TL;DR
OpenAI and Anthropic made parallel moves in early May 2026 to build enterprise AI deployment operations, according to source material from Thorsten Meyer AI. The report says both labs are targeting the services layer where companies struggle to move AI systems from pilots into production.
OpenAI and Anthropic moved into enterprise AI deployment services in early May 2026, according to source material from Thorsten Meyer AI, signaling that major AI labs are trying to control not only the model layer but also the work of putting AI systems into production for business customers.
According to the source material, Anthropic announced a $1.5 billion enterprise-services venture with Blackstone, Hellman & Friedman, and Goldman Sachs to embed Claude inside mid-market companies. Hours later, OpenAI announced a $4 billion Deployment Company, called DeployCo, at a $10 billion pre-money valuation, with 19 investment partners.
The source material says OpenAI also acquired consulting firm Tomoro, bringing 150 forward-deployed engineers into the new operation on day one. The model described is borrowed from Palantir: engineers work inside client organizations, study workflows, build systems around frontier AI models, and remain involved until production use is working.
The central claim in the source material is that the labs are responding to a deployment bottleneck. It says model performance is no longer the main constraint for enterprise AI adoption; integration, security review, evaluation systems, and workflow redesign are slowing production use.
Why It Matters
The move matters because enterprise AI revenue may depend less on selling access to models and more on helping companies change how work gets done. The source material cites a six-to-one spending ratio: for every dollar companies spend on software, they spend about six on services.
If that ratio holds, the services layer is a much larger business than model licensing alone. For AI labs, embedding engineers could turn one-off deployment work into recurring model usage, higher switching costs, and expanded token-metered revenue as AI systems take on more work inside a customer’s operations.
The risk is that this business may behave more like consulting than software. If each customer requires large amounts of custom engineering, margins could be lower and scaling could be slower than investors expect.
Your AI Survival Guide: Scraped Knees, Bruised Elbows, and Lessons Learned from Real-World AI Deployments
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Palantir developed the forward-deployed engineer model through years of work with defense, intelligence, and commercial customers. In that model, engineers do not only advise clients; they help build and adapt software systems inside the customer’s operating environment.
The source material argues that OpenAI and Anthropic are applying that approach to the broader enterprise AI market. It also cites MIT research saying 95% of generative-AI pilots fail to move beyond the experimental phase, using that figure as evidence that deployment is the main barrier.
The report frames the May 2026 announcements as part of a wider shift: AI labs are not only competing on model quality, but also on distribution, implementation, and production adoption.
“Within seventy-two hours in early May 2026, the two largest AI labs in the world made the same move.”
— Thorsten Meyer AI source material
“The bottleneck is not the model.”
— Thorsten Meyer AI source material
“almost line for line”
— Thorsten Meyer AI source material
Big Data Analytics: From Strategic Planning to Enterprise Integration with Tools, Techniques, NoSQL, and Graph
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It is not yet clear whether these deployment operations will scale like software businesses or remain labor-heavy services businesses. The source material says the open question is whether embedded engineering becomes a product-formation system or a lasting margin drag.
It is also unclear how much revenue the new structures will produce, how customers will measure return on investment, and whether enterprises will accept deeper operational dependence on a small number of AI labs.
Workflow Automation with Microsoft Power Automate: Design and scale AI-powered cloud and desktop workflows using low-code automation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next test is execution: whether OpenAI, Anthropic, and their partners can turn embedded engineering teams into repeatable enterprise deployments. Investors and customers will be watching adoption rates, retention, margins, and whether AI pilots move into production at a higher rate.
AI Programming Made Practical: A Step-by-Step Guide to Building AI-Powered Applications, Writing Better Code Faster, and Using Modern AI Tools with Confidence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What happened in early May 2026?
According to the source material, Anthropic announced a $1.5 billion enterprise-services venture, and OpenAI announced a $4 billion deployment company within about 72 hours.
Why are AI labs moving into services?
The report says enterprise AI adoption is being slowed by integration, security review, evaluation, and workflow redesign rather than model performance alone.
What is a forward-deployed engineer?
In the model described by the source material, a forward-deployed engineer works inside a client organization, learns the workflow, builds production software around the model, and stays until the system works in practice.
Why is Palantir part of the story?
The source material says the new AI lab deployment structures borrow from Palantir’s long-running model of embedding engineers with customers to build production systems.
What remains unclear?
The main open issue is whether the model can scale with software-like margins or whether it will require too much custom labor for each customer.
Source: Thorsten Meyer AI
