The Revenue Story Is Real — And Staggering

Since we covered Anthropic's confidential SEC filing, the financial picture has snapped into sharper focus.

According to CNBC and TechCrunch, Anthropic's revenue run rate has hit $47 billion — up from $10 billion in annual revenue last year. That is NOT a rounding error. That is a company that roughly quintupled its revenue trajectory in under 12 months.

The driver, per multiple sources including NYT and CNBC, is largely Claude Code — Anthropic's AI coding assistant. Enterprise customers are paying real money for it. This isn't hype-driven lab revenue. It's recurring software spend.

But They're Still Losing Money

Wired reported it plainly: Anthropic "spent more money on cloud computing and thousands of staff, leading to losses."

So we have a company burning cash at scale, racing to go public before the market window closes. That's not a scandal — that's how frontier tech works. But investors need to price that in, not just the $965 billion valuation headline.

The $965 billion post-money valuation — confirmed by NPR, TechCrunch, and Wired — actually tops OpenAI's $852 billion figure from March, according to CNBC. Anthropic is now, by that measure, the most valuable private startup on the planet.

The Security Number Wall Street Missed

According to VentureBeat, Anthropic's own safety disclosure for its Claude Opus 4 model — a 244-page document released May 28 — revealed that in browser agent testing, an attacker successfully hijacked the model 31.5% of the time before safeguards engaged.

In a coding environment, that number drops to 7.03% on single attempts, pulled down to 2.09% with safeguards active.

Anthropic's defenders will correctly note that publishing this data at all puts them ahead of every competitor. OpenAI reported only one surface. Google moved the whole subject off its model card into a separate document. Meta shipped NO closed-model card at all, per VentureBeat.

But a 31.5% browser hijack rate demands serious scrutiny. Any enterprise deploying Claude agents in browser environments needs to understand that an adaptive attacker can compromise the model nearly one-third of the time before the system catches it. Adam Meyers, Senior VP of Counter Adversary Operations at CrowdStrike, told VentureBeat that AI deployment expands attack surfaces faster than legacy defenses respond. This is real exposure, not theoretical.

CNN, NYT, NPR, and most mainstream outlets treated the IPO filing as a pure finance story — without a single mention of the security disclosure. That's a notable gap in coverage.

The Three-Way IPO Race Gets More Complicated

The competitive landscape has new texture since our last coverage.

SpaceX is targeting a June 12 Nasdaq debut under ticker SPCX, according to CNBC and The Verge. It's seeking to raise more than $75 billion at a targeting valuation of roughly $1.75 trillion, per Wired and Reuters reporting cited by Wired. That would be the largest IPO in history.

But TechCrunch reported a buried line in SpaceX's IPO filing: the company warned investors it "may issue a significant amount of equity in connection with future transactions." That language was added amid industry speculation about a potential SpaceX-Tesla merger — something Elon Musk has floated for years. SpaceX's share structure gives Musk 10 votes per share on his Class B stock, meaning he could execute a dilutive merger without meaningful shareholder resistance on the SpaceX side.

OpenAI, meanwhile, is still preparing its own confidential filing. CNBC reported it was "readying" that filing as of June 1. Wired cited a September IPO target. The Polymarket prediction market, cited by ZeroHedge, was genuinely surprised by Anthropic moving first — suggesting the street expected OpenAI to jump the queue.

Wedbush Securities analysts told NPR this represents "an opening of the floodgates for the IPO market" after years of dormancy. Three potential mega-listings — Anthropic, OpenAI, SpaceX — all in a compressed window. Markets haven't seen anything like this.

The SpaceX-Anthropic Deal Everyone Forgot

The Verge flagged something worth noting: Anthropic has a $15 billion per year deal to use SpaceX data centers. These two companies are simultaneously racing to IPO AND locked in a major infrastructure partnership. That's an unusual dynamic — competitors in the capital markets narrative, customers and vendors in the actual business.

It also means Anthropic's cost structure is materially tied to Musk's infrastructure empire. That's a risk factor investors will want disclosed in the eventual public S-1.

What This Means for Regular People

If you own index funds, you will almost certainly own pieces of all three of these companies within 12 months. The AI infrastructure bet is being made FOR you whether you like it or not.

If you work in enterprise IT or security, the VentureBeat security data is NOT something to dismiss as a PR problem. A 31.5% browser-agent compromise rate before safeguards engage means your procurement and security teams need to ask harder questions before deploying agentic AI at scale.

And if you're a taxpayer wondering whether this AI boom is real or another dot-com bubble — the revenue numbers suggest it's real. But Anthropic is still losing money. At a $965 billion valuation. On $47 billion in run-rate revenue that didn't exist in any meaningful form 18 months ago.

The math will eventually have to work. It doesn't yet.