Anthropic Just Told Investors It Makes $65 Billion a Year — Now Comes the Hard Part

Anthropic Just Told Investors It Makes $65 Billion a Year — Now Comes the Hard Part

Let me tell you something that's been sitting with me since Monday. A company that wasn't even a blip on the public markets a year ago just told its investors it's on track to pull in $65 billion a year. Seven times what it was doing twelve months earlier. And the market's response wasn't a stampede to get in — it was a nervous shuffle toward a price tag that would make this the biggest IPO in the history of money.

I've been running hosting infrastructure for over a decade. I've watched every hype cycle this industry has ever had come and go — the dot-coms, the cloud wars, the crypto miners — and I've learned to separate the noise from the numbers. So let me be straight with you: the $65 billion number is real. The $2 trillion number is the part that should scare the hell out of you.

The Number That Changes the Conversation

Bloomberg reported Monday that Anthropic told investors its annualized revenue run rate hit $65 billion by the end of July. CNBC confirmed it the same day. That's up from $47 billion in May and roughly $9 billion when 2025 closed. Do the math: seven months, roughly sevenfold. The company also shared a preliminary revenue figure of about $11.5 billion for the second quarter of 2026 alone.

Here's the part that matters even more. That number puts Anthropic ahead of OpenAI. Reports put OpenAI's run rate around $40 billion, which means the lab that started as the safety-focused underdog is now clearing its rival by about $25 billion — roughly 62 percent. That's not a rounding error. That's a leadership change in the AI revenue race, and it happened in a single quarter.

Think about what $65 billion a year means in context. That's bigger than most of the enterprise software companies that took twenty years to build. That's more than the entire revenue of most Fortune 500 companies. And it's coming from a business that sells tokens — the same business model as a metered utility, except the fuel is silicon and electricity instead of gas and water.

For anyone who's been telling you the AI boom is all hype with no revenue behind it: you need a new argument. Enterprises are paying real money for Claude, and they're paying more every month. This isn't a demo. This is a $65 billion annualized check being written by actual customers with actual budgets.

The Other Number Nobody's Addressing

Now the hard part. Investors and bankers reportedly expect an October IPO at a valuation of $2 trillion or more. That would blow past SpaceX's $1.77 trillion listing in June to become the largest public offering in history. For context: Anthropic's last private round — a $65 billion Series H in May — valued the company at $965 billion. So in about three months, the price tag more than doubles.

Let me do the math for you, because nobody on the cable shows will. At $2 trillion against a $65 billion run rate, you're paying roughly 30 times annualized revenue. For a company that, by every indication, still burns cash heavily on a GAAP basis, whose entire product is renting intelligence that runs on someone else's chips, and which is simultaneously trying to buy a $6 billion Israeli AI startup — Decart — just to keep its own roadmap moving.

Ent? And I say this as somebody who believes the underlying demand is real. The revenue is real. But a 30x multiple on a run rate is not a valuation. It's a bet that the growth curve stays vertical for the next five years without a single stumble — no pricing war, no open-source challenger, no compute shortage, no regulatory shock. That's a lot of things that have to go right.

The Toll Booth Nobody Wants to Talk About

Here's the reading I don't hear on any of the financial shows, and it's the one that matters most if you actually run infrastructure. Anthropic doesn't own the compute. It rents enormous volumes of it — from Google, from Amazon, from Nvidia and the whole GPU cloud layer. And here's the uncomfortable part: the same companies are also its investors. Google and Amazon both run the clouds where Claude is trained and served, and both hold stakes in the company. Nvidia is in there too.

That's the circular piece of the AI economy I keep coming back to: the landlords of the AI buildout are also the shareholders of the biggest tenant. When Anthropic reports a $65 billion run rate, a meaningful slice of that revenue is a pass-through — token sales on top of rented GPUs, with gross margins that depend entirely on how cheaply the hyperscalers let them lease the hardware that's in short supply.

This is the secondary bottleneck nobody's watching. Not chips. Not power. Not water. The margin stack between the model layer and the compute layer. The moment Anthropic has to file a public S-1 and then report quarterly as a listed company, the entire world gets to see exactly how that stack works — and I promise you, the number that moves the market won't be revenue. It'll be the cost of goods sold.

What This Means for Independent Hosting Providers

If you're running a hosting business, a colo, or any independent infrastructure operation, here's what the next six months look like:

First, treat the run rate as demand confirmation, not a license to overbuild. A $65 billion run rate means enterprise AI demand is real and it's going to keep buying GPUs, power, and network capacity. But it doesn't tell you what demand looks like in 2027. The hyperscalers will use this number to justify even more capex. You should use it to be careful about debt-funded expansion — the same way you'd treat any customer who shows you a hockey-stick forecast.

Second, watch the margin disclosures like a hawk. When Anthropic's S-1 lands, the gross margin line will tell you whether the model layer is a toll booth or a money pit. Thin margins mean the squeeze flows downstream — cheaper token pricing, tighter margins on resold compute, pressure on everyone in the middle of the stack. Fat margins mean the hyperscalers raise the rent and capture more of the value. Either way, the pricing signals ripple through every colo and cloud in the industry.

Third, position as the independent layer. The more the model companies squeeze their own margins, the more value moves to the people who own real infrastructure and real customer relationships. Enterprises are already asking hard questions about concentration risk — one model vendor, one cloud, one region. Independent hosting is the answer to that question, and this IPO cycle is the moment to start making that argument to your customers.

The Transparency Shock Is Coming

Here's the structural reality. For the last two years, the AI buildout has run on private numbers: private valuations, private revenue disclosures, private compute deals. The $65 billion run rate was shared with investors in a routine update — it wasn't filed with the SEC, and nobody outside a small circle of funds and bankers saw the underlying books.

That's about to change. When Anthropic goes public — and every signal says October — the AI industry gets its first giant, GAAP-filed, audited financial statement. And with it, the entire AI capex thesis gets its first real report card. Every hyperscaler capex number, every GPU backlog, every circular financing arrangement suddenly gets priced against actual, auditable revenue from the industry's biggest customer.

I've said it before and I'll say it again: this isn't a warning about something that might happen. The run rate is already here. The IPO date is already on the calendar. The only question is whether the market prices the $2 trillion as the beginning of a new era or the top of a very expensive trade.

The Bottom Line

Here's what I know. A company that was begging for compute a few years ago is now on pace to become the largest public company debut in history. That's not a bubble story. That's a demand story — and it's real.

But buh, let me tell you something, ent? The number that made the headline — $65 billion — is a projection. The number that will define the trade — $2 trillion — is a bet. And in between sits the part of the business nobody's talking about: who owns the compute, what it costs, and what happens when the whole world finally gets to read the margin line.

Plan for the transparency shock. Lock your costs. And don't blink when the S-1 lands.

— Allan Ali, Founder

This article was produced with AI-assisted research and editorial support. Sources: CNBC Television (Anthropic revenue run rate report, Aug 18, 2026), Bloomberg (Anthropic revenue run rate surpasses $65 billion ahead of IPO, Aug 17, 2026), Financial Times (Anthropic investors bet on $2 trillion valuation in record IPO), Fortune, Reuters, TechCrunch, TipRanks, Benzinga, StockTwits (August 2026).