, and the lab's own fine-tuned variants.Confirmed by [1] Crusoe: production workloads include "its Inkling models, GLM 5.2 and 5.3, and its own fine-tuned variants."Verified
- Crusoe Managed Inference passed $100 million in contracted ARR less than a year from launch.Confirmed by [1] Crusoe: the roster "has taken Crusoe Managed Inference past $100 million in contracted ARR", less than a year from launch.Verified
Sources
- [1]Crusoe (opens in a new tab)Crusoe and Thinking Machines Lab Partner to Power Open-Model Inference at ScaleCrusoe Managed Inference took us from evaluation to production quickly
