Anthropic Didn't Buy a Customer. It Bought the Buyer

追加: 2026-05-26

[00:00:00]While everyone's refreshing their feed waiting for Claude 4, the most important AI story of May 2026 already happened.

[00:00:06]It wasn't a model release. It was a check, $1.5 billion. And here's the thing nobody's saying out loud. The number isn't the news. The number is the distraction. By the end of this video, you'll know exactly which mid-market software categories are now on the clock and what to do about it. Whether you build that software, sell it, invest in it, or just open it every morning to do your job. Something's shifting here. Let me show you the shape of it. This deal tells us the AI game just stopped being about whose model is smartest. It's now about whose AI is wired into actual businesses, real ones with payroll and warehouses and call centers. And once you see that, you can't unsee it. So, that's what I'm going to do here. Take this deal apart layer by layer until the real machinery is showing. And about halfway through, there's one line. One.

[00:00:53]If you take nothing else from this video, you'll take that line and the whole month of May suddenly reads differently. Oh, and stick around to the end because there's one kind of company inside this portfolio that's about to become the live test case for the entire playbook. Almost nobody's watching it.

[00:01:08]I'll tell you exactly what to watch for later. Build software, sell it, invest in it, or just use it. This hits all four. Hold on. If this seems weird to you, good. It should. AI labs partner with clouds. That's the normal move.

[00:01:22]They don't partner with the people who own the factories and the call centers.

[00:01:25]Until now. Quick grounding because I'm not going to overexplain either of these. Anthropic, the lab behind Claude.

[00:01:31]Blackstone, one of the biggest private equity firms on Earth. Roughly a trillion dollars under management.

[00:01:36]Somewhere around 250 companies in the portfolio. Treat those two as ballpark.

[00:01:41]Go verify them before you quote them.

[00:01:43]But, the order of magnitude is the point. And one correction the headlines mostly skipped. It's not just Blackstone, it's a joint venture.

[00:01:51]Blackstone, Hellman and Friedman, and Goldman Sachs all wrote checks, too.

[00:01:55]That's not trivia. That's the whole story. Let me show you why this is the most logical move on the board. The press release version is clean. A $1.5 billion joint venture to accelerate AI adoption across enterprises. Embed engineers, redesign workflows, deploy Claude, corporate, boring, easy to scroll past. Now read it underneath.

[00:02:14]Underneath what you've got is hundreds of portfolio companies whose owners just decided from the top what AI they're going to run. A pre-installed customer base. Suppliers that are functionally steered toward one model by contract.

[00:02:26]$1.5 billion sounds like a lot until you realize what Anthropic just bought is the one thing OpenAI can't get for any amount of cloud money. They bought the buyer. Think about what's actually sitting in a portfolio like that.

[00:02:39]Logistics operators, healthcare services companies, industrial and manufacturing businesses, retail chains, and I'm speaking in categories on purpose. I'm not going to name and accuse some specific real company that hasn't done anything wrong. But these aren't tech companies. These are the businesses that run the boring middle of the economy, and they all just got told which AI they're standardizing on. $1.5 billion doesn't buy a customer. It buys a distribution channel. Now layer two. Why does Blackstone need Anthropic at all?

[00:03:08]They could call any vendor on Earth.

[00:03:10]Here's the trap these companies are stuck in. A mid-market company, 300 employees, solid revenue, real operations. It wants AI. Of course it does. Everyone above it is asking about it on every board call. But it can't hire AI engineers. The going rate for that talent is around half a million a year. That's a rough number. Verify it before you quote it. But it's in that zone. And that company is not winning a bidding war against a frontier lab for the same five people. And say it could hire them. Look at what they'd be hiring them to fix. Most of these companies are running software that predates 2015.

[00:03:43]Some estimates put it around 70% of mid-market software. Again, treat that as roughly, not gospel. But anyone who's worked inside one of these places knows the shape of it is true. There's no clean path from that stack to a modern AI integration. There just isn't. So, here's the trick. Anthropic just skipped the entire enterprise sales cycle. They didn't have to win each of those companies one painful deal at a time.

[00:04:05]And this part, the press release says it openly, the joint venture embeds engineers inside those companies and rebuilds the workflows for them. Sit with that. Anthropic doesn't sell them an API and wish them luck. Blackstone sells them an operational transformation and Claude is the deliverable. That's not a software license. That's the consulting industry's lunch taken right off the table. They didn't win the deal, they bought the buyer. Which raises the obvious question. If this is so clean, why doesn't OpenAI just do the exact same thing tomorrow? Hold that thought.

[00:04:33]We come back to it and honestly the answer is the best part. Stop. This is the line. The one I told you about. If you only keep one sentence from this whole video, keep this one. OpenAI is becoming cloud infrastructure. Anthropic is becoming business infrastructure.

[00:04:48]Those are not the same race. They're not even on the same map. One wants to be the layer everything else gets built on.

[00:04:54]The other wants to be inside the building where the work actually happens. We pulled apart the cloud infrastructure half of this in the OpenAI AWS video. Links at the end. This is the other half. If you build mid-market SaaS, this is the part to watch closely. This is the one that's going to cost somebody their company.

