Behind Anthropic’s Trillion-Dollar Valuation: Can AI Be More Than a Layoff Machine?

Private equity as a distribution channel is a better way to transform the real economy

Last week, I wrote about the much-anticipated IPO paths of SpaceX and OpenAI, under the broader question of whether software, now old enough to feel like the establishment, can still “eat the world.”

A few days later, Anthropic showed up with a number that made the whole conversation feel outdated before the ink had dried.

On May 28, Anthropic announced a $65 billion Series H financing round at a $965 billion post-money valuation, led by Altimeter Capital, Dragoneer, Greenoaks, and Sequoia Capital. Reuters reported that this valuation surpassed OpenAI’s $852 billion valuation from March, while Anthropic said its run-rate revenue had passed $47 billion earlier in May.

My first reaction was not shock.

It was a question mark.

Not because the number was unimportant. Quite the opposite. The number was so important, and frankly so absurd, that it forced a more basic question: what exactly are we supposed to compare it with?

If you only read the headline, this looks like an irresistible social media story. Claude beats ChatGPT. Anthropic overtakes OpenAI. The AI valuation war has reached the doorstep of one trillion dollars.

That is also the easiest way to misunderstand the story.

A private market valuation is not the same thing as a public market valuation. Even a private market valuation and a future public-market-ready valuation are not the same thing. Private rounds can include structured terms, liquidation preferences, strategic resource commitments, cloud capacity relationships, IPO signaling, and side letters ordinary readers will never see. Public markets are crueler. Quarterly earnings, sell-side models, secondary liquidity, index flows, retail sentiment, and short reports all get to vote every day.

So the question of whether Anthropic is “really” worth more than OpenAI is not easy to answer. If we stop there, whether we call it another AI bubble or an Anthropic victory lap, we are probably misleading ourselves.

The more interesting question is what this near-trillion-dollar valuation reveals underneath the headline.

AI is looking for a path into the real economy.

That was also the hidden thread in the SpaceX and OpenAI discussion. The point was not simply which company might go public first, or whether their valuations were too high. The more interesting part was that both companies were positioning themselves as next-generation AI infrastructure companies, but from opposite ends of the stack.

SpaceX is the physical infrastructure version: launch, connectivity, satellites, Starlink cash flow, Starship operating leverage, and eventually the narrative option of orbital compute. OpenAI is the intelligence-entry-point version: models, product, distribution, compute procurement, ChatGPT user scale, and a corporate structure designed to carry it toward the public markets.

In plain English, SpaceX sells the physical-world version of AI infrastructure. OpenAI sells the consumer-and-developer-facing version.

Neither looks like a traditional software IPO. Both ask investors to price a future infrastructure position, not merely trailing twelve-month revenue.

Anthropic pushes the same story one layer deeper.

It is clearly an AI infrastructure story. Models are getting stronger. Compute is getting more expensive. Cloud access and chip supply chains increasingly look like strategic resources. Reuters, citing Bloomberg, reported that Apollo and Blackstone were working on roughly $36 billion of debt financing tied to Anthropic, intended to support AI infrastructure expansion and the purchase of Google’s custom TPUs, which Anthropic would lease.

This is not a small detail.

We used to talk about AI companies with SaaS vocabulary: users, subscriptions, gross margin, net retention, CAC payback. Those words still matter, but they are no longer enough. Frontier AI companies increasingly look like a strange hybrid of cloud computing, telecom, energy, semiconductors, and financial engineering.

They sell intelligence. But first they have to buy chips, power, racks, cooling, land, networks, debt financing, and cloud capacity.

Seen that way, Anthropic’s near-trillion-dollar valuation is not an isolated madness. It belongs to the same capital market story as SpaceX and OpenAI. SpaceX is pulling public markets into the imagination of AI physical infrastructure. OpenAI is pulling public markets into the imagination of AI as a consumer entry point and intelligence platform. Anthropic is starting to pull private credit, enterprise deployment, and real-economy demand into the same frame.

But even that framing is still too broad. It is correct, but not sharp enough.

The most easily overlooked signal in Anthropic’s recent news is not the valuation. It is not even the debt financing.

It is the enterprise AI services company Anthropic announced on May 4 with Blackstone, Hellman & Friedman, and Goldman Sachs.

Anthropic said the new company will work with mid-sized companies across sectors to bring Claude into their most important operations. Anthropic’s applied AI engineers will work alongside the new firm’s engineering team to identify high-impact use cases, build custom solutions, and support customers over the long term.

Blackstone’s announcement was even more explicit. The new company will embed Anthropic engineering and partnership resources into the team, while using the networks of alternative asset managers across hundreds of companies to design, build, and maintain enterprise AI deployments. Blackstone’s president Jon Gray said the firm hoped to deploy Anthropic’s technology across its portfolio and the broader enterprise market; Goldman Sachs also highlighted mid-market companies and portfolio companies in its asset management business.

This is where the story gets interesting.

Private equity and alternative asset managers are not just capital providers here. They may become a distribution layer for AI into the real economy.

For the past two years, we have mostly understood AI distribution from a Silicon Valley angle. ChatGPT captured individual users first. Developers adopted APIs. Teams experimented from the bottom up. Companies then moved toward centralized procurement. The path looked familiar, almost like the consumerization of enterprise software: personal productivity first, team collaboration next, enterprise IT budget later.

OpenAI is still the king of that path.

In its March financing announcement, OpenAI described ChatGPT’s broad consumer adoption as a powerful distribution channel into work. It framed consumer adoption, enterprise deployment, developer usage, and compute access as parts of the same flywheel.

But OpenAI has also realized that self-serve product growth and API access are not enough. On May 11, it launched the OpenAI Deployment Company, majority-owned and controlled by OpenAI, with more than $4 billion of initial investment. The purpose is not to launch another model. It is to help enterprises deploy AI into real use cases, redesign critical workflows, and work with consulting and systems integration partners on change management. The partnership is led by TPG, with Advent, Bain Capital, and Brookfield as co-lead founding partners.

That tells us something important.

The next enterprise AI battle is not about winning a few more points on model benchmarks. It is not about showing the most impressive demo video. It is about who can enter a company’s workflows, permissions, data systems, organizational structure, and budget process.

Model capability is the entry ticket. Deployment is where value gets realized.

OpenAI and Anthropic both understand this. The difference is where they start.

OpenAI has ChatGPT as a consumer entry point. It can seep into companies through individual habits. One employee uses ChatGPT to write emails, summarize research, clean up code, or draft a deck. That habit spreads across a team. The team pushes the company to buy enterprise access. This route is like water slowly moving from the surface into the ground. It is slow, but broad.

The OpenAI Deployment Company is a reinforcement of that path, because even the strongest individual habit cannot automatically solve enterprise permissions, audit trails, compliance, data isolation, and workflow integration.

Anthropic does not have the same level of consumer habit.

Claude has a strong reputation among developers, enterprise users, safety-conscious buyers, and increasingly in coding workflows. But it does not have ChatGPT’s mass-market mindshare. So Anthropic needs another path. Not from individual users upward, but from enterprise control and operating pressure downward.

This is where PE-backed companies matter.