When the pitch moves from ‘buy our model’ to ‘let our engineers move into your org chart,’ you’re not buying software anymore. You’re buying a new operating system for the company.
Anthropic and OpenAI are both leaning into a similar play: partner with alternative asset managers to build a new enterprise deployment channel. It looks like helpful services. It is also a blueprint for how AI becomes a financial product.
What happened
Anthropic announced the formation of a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs, aimed at bringing Claude deployments into mid-sized companies via hands-on engineering engagements (Anthropic). Anthropic says applied AI engineers will work alongside the new firm’s team to build custom solutions and support customers over the long term.
TechCrunch reports the move lands alongside a similar OpenAI effort: Bloomberg reported OpenAI raising money for a venture called “The Development Company,” with OpenAI’s venture described as larger in scale (TechCrunch). The shared logic is blunt: raise money from alternative asset managers to create new channels for enterprise deals — and capture more of the value created by those contracts.
The non-obvious angle: this is “AI distribution” becoming a financial instrument
The story isn’t “labs are doing services now.” Labs have always done a bit of “we’ll help you integrate.” The shift is the structure: asset managers + privileged sales access + long-term deployment work. That’s a machine designed to compound.
Once your deployment channel is tied to capital pools, the incentives change. You stop optimizing for “best tool for the job.” You optimize for “the tool that makes the portfolio sticky.” And because this is implementation work, it becomes hard to unwind. Nobody rips out the system that rewired their workflows just because a benchmark chart moved.
Why the Palantir comparison keeps showing up (and why it matters)
TechCrunch notes the forward-deployed engineer (FDE) model — the Palantir-style approach where engineers sit close to the customer and adapt the product to real operations (TechCrunch). That model is effective because it converts software into an institutional relationship.
But it also changes the buyer’s risk profile. You aren’t buying a SKU. You’re letting a vendor co-design your workflows. The switching costs are no longer “migrate data.” They’re “retrain the organization.”
Scenario analysis: three plausible futures for enterprise AI deployment
1) The ‘Services Layer’ future (most likely)
Enterprise AI becomes like cloud in 2014: the winners are the ones who can ship product and staff an army to implement it. The labs become platform vendors; the asset managers become distribution; the integrators become power brokers. Everyone takes a cut; customers pay in both money and dependency.
2) The ‘Commoditized Models’ future (inevitable, just slower)
Models get cheaper and more interchangeable, but the deployment layer gets more valuable. The services company is the moat. Your “AI vendor” becomes whoever owns your change-management plan and your internal toolchain, not whoever has the fanciest model card.
3) The ‘Procurement Backlash’ future (the regulatory wildcard)
Once a few high-profile deployments go wrong — privacy incidents, hallucinations in regulated workflows, budget blowouts — buyers start demanding auditable controls, logs, and liability clarity. That pushes the market toward governance features, not just clever demos. The services channel becomes the compliance channel.
The Singularity Soup Take
This is financial engineering wearing a hoodie. The labs need deployment velocity, the asset managers need yield, and enterprises need someone to do the boring work. So we’re inventing a new thing: “AI as an operating partnership.” It will work — and then everyone will be shocked that it also creates lock-in.
What to Watch
- Pricing power: do these ventures bundle model spend + services into long-term contracts that look like “compute mortgages”?
- Portfolio routing: do the asset managers steer deployments toward their own portfolio companies first (creating a closed loop of “preferred customers”)?
- Liquidity tells: secondary markets have become a trust/governance signal for labs. Watch whether deployment JVs change the narrative about who has durable demand.
Sources
Anthropic — "Building a new enterprise AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs" (May 4, 2026)
TechCrunch — "Anthropic and OpenAI are both launching joint ventures for enterprise AI services" (May 4, 2026)
OpenAI — "OpenAI raises $122 billion to accelerate the next phase of AI" (Mar 31, 2026)