Anthropic files for IPO, joining OpenAI and SpaceX in the race to go public

[Julio Franco]

Why it matters: Anthropic's filing turns the AI race into something Wall Street can finally measure. For the last few years, frontier AI companies have been valued by private investors on growth, ambition, and the fear of missing the next platform shift. A public filing will eventually force Anthropic to disclose the numbers that matter most: revenue, losses, infrastructure costs, margins, and how much money it takes to keep Claude competitive.

Anthropic has officially started the process of going public, confidentially submitting a draft registration statement to the US Securities and Exchange Commission for a proposed IPO. The company did not disclose how many shares it plans to sell or at what price, and said the offering will depend on market conditions and other factors.

The timing is striking but not unexpected. Anthropic recently raised $65 billion in a Series H funding round that valued the company at $965 billion, putting it ahead of OpenAI's most recently disclosed $852 billion valuation. That makes Anthropic one of the most valuable private companies in the world and potentially one of the biggest technology IPOs ever if the listing proceeds near that level.

It also gives Anthropic a narrative advantage over OpenAI. After months of speculation over which AI lab would go first, Anthropic has now moved ahead. OpenAI is also expected to go public, while SpaceX is preparing a blockbuster offering of its own. Together, the three companies are bound to test whether public investors have enough appetite for an unprecedented wave of mega-cap tech listings built around AI, compute, and the data center infrastructure build-out.

Because Anthropic submitted "confidentially," investors cannot see the details that will determine whether the company's valuation is grounded in durable economics or another chapter in the AI hype cycle. Analysts are already focusing on gross margin, cash flow, and the cost of serving increasingly powerful models, especially as AI companies spend aggressively on chips, data centers, and cloud capacity.

That is the central tension around Anthropic's IPO. Claude has become a serious enterprise AI product, especially in coding and professional workflows, but the company still operates in a brutally expensive market where every advance requires more compute and every rival is raising at historic scale.

For now, Anthropic has raced to the starting line. The bigger question is whether it can convince investors that the AI boom is not just growing fast, but eventually capable of producing profits large enough to justify valuations approaching $1 trillion.

Permalink to story:

Anthropic files for IPO, joining OpenAI and SpaceX in the race to go public

 

[RedGuard]

Huh, this feels like putting a name to the AI bubble.

 

[OortCloud]

They've bled the private equity dry now they go public, cash-out and bail before the reality of the cost of AI versus the revenue hit home and the huge debt owed comes due.

 

[Plutoisaplanet]

They've bled the private equity dry now they go public, cash-out and bail before the reality of the cost of AI versus the revenue hit home and the huge debt owed comes due.

Unlike xAI and OpenAI, Anthropic didn’t make massive infrastructure commitments. They had become compute constrained until they signed a deal to use xAI’s Colossus data centers: https://www.networkworld.com/articl...e-of-ai-compute-as-a-standalone-business.html

 

[yRaz]

Unlike xAI and OpenAI, Anthropic didn’t make massive infrastructure commitments. They had become compute constrained until they signed a deal to use xAI’s Colossus data centers: https://www.networkworld.com/articl...e-of-ai-compute-as-a-standalone-business.html

It could be argued that signing a deal with xAi is a massive infrastructure commitment. It's the same thing but with extra steps

 

[mbrowne5061]

They've bled the private equity dry now they go public, cash-out and bail before the reality of the cost of AI versus the revenue hit home and the huge debt owed comes due.

Exactly. IPOs are now the investor off-ramps, not the "final round" of funding for new companies as they get established.

Unlike xAI and OpenAI, Anthropic didn’t make massive infrastructure commitments. They had become compute constrained until they signed a deal to use xAI’s Colossus data centers: https://www.networkworld.com/articl...e-of-ai-compute-as-a-standalone-business.html

Ahh. So Anthropic doesn't have infrastructure commitments... just infrastructure dependencies.... on a company with massive infrastructure commitments that is themselves trying to take advantage of the same hardware they've committed to someone else. That certainly sounds sustainable and stable, and like there is no risk to Anthropic here.

Look, maybe in next couple of years, someone will come up a more efficient processor, or more efficient way to train and operate these models. But I doubt it. These models aren't new. They've been around since the 80s; they were too computationally to run at all until the 2000s~; the 2010s saw them get developed in university labs; and they finally 'broke containment' from universities at the end of the 2010s/early 2020s.

So, ~40 years of mathematics development, ~20 years of hardware development, and ~10 years of software development, and we are still stuck with something that is incredibly expensive to run and doesn't scale well. We're not seeing efficient LLMs for quite some time.

When they do figure it out (or computer finally outpaces what LLMs can reasonably use), I suspect you are touching upon how the industry will shake out, however. There will be companies that exist just to setup, maintain, and sell hardware compute services; and there will be companies writing software that utilizes these services (including LLMs). Having a market of vertically integrated AI companies does not seem sustainable to me. Too much duplication of infrastructure, driving up costs. Like how no one hosts their websites on their own hardware; too expensive when everyone tries to do it themselves.