Update: this inspired a note on my blog: https://simonwillison.net/2026/May/29/anthropic/
This one client, then, is 12.8% of Anthropic's run-rate revenue? That does not exactly fill me with confidence that it's a meaningful number. Doesn't this suggest run-rate revenue will fall off a cliff if Anthropic customers start applying cost controls?
If they really did count that one customer as $6B it means they've gone from $30B in April to a mere $41B in May.
im really not sure why he keeps parroting on about this. its all irrelevant frankly. companies play games. non-gaap revenue recognition, adjusted operating income.... boo-ya.
wait for the somewhat official doc's to come out, then its worth talking about.
As an anecdote, Github is changing their copilot plan to usage based billing next month. They released a tool that allows their users to estimate what their bills will look like under the new plan based on their past usage. There are some screenshots online from users showing their plans will go from $40/month to $3-5k/month. I imagine this is happening everywhere. These tools absolutely can do more than they were capable of just six months ago. But if the true costs are as high as it appears, folks are going to be much more judicious with how they use them moving forward.
People confuse price with cost all the time. The price of Opus has dropped from $75/1M output tokens to $25. That's the price. The cost is much lower and according to Dario, about a month ago, they had about a 73% margin.
I don't understand how anyone would use GitHub copilot...it's basically running a custom harness and using close to API pricing for Opus. This is why Microsoft is cooked in this game.
But yeah, I don't understand why people switch from subscription to API prices for Claude. They're way higher, but again that's price and not necessarily the cost to Anthropic to serve.
Do you have a source for this? I missed this and it does not match my impression.
And FYI, most enterprise accounts were forced to switch to a hybrid monthly seat license plus API based usage earlier this year. So that is why we are probably seeing so much alarm over Ai bills at the enterprise level. Companies went whole hog on agentic workflows not fully appreciating the costs structures of their new plans. Didn't help that pretty much every VC and board was probably breathing down their neck that if they didn't jump on the AI bandwagon they would get left behind.
There were a lot of people (here included) that were absolutely abusing github - getting Opus to generate its own subagents for days on end with a single premium request. If THAT'S $3k/mo, I'm honestly not worried.
I was just asking sonnet, 5.4, opus, in single agent session to fix a problem for 20 minutes...if THAT'S $3k/mo, then AI is truly cooked.
I think for your use case, the most likely outcome is that you are going to need to be on a $100-200 month plan if you want access to the cutting edge models. But on the other end of the spectrum, you could probably get closer to $10-20 month using Chinese models. My usage was closer to yours, with the occasional tokenmaxxing sprint just to experiment a bit and test out the limits of the plan. I am not quite sure what I will be doing next month once it moves to API billing but I suspect I will move to openrouter and one of the open source CLI harnesses.
If you think investment banks will catch them during the IPO: They have launched many overpriced IPOs in 1999/2000 and also want a high IPO price.
Historically, listing rules asked: "How much money did you make last year, and do you fit our standard corporate governance box?"
Today, NASDAQ's rules ask: "Do you have the massive market capitalisation, sufficient institutional public float, and transparent liquidity to ensure fair and orderly trading?"
Here are the obvious ones:
1. Free float - Every company that intended to list was required to have at least 20% free float.
2. Index weights were based on the free float.
3. Time before inclusion into the Indices (min 12 months) now 15 trading days
4. Lockup period - minimum 12 months up to 24 months - not 180 days.
I wish people would understand that if America had a functioning criminal justice system, no one would have heard Elon Musk nor Donald Trump.
"The passive funds holding trillions of dollars of 401(k)s and other investments are rushing to change their rules as the IPOs of SpaceX, OpenAI and Anthropic draw closer."
Those index providers are the same interest class with VCs. With such moves they inflate demand post-IPO (hoping it holds for 180+ days), but also allows them to lure buyers in private secondary market and offload that shit pre-IPO.
https://www.wsj.com/finance/stocks/stock-indexes-are-contort...
ARR seems like a interesting concept which hides away the concrete details and is counting on a futuristic possible commitment especially for companies at this early stages. The crucial detail is probably the money being spent, that is what is fueling this sale of equity at this point.
Yes, it is a meaningful financial measure of actual progress, exactly where progress is most needed for a new company.
> which hides away the concrete details
A normal metric isn't a magic trick. It's just the number it is.
> The crucial detail is probably the money being spent
Accelerating (not just fast) revenue growth at an astounding rate, with absolute numbers that are enormous for a new company, bonkers for a 2500 employee startup [0], is the crucial "detail".
There are lots of companies with deep pockets and compute, making great AI efforts with teams of smart people, that would love to be doing, but are not doing, what Anthropic is doing and doing well. That might be the second crucial detail.
