Dissecting Anthropic
Why now? Is this it? Who benefits?
The AI bubble gained a few more weeks or months of life as a result of Anthropic’s announced Initial Public Offering (IPO) actually seems to be materialising with its prospectus now available to everyone. Most of the takes have thus far been cautiously optimistic, the business is not doing well but not as awfully as sceptics would have guessed either. Still, for any analyst willing and able to think beyond the confines of spreadsheets, filings, and accounting it raises some tough questions. Why now?
Let’s begin by giving Anthropic some credit, earlier this year I wrote a post asking where all the AI IPOs were, making the case that the most plausible reason for AI labs’ reluctance to going public implied a fundamentally poor business. I never expected Anthropic’s IPO to happen, and while it is brave of them to go through with it, I can’t say that the prospectus has relieved any worries about the labs’ business model. Still, it is undeniably a shot in the arm for the AI sector.
First point to make is that the company is not profitable. The bleeding of cash has not stopped, but at least slowed recently, but it remains unprofitable To be fair, it has had one (1) profitable quarter in its existence, Q2 2026, but the losses in Q1 outweigh the profits in Q2 meaning that the company has not had a profitable year yet. We don’t know how Q3 went, but as Gary Marcus pointed out, Anthropic curiously did not give any guidance, indeed it submitted a prospectus where figures for Q3 were literally missing. There are reasons to believe that Q2 was a one off.
All throughout the spring of this year, there was a frenzied discussion climate surrounding AI which led to some companies establishing internal leaderboards to see who used the most tokens, as a proxy for who was most proficient in working with AI. Obviously, incentivising employees to use more AI through a competition devolved quickly into people gaming the system by deliberately using a lot of tokens without actually needing to, leading to high AI costs with little to show for it. Uber burned through its yearly AI budget in [three months], Meta made career advancement conditional on AI usage. Tokenmaxxing definitely inflated AI usage, which directly leads to more revenue for Anthropic, making it seem like the patient was on the path towards full recovery. Almost quadrupling revenue between FY2025 and FH2026 is impressive, but as it was driven by an irrational usage surge it is unlikely to repeat. Secondly, Ed Zitron reported that Anthropic rented out SpaceX’s colossus data centre at a reduced rate in June which also boosts profitability in a deceptive way. Now that the price is back to normal, we should expect compute costs to increase at the same time that the tokenmaxxing trend’s demise produces lower revenue.
There is also the threat of Anthropic’s capital commitments as pointed out by Ed Elson. Anthropic has promised a lot of companies an incredible amount of money, and nothing suggests that the companies revenues will be able to cover those commitments, let alone its (so far illusory) profits. The potential return for an equity investor is far off given just how many businesses will have their hands in Anthopic’s pockets before anything trickles down to the bottom line.
Speaking of net profitability, time to stop circling the issue. Anthropic made an operating loss of almost $700 000 000 in the first half of 2026, at a time where I would argue it had ideal conditions for its business, what with subsidised data centres and tokenmaxxing. The company further had a net loss of $80 000 000 000 in the first half of 2026. Sure, that is largely due to accounting rules, but $80 billion is a lot of money.
This brings us to the bigger picture. Anthropic is one of the most hyped companies since the South Sea Company, and could have chosen to IPO at any time. Its rival OpenAI has chosen to postpone its IPO, and so we should expect this to be an option for Anthropic too, especially as the company’s prospectus reports that it has $112 billion in cash. Which brings us to the three main questions I’d like to pose in this post:
1. Why now?
2. Is this it?
3. Cui bono?
As I’ve argued, I think now, as in Q2 2026 being the company’s most recent quarter is the best possible time for the business to go public. The figures have never been better, and crucially, are likely to be worse in the future. Anthropic’s leadership have had all the advantages needed to pick the optimal time for an IPO in order to guarantee success and raise the most money possible, and this is what they give us? Again, under optimal circumstances the company made an operating loss of almost $700 million. If they thought they would remain profitable, they would surely wait a few more quarters and get a higher valuation, as profitable companies tend to be better liked by investors than the sort of anaemic money waster Anthropic looks to be.
