Crusoe Secured $65 Million Annual AI Inference Deal
Thinking Machines Lab will deploy open models on Crusoe Cloud to scale research infrastructure.
Updated on Sept. 23, 2026 in Artificial Intelligence

Live Poll
Is now a good time to invest in companies heavily focused on artificial intelligence infrastructure?
Thinking Machines Lab has signed a $65 million annual agreement to run its open model lineup, including Inkling and GLM 5.2 and 5.3, on Crusoe Cloud. This move follows Crusoe reaching $100 million in contracted annual recurring revenue for its inference business.
Why it matters
The deal provides Thinking Machines Lab with the computational scale and economic efficiency necessary to support its ongoing research initiatives. It highlights the growing reliance of AI research labs on managed infrastructure providers to handle intensive model training and inference.
Thinking Machines Lab will utilize dedicated Nvidia HGX B200 systems, a significant performance upgrade over previous generation compute, connected via Nvidia Quantum-2 InfiniBand networking for high-throughput data transfer.
The players
Crusoe
A cloud infrastructure provider specializing in high-performance compute clusters and managed inference services.
Thinking Machines Lab
A research organization focused on developing and running open-source large language models.
The details
Crusoe will deliver this capacity as a Tailored Deployment, a managed infrastructure service that enables fine-grained control over the underlying compute and networking stack. By leveraging the Nvidia HGX B200 architecture, the system provides the hardware acceleration required to run sophisticated open models while Crusoe handles the maintenance and support of the cluster.
Timeline
September 2026: Crusoe raised $3.9 billion at a $30.9 billion valuation and secured a $13 billion deal with Jane Street.
September 23, 2026: Crusoe and Thinking Machines Lab announced the $65 million agreement.
The Tech Race
This agreement positions Crusoe as a primary infrastructure partner for research labs aiming to achieve performance parity with industry giants. It reflects a wider industry shift where specialized cloud providers are increasingly competing with hyperscalers by offering bespoke hardware deployments for AI.
Researchers and developers can expect wider availability of the Inkling and GLM 5 models as Thinking Machines Lab increases its inference capacity. The potential implementation of batch inference for synthetic data may eventually accelerate the development cycle for future AI model iterations.
The takeaway
This $65 million deal underscores the capital-intensive nature of sustaining top-tier AI research in the current landscape. Watch for future developments regarding Thinking Machines Lab's reported efforts to raise $6 billion to fund its ongoing model research.
Further reading
For broader trends in infrastructure scale, explore our coverage of Artificial Intelligence.
Live Poll
Is now a good time to invest in companies heavily focused on artificial intelligence infrastructure?









