May 4, 2026. Two announcements landed at once.
Anthropic formed a $1.5 billion joint venture with Goldman Sachs and Blackstone. The goal: embed Claude directly into private equity portfolio companies. Redesign their workflows. Send Anthropic engineers inside.
Same day, OpenAI launched The Development Company — raising $4 billion from 19 investors to do the same thing at larger scale.
Most coverage framed this as "AI companies going enterprise." That framing misses what's actually happening.
The real story: AI model vendors just entered the competitive set for enterprise software.
Here's the scenario I kept running through after reading those two pieces.
Say you sell software to a 200-person hospital network or a regional law firm. Last year, that firm was acquired by private equity. This month, Anthropic's engineering team is on-site doing workflow analysis. Next quarter, the software budget is up for review.
What might replace you isn't a competitor product.
It's Goldman-backed capital, Anthropic engineers, and a PE board with a mandate to create value — arriving at the same time.
That combination didn't exist before.
PE risk always existed in B2B SaaS. Customer gets acquired, budgets get cut, old vendors get evaluated. Standard playbook. And typically, the best product survived.
But this version adds something new.
Anthropic's Goldman partnership is explicitly building deployment templates — take what worked in one firm and replicate it across thirty more. A sector template. Starting in financial services, moving outward.
That template will eventually reach your customers. And your customers' competitors. Inevitable.
There's also a structural shift in incentives. Anthropic isn't just selling licenses here. It wants a share of business outcomes. That changes who the model company is aligned with — and it's not necessarily your interests as a software vendor built on their API.
As a PM, here's how I read this.
Your product's integration layer is now worth more than the product itself.
If Claude can operate inside your software — access your data, understand your workflows, call your APIs — then when Anthropic engineers arrive at a shared customer, they have no reason to replace you. You're already part of the infrastructure. If it's working, why rip it out?
But if your product is a closed box — no public API, no agent compatibility, no way for external processes to interact with your data layer — then "why are we still paying for this?" becomes a question a PE CFO asks. And the answer gets harder to defend.
Consider a product with 17 enterprise customers. How many have been PE-acquired? How many will be in the next 18 months? This used to be an edge case. Now it's a risk metric worth tracking explicitly.
The second thing is subtler.
"We use Claude" is no longer a differentiator.
Eighteen months ago, saying "we're built on Claude's API" meant something in a demo. Now every competitor says the same thing. And soon, Anthropic will be distributing pre-built agents for specific verticals, directly to Goldman's portfolio companies, at scale.
So your competitive set now includes: the model company's engineers + the PE firm's capital + the portfolio board's mandate to cut costs and drive outcomes.
That combination hitting at once is a different kind of pressure than losing a deal to a better product. A better product, you can respond to. This is structural.
I don't have a clean answer here. But the direction seems clear enough.
Vertical depth is the moat that doesn't transfer. A healthcare workflow tool that understands three years of patient data models, local compliance rules, and how clinicians actually deviate from protocol — that knowledge doesn't show up in a deployment template. It takes time and proximity to build. It's hard to replicate in six months, no matter how good the engineers are.
And becoming part of the agent infrastructure, not competing against it. If your product is the source of truth that agents query, you're inside the workflow. If you're a standalone UI that users interact with manually, you're the thing agents eventually route around.
The distance between those two positions is growing.
Enterprise AI competitive analysis looked very different 12 months ago. Worth updating the analysis.