ARR as a number for steady business or public company is much simpler, ARR for companies at this early stage fueled by the motivation for IPO is more than just a number. Its an attempt to convince the market for a certain outcome without revealing the books
> Accelerating (not just fast) revenue growth at an astounding rate
This acceleration is fueled by somewhat artificial demand (AI mandates by executives across the board, still figuring out the realistic use cases) and subsidized pricing. The expenses matter cause they are the ones which are going to dictate if the business is sustainable or not.
Of course ARR is simpler for less dynamic companies. Tautology.
Of course a PR is a communication to the market. Also a tautology.
You are framing functional numbers and behavior as fraudulent because... a newsy PR isn't a formal disclosure?
Cynicism is fine. However: accusations should reference some evidence, not just distrust.
When one is under the grip of irrational exuberance and FOMO due diligence goes out the window.
For example look at Sequoia’s investment into FTX and article about Sam that accompanied.
If for no other reason than they would rather spend their days procrastinating on everything else, so they can obsess over some of the most interesting numbers they will ever see.
Not to mention even a total shitshow from an obvious crackpot like Theranos got $1.2 billion total funding, and a 9B valuation. Or FTX.
What we are talking about is entirely based on this one term in the announcement: run-rate revenue. This is a meaningless term just like how “clinically proven” for vitamins is a way for the companies to use weasel words to imply something without truly being able to back it up, but also not actually lying. There is no legal definition of “clinically proven” so, what exactly would you sue them for? The same thing is true for run-rate revenue. They can cherry pick numbers to use to generate the run-rate revenue value and they are not lying, but this isn’t exactly honest either. We have no transparency and run-rate revenue is not an accounting term.
This goes back to my entire point, this is a vibe. This announcement does not provide much in the way of substance, so a reader will take it to say whatever they want.
IDK if these tricks would work anymore but then again fraud is legal now so who knows.
In the US?
Hahaha, that was a good one, buddy.
> They are not obligated to be truthful here
1) I have a new extra cost 2) How does that make me more profit, and improved cashflow
I know I'll be bombarded with metrics about productivity, feature completion, bug fixes etc. But someone is going to have to tell me how that equates to more sales. And who is going to do that? I'll be worried that the feature wishlist will just creep up now we can handle more throughput, and yet everyone will be telling me that I don't need to worry about the lack of new customers this quarter, one more new feature and we will catch up in Q4. You can see how that could make one grumpy. Then when the sales don't come, I'll tell the CTO that they need to balance the books, if they want to use expensive tools to make each developer more productive then they will need fewer developers...
...but I don't want that, I want more great features that people will pay for, and faster. But they have to pay for themselves.
Existing corporate command and control has optimized for people control, because people cost money and performed work. Control their assignments, and you control costs and what's worked on.
Widespread unmetered AI turns this on its head, because suddenly each employee is directing their own work and the AI spend that comes with it.
F.ex. Bob in accounting may think it's a brilliant idea to rebuild Lotus 1-2-3.
That may help Bob, but 10x'ing Bob's spreadsheet output doesn't change the company's profitability, because it wasn't a limiting factor. It was to Bob, but not to the company's revenue generators.
HN didn't like this article when it hit, but I thought it made a good point that corporate C2 is going to be the first thing AI adoption breaks: https://www.forbes.com/sites/jasonwingard/2026/04/23/vibe-co...
Increasing AI spend without profitability improvements is a symptom that C2 is failing (or was insufficient to begin with).
Seen through a charitable "CEOs know what the fuck they're doing" lens, the preemptive layoffs are about forcing AI efficiency gains in areas CEOs expect them: instead of allowing those departments' remaining employees to build their own apps, they're forced to deploy AI to cover for their missing 3 team members.
Unfortunately, the layoffs were executed before there were solid results about which departments could benefit from AI use (and without a plan for continuity of institutional knowledge), so... we'll see.
I get what you're saying about how every employee becomes a liability when they are let loose with an AI, and totally agree. But I think the layoffs have a much more financial root because they are so widespread across so many companies and even industries (not just tech)
Layoffs + access to AI forces most of that AI use to make up for the layoffs.
Which is a pretty shitty way to burn your employees out... but it has a method.
(Also, I'd say AI infra companies vs everyone else are apples to oranges. This point would be more applicable to a non-AI infra company that's paying for AI use rather than AI infra capital)
Turn a potential existential competition into (a) maybe I win or (b) I commoditize the market and you can't make money at it either.
How'd Microsoft's mobile efforts go?
Sometimes the lead quickly becomes insurmountable, so if you're sitting on a cash flow machine it's not a terrible idea to at least keep pace with new markets.
In my new job I haven't written a single line by hand. I now almost entirely work in claude code / codex and in github PRs. Occasionally I open vscode to read code, but very rarely. My company probably spends ~$100 a day for my token use. I'm not even going crazy with parallelization or subagents and such.
I 100% believe the demand growth based on my personal experience.
like when you owe a bank a billion dollars, thats the banks problem.