About two years ago, Sam Altman and Garry Tan sat down to talk on a podcast about Altman’s recently published techno-optimist essay “The Intelligence Age.”
In their chat, the former and current presidents of Y Combinator postulated in hushed and reverential tones that “super intelligence” could be a mere “thousands of days” away.
Some of the benefits Altman and Tan said humanity could look forward to include, in their words:
- “fixing the climate”
- “abundant energy”
- “nearly-limitless intelligence”
- “establishing a space colony”
- “the discovery of all of physics”
Checking back about 740 days after these exciting announcements, their predictions are not looking very prescient — at least not yet. We’ll check back on that abundant energy next year, and the year after that.
One area where abundant growth has been found, however, is in the valuation of key AI companies. Indeed, Altman’s OpenAI was valued at $157 billion when he appeared on Tan’s podcast in 2024; it’s now grown about 13X to $1.25 trillion.
Meanwhile, Anthropic was a valuation guppy in 2024 worth a pitiable $18.5 billion. The company is set to IPO tomorrow at a valuation of about $1.4 trillion – a healthy 75X growth in two years.
These valuations are wildly optimistic and predicated on the companies generating several times their current annual revenues and, one would hope, achieving significant profitability. For now, the values are mainly driven by the FOMO-based largesse of investors.
But despite turbo-charged revenue growth over the past two years, OpenAI and Anthropic have not yet figured out how to stop losing money hand over fist.
OpenAI grew its revenue 50% quarter-on-quarter to $10 billion in Q3 of this year, but still saw operating losses of nearly $11 billion in the period. In its S-1 IPO disclosure, Anthropic announced $17 billion in revenue in Q3, a 47% increase from Q2, with a much smaller operating loss of about $567 million, but a net loss of $23.9 billion.
Anthropic also reported future spending commitments (mainly on AI infrastructure) of nearly $900 billion. Comparing the net losses and cost commitments of both companies is tricky because of their overly complex financials and different methods of accounting — and the fact that one is private and one is about to be public — but no one disputes that they incinerate money at a rate never seen before.
In 2026, AI industry leaders continue to tell us that the age of abundance is nigh and we can expect unlimited intelligence, wealth, and energy. What they’ve forgotten is that everything, no matter what, has a cost, whether it’s clear and present or a background externality that rears its head at the most inopportune time.
In other words, they’ve forgotten that there’s no free lunch.
In the case of the AI industry, despite LLMs working like magic for free or very cheap for consumers, the hidden but very real costs being subsidized by investors include chips, servers, data center construction, absurd amounts of electricity, absurd amounts of water, an army of highly-paid workers in first world countries, and an even bigger but less well-paid one in places like the Philippines and Kenya doing digital grunt work so that consumers can send snappier emails.
These very real (and growing) costs amount to trillions, and there is no clear roadmap for how a return on investment will be achieved. Researchers at Gartner recently estimated in a much-discussed analysis that $7 trillion will be spent on AI infrastructure from 2026 to 2030 which would need to generate $8.5 trillion in revenue to generate a reasonable ROI. The most revenue that the good researchers at Gartner could forecast during the same period, however, was a mere $3.6 trillion. You don’t need Altman and Tan’s “nearly-limitless intelligence” to know that’s a $5 trillion shortfall.
That missing $5 trillion is the free lunch.
That free lunch is what AI luminaries have convinced themselves is inevitable when we discover all of physics, and more, thanks to all that unlimited energy from super intelligence. Production gains will be unlocked, unforeseen and incredible new industries will be developed, cancer will be cured, climate change reversed, and these AI companies will thrive.
But what if they don’t?
If you remember nothing else of this column, memorize that $5 trillion number. That’s the price of the free lunch baked into the AI boom based upon Tan and Altman’s timeline from 2024.
I’ll check back in a year or two to see if the bill ever came.










