The Greatest Show on Earth: AI’s Bubble, Bonanza and Betrayal of the Little Guy
From trillion-dollar valuations to eye-watering subscription fees that lock out ordinary users, the AI industry’s three-ring circus is in full swing.
There is a particular kind of madness that grips financial markets every generation or so. It smells of opportunity, sounds like a revolution, and ends, reliably, with a lot of very clever people looking very stupid indeed. Ladies and gentlemen, welcome to the artificial intelligence gold rush of 2026 — the most spectacular, most overhyped, and most ruthlessly expensive circus since the dot-com boom handed out clown suits to half of Silicon Valley.
The numbers, as they say, don’t lie, although the people presenting them might do. AI investment has ballooned to truly vertiginous heights, with global venture funding pouring into the sector at a pace that would make even the most optimistic tulip merchant blush. OpenAI, the poster child of this particular mania, is now valued at roughly $300 billion, a figure so vast it makes you want to lie down in a darkened room. Anthropic, its slightly more buttoned-up rival, has hoovered up tens of billions in funding. Meanwhile, investors queue around the block, cheque books open, eyes glazed, muttering something about transformative technology and paradigm shifts whilst their financial advisers weep quietly into their spreadsheets.
The question — the rather large, inconvenient, elephant-in-the-data-centre question — is whether any of this is actually worth it.
A Bubble By Any Other Name
Let us be precise about what a bubble is. A bubble is what happens when the price of something becomes entirely detached from its underlying value, sustained only by the collective belief that someone else will pay even more for it tomorrow. By that definition, significant portions of the AI market are not approaching bubble territory — they have pitched a tent there, put up curtains, and started receiving mail.
The uncomfortable truth is that most AI companies are burning through cash at a rate that would make a Premier League football club wince. Training large language models costs hundreds of millions of dollars. Running them costs more. The infrastructure requirements — the chips, the data centres, the eye-wateringly expensive Nvidia GPUs that power the whole enterprise — are so colossal that even the most bullish analyst occasionally pauses mid-sentence, stares into the middle distance, and quietly reconsiders their bonus projections.
Revenue, meanwhile, remains stubbornly modest relative to those valuations. OpenAI reportedly brought in around $3.4 billion in revenue last year, which sounds impressive right up until you divide it by a $300 billion valuation and realise you’re looking at a price-to-sales ratio that would make a 1999 dot-com founder feel conservative. The gap between what these companies are worth on paper and what they actually earn in practice is, to put it in the technical financial terminology, absolutely Madoff.
Everyone’s Rushing for the Exit — Via the Front Door
Which brings us, naturally, to the IPO frenzy. When venture capitalists start talking loudly about public listings, it is generally a sign of one of two things: either a company is so robustly profitable that it deserves the scrutiny of public markets, or the smart money has decided it is time to pass the parcel before the music stops. In the current AI climate, one suspects it is rather more of the latter.
Several high-profile AI companies are circling the public markets with the cautious enthusiasm of someone approaching a suspicious-looking buffet (and we don’t mean Warren). Anthropic has been mentioned in the same breath as a potential listing. CoreWeave, the AI infrastructure firm, already made its debut on the Nasdaq earlier this year in what was, depending on your perspective, either a triumphant validation of the AI thesis or a spectacularly well-timed exit by its early backers. The stock promptly wobbled. Observers were, if not shocked, at least mildly unsurprised.
The IPO playbook here is achingly familiar. Build something impressive-sounding. Attract eye-watering private valuations. Generate breathless press coverage. List publicly before the quarterly results get too repetitive in their disappointment. Repeat. The retail investor, as ever, is invited to the party, just in time to help the early guests find the door.
AI for All (Terms and Conditions Apply)
And here, dear reader, is where the story curdles from merely absurd to genuinely objectionable. Because whilst all this investment theatre plays out in the upper atmosphere of finance, something rather telling has been happening at ground level: AI is becoming increasingly expensive for ordinary people.
Cast your mind back, if you will, to the original promise from just a few years ago. AI, we were told, would democratise knowledge. It would give the sole trader in Swindon the same analytical firepower as a Goldman Sachs research desk. It would level the playing field, disrupt the incumbents, and empower the individual. It was, in the breathless, oft-repeated parlance of the tech world, going to change everything.
What has actually happened is rather different. OpenAI’s most capable models now sit behind a $200-per-month ChatGPT Pro paywall. Anthropic’s Claude at the top tier costs similarly eye-watering sums for meaningful professional use. Google, Microsoft, and the rest have quietly restructured their offerings so that the genuinely useful, generally powerful versions of their AI tools are priced squarely at enterprise budgets — the kind possessed by corporations, institutions, and well-funded startups, rather than freelancers, students, or, heaven forbid, curious members of the general public.
The free tiers, meanwhile, have been trimmed, throttled, and rate-limited to the point where they function primarily as extended advertisements for the paid versions. You can have a taste of the future — just not enough to actually do anything gob-smackingly useful with it.
The Democratisation That Wasn’t
This is not an accident. It is, in fact, entirely logical given the economics outlined above. If you are spending hundreds of millions running these models, you cannot afford to give them away. And if your investors are expecting the kind of returns that justify a $300 billion valuation, you need revenue — serious, corporate, enterprise-grade revenue — not a million freelancers paying £18 a month.
So the great democratisation of intelligence has quietly become the great stratification of intelligence. The gap between what an institution can access and what an individual can afford has never been wider. AI, it turns out, is following the oldest rule in capitalism: technology trickles down, but the best stuff costs a fortune, and the people who most need a competitive advantage are precisely the ones who can least afford to pay for it.
The circus is spectacular, the performers are talented, and the ringmasters are extremely well compensated. Just don’t be surprised, somewhere down the line, when the tent comes down and the field is left rather muddy. It always is.
Keep up to date with The Letts Journal’s latest news stories and posts at our website and on twitter.