If what I’ve laid out is the answer to the question ‘why now?’ is this it? Is this all that however many billions Anthropic has raised gets you? Is this all the golden boy of the new economy, nay the largest disruption in the history of capitalism dwarfing even the industrial revolution, can muster? If so, can you in good faith argue that the AI industry has legs? To really stress this, under optimal circumstances for Anthropic, you would make a higher profit in percentages and in dollars by running a hot dog stand.
The key question in political economy, which is exceptionally applicable here, is cui bono? Who benefits?
With 22 rounds of pre-IPO investment rounds, the answer is venture capital firms who have gone hog wild on AI start-ups. Shareholders from series A through H-4 have a lot of shares to sell, and as Anthropic has been generous to its employees in terms of compensation through stock options, there are a lot of people waiting to cash out on this IPO. This means that there are both top-down pressures from external investors and bottom-up pressures from employees to IPO at peak possible valuation.
We should also consider that there are businesses who may not have direct equity stakes in Anthropic who benefit from a successful IPO. First, and most obvious is the investment banks actually handling the transaction. As Ed Elson and Scott Galloway noted on a recent podcast, with the Trump court’s loosening of banking regulations, shady practices like a bank’s sell side research team setting price targets in order to benefit the bank’s other business lines, like for example taking a company public, seems to be back. A blockbuster IPO will generate a massive amount of fees for bankers, and they will want to do everything they can to make sure it goes through and is considered a success.
Then there are the other characters in the AI circular financing story. If Anthropic’s IPO is successful then the risky bets Nvidia has made in financing it, OpenAI, and the various neoclouds like CoreWeave will look like they paid off. It will also reward the hyperscalers who invested in Anthropic like Amazon, and will serve as an indication that the data centre buildout is necessary and a good investment; Anthropic made a profit in Q2 2026 so there will definitely be someone to pay for all this computing capacity, we promise! As Quoth The Raven pointed out, the accounting loss Anthropic made from its shares being revalued higher in Q2 2026 becomes an accounting gain of billions of dollars for Amazon and Alphabet. If I was to speculate, I’d expect most venture capital firms to sell a share of their stakes in Anthropic but I don’t think the other AI players will. After all, the more shares are on sale at a time drives prices down, which provides an indication that the AI trade is not going well. Indeed, I would not be surprised if Nvidia uses some of its abundant profits to buy Anthropic shares in a convoluted proxy buy back scheme.
But so far this discussion has been limited to sellers of shares, which begs the question who is going to buy these shares? Selling shares is always and everywhere a two party transaction which gives one party cash and the other equities. Will anyone want to buy shares in Anthropic? The honest answer is probably yes, though the price might be a sticking point. The valuation Anthropic initially wanted, $2 trillion, is looking unlikely, and it’s not hard to see why. The business has been a net loser of money, and looks set to haemorrhage money going forward.
Unlike Elon Musk, the people running Anthropic seem fairly unacquainted with the intricacies of high finance, and have thus not been smart enough or perhaps simply unable to game the system in the way Musk did with SpaceX’s IPO. SpaceX looks like a better business than Anthropic by grafting on a healthy business to the side of the sickly AI business, and crucially Musk used his clout and connections to make sure SpaceX would be fast-tracked into an index. This guarantees a stead supply of buyers seeking to mimic the index, which combined with a small number of shares being available to buy and sell on the open markets increases the price; or at least makes it less prone to crash.
Anthropic will lack that safety net, which is acutely relevant if my answer to the question of why now is correct. If the numbers will look worse after the IPO, if the first quarterly report as a public company contains a nasty “surprise” like a return to operational loss making, say, or an end to exponential revenue growth, perhaps both, then a violent market reaction is likely. If Anthropic has a nasty fall during earnings season, the risk is that all the listed AI firms get dragged down too, and suddenly we have a potential trigger for an AI crash.