[00:05:10]And remember, I told you up top there are categories on the clock. Here's where I start naming the clock. Take one of those portfolio companies and actually count the software it runs. A typical one is paying for somewhere between 30 and 50 separate SaaS tools. A niche CRM, a scheduling tool, a billing system, a support ticketing layer, some vertical analytics dashboard nobody fully understands but everybody's afraid to cancel. Now, drop a capable agent into that company's workflow. One that can read the data, take the actions, talk to the systems directly. What happens to the niche tool that just did invoice processing? Redundant. The route optimization tool? Redundant. The bolt-on support ticketing layer redundant. And here's the part that turns a trend into an extinction event.

[00:05:52]The decision to rip those out and replace them doesn't happen one company at a time with some slow committee. In a PE portfolio, it gets made once at the top and pushed across the whole portfolio. That's roughly 250 companies at the same time. So every vertical SaaS company that leans on PE owned mid-market customers is now a clock.

[00:06:12]Some of them have 18 months. Some of them have six. And I'll be straight with you because this channel doesn't get to stand above this. Half the tools I cover here sit in that exact crosshair, including ones I use myself. I'm not narrating this from a safe distance.

[00:06:25]Nobody in this conversation is. So, why doesn't Open AI just do the same thing tomorrow? It can't. Not easily. And the reason is structural, not strategic.

[00:06:36]Look at the stack. Hardware at the bottom. Cloud above it. Models above that. Applications on top. Open AI has spent the last year wiring itself into the bottom of that stack. The AWS deal, the enormous compute commitments. And up at the application layer, Microsoft's Copilot already owns a huge chunk of enterprise distribution. Distribution Open AI is tied to. You can't be the operating system and the app at the same time without breaking your own commitments. If Open AI walks into mid-market companies and starts replacing the apps, it's competing with the very partners it just signed up to power. This is the difference between playing chess and playing poker. Open AI is optimizing for compute, for raw position, for the biggest pile of chips.

[00:07:20]Anthropic is optimizing for control of where the work actually happens.

[00:07:24]Anthropic picked the one side of the stack Open AI structurally can't walk into. That's not luck. That's chess.

[00:07:31]Remember the test case I teased at the start? Here it is. And I want to be honest about what it is and what it isn't. I'm not going to point at a real named company and tell you it's about to gut its workforce. That's irresponsible.

[00:07:42]and frankly nobody outside the deal can verify it yet. So let me be clear. This is the kind of company to watch, a pattern not a name. Picture a Blackstone owned logistics company, roughly 3,000 employees. The reason this kind of company is the test case is simple. It's head count and its revenue are both trackable from the outside, its workflows are exactly the automatable kind, and the contract is fresh. So here's your watch list. Two metrics.

[00:08:08]Watch the head count, watch the revenue.

[00:08:10]If head count drops 8 to 15% while revenue holds or grows in Q4 2026, the playbook worked. And if it works once, it's coming for every PE owned business in the country. That's the signal. Mark it down.

[00:08:25]Before I tell you which SaaS category I think dies first, what's your guess?

[00:08:29]Drop it in the comments. I want to see if the room reads this the way I do.

[00:08:33]Okay, the honest part. What do you actually do with this?

[00:08:37]And it depends on who you are. Say you're an engineer at a mid-market SaaS company. Here's the uncomfortable spot you're in. The product you're paid to make better might be on the kill list, and the kill order doesn't come from your CEO. It comes from an owner three levels up who you've never met. You can't outwork that, so don't try. Two moves work. Move toward the integration layer, the deep ugly hard to replicate part of the product an agent can't just absorb, or move toward a customer base that private equity doesn't own. One of those two, not neither. Say you're a founder building vertical AI. Your instinct is going to be to ship a better model, a smarter agent, a slicker product. And that instinct will sink you. Your moat is not the model. It was never going to be the model. Your moat is distribution. If you don't have a PE style channel into your customers, you need one. Or you need a customer category PE hasn't bought yet. And there are still plenty. Go find one before it has an owner. And say you're an observer or an investor. You've spent two years training yourself to read benchmark scores, and that habit is now actively lying to you. Stop tracking model benchmarks. The leaderboard is theater now. Start tracking distribution deals.

[00:09:46]The next three of these, and there will be a next three, will rewrite the map faster than any eval score ever could.

[00:09:52]So, here's the reframe. Irrelevance isn't a forecast. It's a position. You can choose which side of this deal you stand on, but you have to actually choose. Doing nothing is also a choice.

[00:10:04]It's just the losing one.

[00:10:06]6 months ago, the AI war looked like OpenAI versus Anthropic versus Google, model versus model. It's not that anymore. It's labs versus clouds versus operators. And the operators, the ones who own the actual businesses, just won round one without shipping a single model. We've said this before on this channel, in the cloud lie. The visible fight is never the real fight. We said it again with the NPU bottleneck. We're saying it here. And we'll keep saying it until the news cycle catches up. So, loop A. The thing I promised you at the very start. Which categories do I think go first? This is my read, and I'll defend it. Mid-market invoice and accounts receivable automation goes first. It's pure workflow, no defensibility. Then vertical CRMs sold into PE-owned industries, because the owner can swap them out wholesale in one decision. And then the niche support ticketing layers that just sit on top of a CRM and add a little structure. Those three are the most exposed. That's where I'd be nervous, or short, or already moving. And if you want the other half of this map, why OpenAI broke up with Microsoft to chase AWS, that video's on screen right now. This deal is one half of the May 2026 board. That's the other.

[00:11:17]Watch them together, and the whole month finally makes sense. If this is the lens you've been missing, that's the channel.