You Can Invest in AI. Will You?

I’ve just realized how old I am — I remember how, back in my childhood (and even a bit later), older people would talk about computers along the lines of “it’ll never catch on.”

When I was entering the job market, the internet was still a novelty — I even dug up a report according to which, in 1999–2000, only a little over half of medium-sized companies in Poland had internet accessThe report says that at the time 99% of medium-sized firms had a computer, and of those, 56% had internet access. Since the report doesn’t define exactly what a medium-sized firm is, I’d assume it means 50–249 employees, which is how Statistics Poland (GUS) defines it. .

You probably also remember the crypto craze from a few years back — plenty of lucky ones got rich on it, plenty lost serious money.

So, looking at today’s investment landscape, one might wonder: since AI looks at least as groundbreaking as the computer or the internet (I’d even venture to say it’s more groundbreaking than both of those changes put together), then can you make money on it?

Of course you can.

The problem is, most people are going to lose anyway.

Why do I think so?

Anthropic is going public, with OpenAI right behind it

The landscape of companies developing AI isn’t all that diverse, after all: we’ve got two leaders (OpenAI and Anthropic), followed by labs like Google’s DeepMind, Meta, xAI, and Chinese players like DeepSeek, Alibaba (Qwen), MoonshotAI (Kimi), and Z.ai (GLM).

Okay, maybe it is diverse after all, since the list could go on 😅

Still, there are two clear leaders, and both are planning to go public right now. On paper, that means you or I could become part-owners and profit from their earnings, or simply from the stock going up.

So what’s the problem?

As usual, the devil is in the details, so let’s take a closer look.

On October 26, Anthropic officially filed its S-1 registration statement with the US SECSecurities and Exchange Commission : it’s a document meant to give prospective investors information about the business model, finances, and risksThe data presented in this document ends on June 30, though, so it says nothing about the company’s most recent months — which could be damn important for judging how fast revenue or costs are growing. .

First, let’s see what the reported numbers look like on a chart:

• The company is aiming for a valuation of $1.5–2 trillion. Is that a lot? Well, it’s roughly the GDP of Canada!
• Tech CEOs (that is, the main customers of AI model vendors) have gone weirdly rabid over AI: they force their employees to use AI no matter what, and as a result the term “tokenmaxxing” popped up a while back: employees at big companies were literally being evaluated on the number of tokens they used!
  • The problem comes when incentives like that make costs snowball and companies start cutting back on usage; you can guess how that will play out for Anthropic’s revenue.
• Anthropic itself warns that the potentially catastrophic risk from further development and use of AI could be so large that the entire insurance industry won’t be able to cover it!
• The company employs 5,300 people, and yet it says plainly: in September 2026, over 90% of new code at the company was written by Claude. Each engineer today accounts for 45 times more code than in 2024. And headcount still grew 5.6x — and dozens of positions are still open. What are all those people going to do there?

What’s the industry saying?

There’s a platform I didn’t even know existed until recently: Blind is an anonymous forum for employees of big tech companies. You sign up with your work email, and every post shows the poster’s employer. And that’s exactly where polls started showing up asking those employees whether they’d invest in Anthropic/OpenAI.

See for yourself what the answers looked like a few months ago and recently, after the details from the S-1 came out:

But maybe they’re just jealous they don’t work at Anthropic?

Let’s zoom out to social media like X (formerly Twitter), Reddit, and Hacker News, plus the general take of experts in industry media:

The verdict is clear: 70% of voices are skeptical!

Okay, but think about it: since when has the crowd been any good at spotting investment opportunities? On Reddit there’s even a running joke, “inverse WallStreetBets”: when most voices say one thing, you should usually do exactly the opposite!

Here comes China

Remember how I mentioned the Chinese models in the AI vendor landscape?

Well, they stand out for one damn important feature: nearly all of them are open-weight, meaning you can download them and, given beefy enough hardware, run them locally.

That means you’re not locked into US vendors, which for us Europeans matters all the more, because ever since Trump became president again we’ve seen clearly that we can’t count on a stable relationship with the US the way we used to.

Anthropic itself admits in the prospectus that open-weight models may put pressure on prices. It also writes that competitors have tried to copy its models by training on Claude’s responses.

Ask yourself: if Chinese labs, hobbled by export controls on chips, could train models nearly as strong as Claude on Claude’s own outputs, how long can Anthropic hold its lead and its ever-growing revenue?

Dependencies and financial risks

There’s one more risk worth knowing about: it seems everyone is making money in such a way that you (as a retail investor) earn the least.

You probably realize that Anthropic, being a relatively young company, hasn’t built up the massive infrastructure it needs to deliver its services, and still has to keep buying chips to handle the growing traffic, right?

Exactly. So it has to sign contracts to lock in that infrastructure and hardware and guarantee uninterrupted operation, and that’s not chump change! Worse still, Google, which has a huge amount of infrastructure for rent, is also developing its own models!

The upshot is that Anthropic signs purchase commitments because it has no choice, and for every dollar Google has invested in Anthropic, it gets almost $5 back in commitments. Amazon does even better: over $8 for every $1. How much Microsoft put in, nobody knows (hence the fogged-up pressure gauge in the graphic below).

As long as the company keeps growing, it’s all good — but once growth stalls, the installments will still have to be paid, and there’ll be a problem.

What about OpenAI?

For Anthropic’s main rival, OpenAI, we don’t have juicy details like the ones from the published S-1. But there’s one interesting thread that turns on another warning light on the dashboard: The Wall Street Journal reports that the company aimed to raise $30 billion in its latest round and managed to raise… less than half!

I don’t have a clue what these fundraising talks look like or what specifics Sam Altman shows in them. But the mere fact that he raised less than half of the target says something.

The pros, and last call

According to the Financial TimesIf you don’t have access to ft.com, use https://wallhop.io/https://www.ft.com/content/59383c7f-8d82-493f-bf01-d2e862a1018f?syn-25a6b1a6=1 , Blackstone warned investors that after its stock market debut Anthropic could get a rating below investment grade, probably in the BB range, which in plain English means a junk rating.

On the other hand, Amazon had a junk rating for over a decade, so a rating like that doesn’t write a company off by itself.

⊕

You know, I have a feeling that a public offering of a company like Anthropic is something like a party at a trendy club. A party that started with a quiet, well-guarded VIP entrance opening at 9 p.m.

Those who got in then drank the free champagne, hit the best of the buffet, had a grand old time, and happily moved on to the next club. Then, around 3 a.m., the doors graciously open to the general public, with a steep cover charge.

You walk in hoping for the night of your life, and you get the bill for the broken glass and a broom.

Or maybe it’s time to run?

A few days ago, a lot of tech stocks took a hit.

In my view, that drop simply came from a market looking for an excuse to correct and reacting with disappointment to the earnings of Google, Intel, and Amazon.

I keep wondering: what happens when a share offering as big and significant as Anthropic’s or OpenAI’s faceplants spectacularly?

As you’ve probably guessed, I’m not putting my money into Anthropic. In fact, over the past few months I’ve pulled out of some stocks to reduce my exposure to what looks like an overheated market. I don’t know when the correction will come. But I can feel it in my bones: when the shit hits the fan, it’s going to be ugly.

I hope I’m wrong.