Still, the fact that Anthropic seems likely to actually IPO makes me far less anxious about an AI bubble crash than I was last week. Unless the venture capitalists are exceptionally stupid and decide not to sell when they can dump their garbage on public market investors, the risk of contagion will be significantly lower when a lot of concentrated bets on Anthropic are cashed out. Assuming, of course, that the IPO actually happens.
I’m still not entirely convinced, because as mentioned finding buyers seems difficult. And, it is important to note that the type of buyer at an IPO is very different than the types of buyers at a venture capital funding round. Sam Bankman-Fried was memorably caught playing League of Legends at a Sequoia pitch meeting, and incredibly got the money. That type of behaviour won’t fly at a Goldman Sachs road show. And the machinery which has propelled AI to the valuation highs the sector currently enjoys is located almost entirely in Silicon Valley. Now that the S-1 is out the centre of gravity is shifting from Sand Hill Road to Wall Street, and portfolio managers are not making investment decision based on what they hear on All In or Dwarkesh.
For venture capitalists, even the late comers who invested just a few months ago, there was always a future pay off in the form of an IPO on the horizon. But what pay off can a teacher’s union pension fund expect from buying in to Anthropic now? Unless it is priced like the highly speculative investment it is, finding the kinds of mainstream buyers who stick to public markets could be difficult. Investing at 100x price/sales ratio is not on the table for most investors unless it’s a Musk joint. At the same time though, there is a valuation floor at or under which the operation of listing publicly is not an option for Anthropic; series H-4 investors are not looking to take a haircut on the IPO.
One of the important changes an IPO results in is the shift in the balance of power between insiders and outsiders. When you’re a venture backed start-up you can be secretive and disclose what you like, which won’t fly on the open market. Public market investors could easily demand, for example, the Q3 2026 numbers to actually be in the prospectus before they consider investing. Australian neocloud [firmus?] IPO was triaged recently as investors didn’t put up with the company’s bad financials and high target financials. The same fate could await Anthropic.
Consider the negative feedback loop some scrutiny could trigger:
Investors say they won’t consider investing in Anthropic releases its Q3 numbers
--> the numbers turn out to be bad and the company is still unprofitable
--> investors demand a lower price
--> either Anthropic accepts a lower valuation than the rumoured $1.4 trillion which triggers contagion in the AI sector as this means that OpenAI is also worth less than people thought
Or
Anthropic does not accept a lower valuation and the IPO is cancelled or more likely postponed which also triggers cold feet about the AI sector.
--> the new reality makes it far more difficult to raise money for private AI companies as the labs’ aura of invincibility has been pierced. As the kinds of start-ups which form the labs’ customer base, think Lovable, Replit, Harvey, can’t raise more money which forces them to spend less on OpenAI and Anthropic or go bust. The investors in those companies get burnt and stop funding them creating a negative feedback loop.
--> suddenly it becomes clear that the hyperscalers and neoclouds have massively overbuilt data centre capacity, now that AI companies are going bust.
--> worst case scenario this triggers a financial crash as leveraged bets on the AI sector – à la Leopold Aschenbrenner – go bad and margin calls lead to forced liquidations.
So, back to our third question: who benefits? Well, if things go to plan then a bad company with poor outlooks for the future is dumped on mainstream investors at a ridiculously inflated price, with a small group of already rich founders and venture capitalists cashing out massive gains. If it doesn’t, no one benefits.
If you liked this post you can read a previous post about the war in Iran here or the rest of my writings here. I also have a section for longer reads I call essays here, I particularly recommend my series called The Bird & The Technoking exploring Elon Musk's takeover of Twitter, and its political and cultural implications. It'd mean a lot to me if you recommended the blog to a friend or coworker. Come back next Monday for a new post!
I've always been interested in politics, economics, and the interplay between. The blog is a place for me to explore different ideas and concepts relating to economics or politics, be that national or international. The goal for the blog is to make you think; to provide new perspectives.
Written by Karl Johansson
Sources:
Cover photo by Mark Stebnicki from Pexels, edited by Karl Johansson
Kommentarer