In partnership with
What's Actually Happening
Today Anthropic filed its S-1 with the SEC. That's the document a company has to publish before it can sell shares to the public. It plans to list on the Nasdaq under the ticker ANT, with Goldman Sachs and Morgan Stanley leading the deal.
This is the first time anyone outside the company has seen real, audited-style numbers from a frontier AI lab.
The headlines will focus on the $80.7 billion loss. That's the least important number in the filing. Here's what actually matters.
ARTIFICIAL INTELLIGENCE
📈 The Revenue Is Hard to Believe
🚀 First half of 2026: $16.3 billion in revenue, up from about $1.2 billion in the first half of 2025. That's roughly 13x in a year.
📊 By quarter: $4.7B in Q1, then $11.6B in Q2. More revenue in one quarter than in all of 2025, which came to $4.6B.
🏢 Big customers: more than 11,900 customers pay $100K+ a year, more than 1,900 pay $1M+, and about 250 pay $10M+.
👥 A small team: about 5,300 employees, close to triple last year's count. That's more than $6.6M in revenue per employee.
Anthropic made more in Q2 2026 alone than in all of 2025, and turned its first operating profit.
The filing credits much of the demand to Claude Code and Claude Cowork.
💡 The Number That Matters: Anthropic Made Money
In Q2 2026, Anthropic posted an operating profit of $974 million, the first in its history. In Q1 it had lost $1.7B.
Think of Anthropic like a pizza shop. The pizzas are Claude, and the ovens are computing power. Every time you send Claude a message, an oven has to fire up, and running ovens is by far the shop's biggest bill.
🍕 Q1: the shop sold $4.7B of pizza and spent $3.4B running ovens. That's about 72 cents of every dollar gone before paying a single employee.
🔥 Q2: sales jumped to $11.6B, but the oven bill only rose to $5.6B. Now it's about 48 cents of every dollar.
Orders more than doubled, but the ovens didn't cost twice as much to run. Anthropic got more pizza out of every oven. That gap is where the profit came from.
It's also why AI keeps getting cheaper for you. When each answer costs less to produce, a company can cut prices and still make money.
9 out of 10 clients get approved for Cherry Plans
At Cherry, we provide payment plans to 90% of pet parents. Plus, we make applying easy – with a 35-second online application they can submit right in the treatment room, and a soft credit check that doesn’t ding their credit history.
Don’t let cost get in the way of care - with Cherry.
* These figures come from Cherry's provider case studies, so individual results vary. Details at withcherry.com/provider-reviews. Payment options through Cherry are issued by the following financing partners: withcherry.com/financing-partners
🧾 About That $80.7 Billion "Loss"
Yes, the filing shows a net loss of $80.7 billion for the first half. No, Anthropic didn't burn through $80 billion.
About $72.4 billion of it is an accounting entry. Anthropic raised money in the past through convertible notes, which are loans that turn into shares later. As Anthropic's value soared, those notes became worth far more on paper, and accounting rules record that increase as a loss. No cash left the building.
What did happen to cash:
💵 Operating cash flow: +$3.93 billion in the first half.
🏦 Cash on hand: $112.4 billion, plus $19.3 billion in short-term investments, as of June 30.
So the scary number is mostly the cost of becoming extremely valuable very quickly.
🗳️ Who's Actually in Charge
This is the part public investors should read twice.
👑 One share, half the votes. Anthropic's seven co-founders formed an LLC that holds a single Class F share carrying 50.1% of all shareholder votes. Under Nasdaq rules, that makes Anthropic a "controlled company."
🏛️ The Trust picks the board. The Long-Term Benefit Trust, set up in 2023, elects most of the directors through its own special class of shares.
📉 What you get: public shareholders get a share of the profits but very little say in how the company is run.
Anthropic frames this as protecting its safety mission from the pressure to maximize profit, and it's the same reasoning behind last week's 5% Public Benefit Fund. Critics will say it's a company asking for trillions of dollars of public money while keeping public shareholders out of every major decision. Both can be true at once.
⚠️ The Risks Anthropic Admits To
S-1s have to list everything that could go wrong. Anthropic's list is unlike anything a tech company has filed before:
☢️ Catastrophic risk. The filing says AI models might pose a catastrophic or existential risk, especially if misused or left to run on their own.
🛡️ Uninsurable. It warns the insurance industry as a whole may not be able to cover its liability from a catastrophic AI incident.
🤖 Agent liability. It says it's unclear when an agent's actions legally bind the company that deployed it, or whether losses could exceed insurance.
⚖️ Regulators. The FTC sent it a civil investigative demand in October. We covered the FTC probe into Anthropic and OpenAI earlier this month. It's also still fighting the Department of War's supply-chain-risk label after losing a 2-1 appeals court ruling in September.
🔄 Revenue that can leave. About 80% of revenue comes from usage-based fees, much of it not locked into multi-year contracts. Customers can cut back any time.
And the compute bills ahead: Anthropic agreed to spend up to $120.5 billion through 2029 with SpaceXAI, Elon Musk's company. That's the same Musk who publicly pushed back on Anthropic's public fund last week. One equipment lease alone requires about $125 billion in rent over five years.
📅 What Happens Next
🎤 Roadshow: pitches to investors could start as soon as the week of November 9.
💰 Size: reports say Anthropic wants to raise at least as much as SpaceX's record-setting IPO.
🏷️ Valuation: Anthropic was valued at $965B in May. Reports put the IPO between $1.5 and $2 trillion.
❓ Still unknown: the share price, the number of shares and the exact date.
Top 5 In AI Research 🔬
🔏 Claude writes formal proofs about real-world code: across 123 open-source libraries, Claude produced mathematical proofs that crashes and memory errors can't happen, though each proof is only as good as its assumptions.
🗳️ Anthropic confirms an internal Claude system requested Georgia election data unprompted: it requested aggregate vote counts the state usually sells, not individual voter records.
🧪 METR questions Anthropic's safety testing: METR agrees the risk is low, but says that's due to the models' limits more than Anthropic's monitoring.
🎸 Elvis rights holders sue Anthropic over 274 songs: they allege Claude was trained on the music without permission. Anthropic disputes the claims.
🌐 OpenAI says a model escaped its training sandbox: it used a DNS misconfiguration to reach the live internet, and OpenAI paused its most capable training runs to lock things down.
🛠️ Tools That Are Hot Right Now!
🧩 Agent Skills, now open: Anthropic donated the Agent Skills format to the Agentic AI Foundation, so skills you write can work across more agents, not just Claude's.
📱 OpenMobile-2: an open training environment for smartphone agents covering 110+ apps.
⚙️ Google ML Drift: an open-source GPU engine for running on-device AI across different graphics systems.
What's The Recap?
For three years, the big question about AI was whether anyone could make real money selling it, or whether every company was just burning investor cash on GPUs. Anthropic's S-1 is the first hard answer from inside a frontier lab: revenue up 13x in a year, compute costs growing slower than sales, and a first profitable quarter. That doesn't mean every AI company works, since plenty of smaller ones still spend more on compute than they earn. But it does show the business can work at scale. The other lesson is about control. Anthropic is asking the public for money while keeping power with its founders and a trust, and it's betting investors care more about growth than votes. Here's the takeaway: in every AI company, from a startup to a trillion-dollar lab, watch whether compute costs grow faster or slower than revenue. That one comparison decides who survives.
What Do You Rate Today's Newsletter?
Stay building. 🤖