Most of the AI companies burn investor money and call it growth. Anthropic is the rare exception, the revenue is real, and they are expecting a profit in Q2.
Its run-rate (annual projected revenue) went from about $9B to roughly $47B (5x) in five months. It then raised $65B at a $965B valuation and filed to go public.
But the compute behind all of it is rented, about $15B a year, from a competitor SpaceXAI, cancellable on 90 days' notice, and this is not the only deal Anthropic made.
So how solid is the most solid business in AI?
Model breakdown #002 Anthropic snapshot
01. The Tech
Anthropic's models are genuinely good and they are well known for reasoning and coding tasks. They have four categories of models: Haiku < Sonnet < Opus < Fable, with each specializing in different tasks.
Just a few hours ago, they released Fable 5, a fine-tuned version of the Mythos models for public use with strict guardrails, calling it their best public model release so far.
But I believe the AI race is now beyond good models. It is now about the ecosystem you create around them.
Here, Anthropic clearly has a rich ecosystem. Alongside proprietary products such as Claude Code, Cowork, Claude Designs, and the Claude chatbot itself, there is also a growing ecosystem of open-source integrations built around Anthropic's technology.
Anthropic also offers the Claude Agent SDK, making it easier for developers to build agentic applications on top of its models. This further strengthens its position within the developer ecosystem.
Anthropic introduced MCP (which was later donated to the Linux Foundation), Skills, and Extensions. These have now become industry standards.
This gives Anthropic soft power over its competition while also supporting its philosophy of Constitutional AI.
That is why I believe Anthropic will increasingly serve governments and large enterprises in the future, much like Microsoft has done for decades.
02. The Business Model
Anthropic runs a hybrid model: subscriptions, enterprise seats, and usage-based API revenue.
The good news is that the revenue appears real. Reports suggest the company is approaching its first operating profit in Q2, which is rare in AI. Most competitors still rely heavily on investor subsidies.
The pressure, however, sits beneath the headline numbers.
Anthropic's fastest-growing products are also its most expensive to serve. Coding agents consume enormous amounts of compute, and compute costs are not falling fast enough. The company reportedly committed roughly $15B per year for rented compute capacity through 2029.
That means revenue scales.
But costs scale too.
The AI economy is a circle: the same dollars get booked as investment, revenue, and compute spend, but Anthropic seems to earn some profit out of it.
03. The Real MOAT
Anthropic has some genuine moats.
Currently, around 41% to 46% of all production code globally is generated by AI. Of that, Anthropic commands 41% to 54% market share, thanks to its powerful models.
Workflow Lock-in: AI-assisted coding is becoming the new normal, and Anthropic is a key player in that shift. As a result, it benefits from workflow lock-in. The cost of switching tools for engineering teams is high, especially once AI becomes deeply integrated into their development workflows.
Soft Power: Anthropic plays its game around Constitutional AI, which could give the company an edge in government and large-enterprise contracts.
This is the soft power Anthropic enjoys.
MY TAKE
Anthropic is one of the rare AI labs that is expecting to turn a profit in Q2. This will give investors confidence in Anthropic and, more broadly, in the AI industry itself.
That said, I am still not convinced by its valuation.
Just a few hours ago, Anthropic released Fable-5, a Mythos-class model for public use and what it calls its best public model so far. But the timing is interesting.
Over the past few days, the company reportedly raised $65B in a Series H round and confidentially filed paperwork for an IPO. Speculation suggests the IPO could happen in the third week of October 2026.
Because of that, it is worth asking whether some of the hype around Mythos (Fable) is tied to the upcoming IPO.
That is not to deny the fact that these are state-of-the-art models. But the timing is difficult to ignore.
04. ONE THING TO WATCH
Whether Anthropic can maintain profitability across consecutive quarters?
That will depend heavily on how effectively the company manages its computing requirements. Revenue is growing, but so are the costs of serving increasingly powerful models and coding agents.
That's the teardown.
Reply and tell me your feedback and which AI startup you want torn apart next. I read every one.
- Abhishek
P.S. If you found this useful, forward it to a founder who's tired of the hype.