For years, tech bros, politicians, and Wall Street have been spruiking artificial intelligence as the holy grail of productivity growth.
Almost four years after ChatGPT brought AI into the mainstream, there is just one problem: the promised productivity boom is nowhere to be seen.
And with Anthropic intending to list on public markets in the next few months, the runway to prove AI is worth the hype is quickly running out of tarmac.
Yesterday, Anthropic’s prospectus was allegedly leaked to the media, apparently revealing the maker of Claude is seeking a valuation of more than $2 trillion, which would, if confirmed, make it one of the largest IPOs in history.
The alleged numbers are extraordinary.
The report suggests that revenue increased twelvefold last year to almost $4.6 billion, while losses are historically significant, with an operating loss of roughly $8 billion in 2025.
It also allegedly plans to spend over $500 billion in cloud, computing and AI infrastructure in coming years. Nearly a quarter of its revenue comes from just two customers.
Perhaps most remarkably, almost a third of the prospectus is devoted to risk factors, including the possibility its own technology could pose ‘catastrophic or existential risks to humanity’.
For comparison, SpaceX went public earlier this year after losing $4.9 billion on $18.7 billion of revenue.
Its shares initially soared 66 per cent before subsequently losing around half their value from their peak.
For valuations like Anthropic’s to make sense, AI eventually needs to create enormous value throughout the economy. So far, it’s proving difficult to find.
Since ChatGPT launched, total factor productivity growth has averaged 1.11 per cent a year, below its historical average of 1.23 per cent, according to the US Federal Reserve.
Over the four quarters to the end of 2025, it slowed to just 0.32 per cent.
Federal Reserve research also finds that the industries adopting AI fastest have not contributed disproportionately to recent productivity gains.
The Reserve Bank of Australia is also struggling to find significant evidence on the promised AI productivity boom.
RBA Governor Michele Bullock recently called AI ‘the great white hope’ for productivity, before conceding there were ‘very few signs yet’ it was improving the supply side of the economy.
The pressures on the demand side, however, are much easier to find. In other words: inflation.
At yesterday’s meeting, the RBA pointed to strong AI-related demand as one of the forces adding to global inflation.
Federal Reserve governor Lisa Cook has echoed those concerns, saying that AI is stoking short-term inflation.
In America, it is also cannibalising home building at a time of a housing crisis, as data centre construction sucks up all available resources, from tradies to copper wiring.
Nobody wants to dwell on another possibility: some of AI’s eventual productivity gains may come from fewer employees being expected to produce more.
A recent Australian survey found 56 per cent said time saved by AI was quickly filled with more tasks, while 47 per cent said expectations of them had increased since AI entered their workplace.
Workers in AI exposed occupations also face wage suppression and lower bargaining power with their employers, with many employees unable to switch jobs under threat of AI replacement.
So far, AI seems much better at generating investment, inflation and tech billionaires than the productivity boom we were promised.
Yet investors remain stubbornly in love with the story of AI.
A handful of AI stocks have driven US equities to record highs while at the same time much of the market underneath them has weakened, with ‘breadth’, the measure of the health of a stock market, falling to levels not seen since the dot-com bubble, according to Goldman Sachs.
There is another problem for Anthropic.
AI keeps getting cheaper, as competitors and open-source developers steadily drive down the price of using increasingly capable models.
That is great news for consumers.
It is less obviously good news for a company seeking a $2 trillion valuation.
It might also explain why frontier AI companies are so keen to warn governments about the dangers of models they do not control: regulation has a useful habit of raising the barriers to entry.
If bubbles tend to peak with their most audacious deal, a $2 trillion IPO for an AI company losing $8 billion a year would be a fitting candidate.
AI may yet produce the economic transformation its creators expect. But being right about a technology is not the same as being right about what its companies are worth.
That was the lesson of the dot-com bubble.
Anthropic will arrive on public markets with extraordinary growth, extraordinary losses, and valuation resting on profits that remain somewhere in the future.
The post AI’s show-me-the-money moment is approaching appeared first on The Spectator Australia.
