May 6, 2026 · 5 min read

Anthropic and OpenAI built the same company on the same day

On May 4, 2026, the two biggest AI labs both announced enterprise services joint ventures with private equity. Same day. Within minutes. Here is what is real, what is hype, and what it means for small businesses.

Charcoal and amber illustration of two identical doors opening on a dark hallway at the same moment, no people, soft warm light, abstract suggestion of duplicate corporate moves

What happened

Two days ago, Anthropic and OpenAI both announced new joint ventures with private equity firms whose only job is to send engineers into companies and build AI systems on site. Bloomberg reported the announcements landed within minutes of each other. [Bloomberg]

The Anthropic deal is a $1.5 billion enterprise AI services firm backed by Blackstone, Hellman & Friedman, Goldman Sachs, plus a deeper consortium that includes Apollo, General Atlantic, Leonard Green, GIC, and Sequoia. [Blackstone press release]

The OpenAI deal is called The Deployment Company. It is a Delaware joint venture, anchored by TPG, with Brookfield, Advent, Bain Capital, and SoftBank among the 19 investors. OpenAI keeps majority control. [Bloomberg]

The viral framing has two errors

The Reddit post that pushed this story called it "$10 billion raised" by OpenAI. That is wrong. The $10 billion is the joint venture's valuation. The cash actually committed at announcement is closer to $4 billion. Bloomberg confirmed that number.

The viral version also collapsed Anthropic's $1.5 billion into a single Anthropic cheque. Also wrong. The $1.5 billion is the total founding-partner commitment across Anthropic plus the PE firms plus the consortium investors. Anthropic's own contribution is roughly $300 million.

The headlines are still accurate at the level the headlines work at. Both labs really did launch enterprise services firms on the same day. Just be careful repeating the numbers.

What they are actually doing

Both joint ventures are paying engineers to sit inside customer companies. Not to sell software. To redesign the actual workflows around AI.

Anthropic's announcement says it directly: "applied AI engineers from Anthropic will work alongside the firm's engineering team to build custom solutions and support customers over the long term."

OpenAI's announcement says the same thing in different words. Engineers embedded inside the customer. Building. Iterating. Staying.

This is not a new idea. Palantir has been doing it for almost twenty years and calls them forward-deployed engineers. Multiple outlets drew the comparison the same morning. SiliconSnark ran the headline "OpenAI and Anthropic Both Invented the Same Company This Morning. Palantir Wants Royalties."

The bet both labs are making is the same. The model is good enough. The bottleneck is the human work of figuring out where to put the model. Whoever owns that human work owns the customer.

Who they are actually targeting

Read the press releases carefully and the customer profile is consistent. Community banks. Regional health systems. Mid-sized manufacturers. Multi-site healthcare groups. Retail chains. Real estate operators.

That is private-equity portfolio company territory. Companies with real revenue, complicated workflows, and a cheque-writing executive who will pay for an embedded engineer at a roughly $400,000 a year all-in cost.

It is not the under-fifty-employee small business. The math does not work at that size. A frontier-lab forward-deployed engineer needs a multi-million-dollar engagement to justify the cost. SMBs do not have that engagement size.

Charcoal and amber illustration of a small wooden ship and an enormous corporate barge sailing in the same direction on dark water, both heading the same way, the small ship more nimble

What this means if you are a small business

You will start to hear about "forward-deployed AI engineers" over the next six to twelve months. Your trade press will talk about it. Your PE-backed peers will talk about it. The vendors that sold you software in 2022 will start pitching engineers in 2026.

Most of it will not apply to you yet. The frontier-lab joint ventures are not coming for the SMB segment any time soon. The deal sizes are too small for them.

But the model itself is right. The model says you do not buy AI. You hire someone to put AI inside your work. That person sits with you. They learn what you do. They build tools that fit how you actually run things. They stay long enough that the tools get better.

That model works at any size. It just needs different economics for SMBs. That is the gap I sit in. Same model. Smaller engagement. SMB economics. No PE-portfolio gating.

The lock-in question

The Reddit post framed this as "get so deep inside the company that removing you becomes impossible." That part is editorial. No analyst said it that way. But the concern is real.

If your forward-deployed engineer is paid by Anthropic or OpenAI, the AI inside your workflows is going to get built around one model from one lab. Switching costs go up over time. That is not bad on day one. It can be bad on day five hundred.

An independent shop like Obsidian AI Labs stays model-agnostic by default. I build with whichever model fits the task. I do not have a parent company that needs me to lock you into one.

For some companies that does not matter. They want the brand-name AI lab and they will pay for the lock-in. Fair trade.

For other companies, the option to swap models in eighteen months when something better lands is worth more than the brand on the engineer's badge.

The signal under all of it

The two biggest AI labs in the world just publicly stated their belief that the model is not the bottleneck anymore. The bottleneck is engineers who will sit inside companies and rewire how those companies work.

That is the exact thesis I built Obsidian AI Labs on. I have been operating on it for the small business segment all year. Watching Anthropic and OpenAI validate it on the same day, with $11.5 billion in combined commitment, is a real signal.

The race to embed AI in real businesses just got expensive. It also just got legitimate.

Sources

← Back to blog