On Monday, OpenAI announced $40 billion in new financing, the largest funding round in history, and one that nearly doubled the artificial intelligence company’s valuation to $300 billion. While no other startup can match those eyewatering numbers, they probably shouldn’t be a surprise given just how much capital is flooding into the technology: AI companies raised a record $110 billion in VC funding last year.
That kind of money might seem like a boon for AI innovation. But it may actually become a burden instead. First, by depriving these companies of invaluable market signals. Second, by driving them to appeal to investors instead of customers.
The leading AI companies are extraordinarily unprofitable, even by historical “growth at all costs” standards. OpenAI’s unprecedented ability to raise capital is critical for its continued functioning, as the company reportedly burned $5 billion in 2024. It’s not alone, as the Information reported that its competitor Anthropic burned $5.6 billion that same year. By comparison, the largest-ever loss by Amazon, another startup that initially prioritized growth, was $1.4 billion in 2000, and it became consistently profitable three years later.
These losses are not just about R&D or capital investment. Most of OpenAI’s users do not pay for the service, and each one costs the company money. An analysis by Ed Zitron, perhaps the industry’s most trenchant critic, found that OpenAI is likely losing money even on paying customers because of how expensive they are to serve (the best version of OpenAI’s o3 model can use more than $1,000 worth of computing power per query).
None of this means that generative AI isn’t useful. I used Perplexity and ChatGPT while researching this column. But it does raise questions about the technology’s extraordinary growth. How many innovations would have seen their uptake skyrocket if they were offered this far below marginal cost? Put another way: How fast would Amazon have grown if it had given away books?
Guessing what customers are willing to pay for isn’t easy. When Steve Jobs predicted that a new product would be as important as the PC and legendary venture capitalist John Doerr said it would be “bigger than the internet” they weren’t describing the iPhone. They were talking about the Segway.
The Segway was a technological marvel, but when it was introduced, prices started at $5,000. Its adoption curve might have looked a lot more like ChatGPT’s, though, if Dean Kamen, the device’s inventor, had been able to give them away. People might have given them a spin out of curiosity. Breathless news articles might have proclaimed the need to prepare for a post-walking world. Major automakers might have launched Segway divisions. Investors might have agreed with Kamen’s prediction that “the Segway HT will do for walking what the calculator did for pad and pencil” and handed over even more capital.
And all of it would have been a huge waste. Today, the least expensive Segway costs less than $500 — but it turns out that, even at that price, people prefer to walk. Often the best way to see what customers will buy at a profitable price is simply to try selling it to them.
Following investors’ whims, in other words, can be dangerous. In 1987, Harvard Business School Professors Bill Sahlman and Howard Stevenson described “capital market myopia,” a recurring pattern they first identified in a new invention called the Winchester disk drive. It plays out something like this: Projected growth in a hot industry attracts capital, driving up valuations and drawing more capital, which provides access to new technical and human resources and spurs even higher valuations. The end result is companies that chase investors (who usually want growth) instead of profitable customers and go down technological blind alleys even as they achieve valuations unjustified by any plausible estimate of future cash flows.
In the case of the disk drive, investors quickly piled into the new tech — until that overinvestment led to price wars and the collapse of the sector; in 1984, the market cap of the 12 leading publicly held disk drive manufacturers collapsed from $5.4 billion to $1.4 billion.
Generative AI seems to be following the same pattern. Deprived of paying customers sending real market signals, companies are chasing investor preferences and pursuing growth instead of profits — even when that leads to burn rates larger than the GDP of some African countries.
For example, last week ChatGPT’s Studio Ghibli-style pictures went so viral that OpenAI CEO Sam Altman said demand was “melting” its GPUs. Given the cost of answering even simple text-based queries, this viral moment likely wasn’t cheap.
Yet, even setting aside legal and ethical questions, how much demand is there for such cartoons? You could go on Fiverr and hire someone to make them for a few dollars. But there’s no sign many people are doing that. The technological feat likely impressed some investors, but what customers will pay OpenAI enough to make it worthwhile?
The pursuit of an “Artificial General Intelligence” that surpasses human cognition presents an even more expensive version of the same problem. Right now, companies are dumping billions of investor funds into scaling their approaches to reach this elusive goal. Yet a new report by 24 of the world’s leading AI researchers found that the overwhelming majority believe their efforts will fail. If it does, AI companies will have wasted resources they could have used to pursue more realistic goals — goals that might have produced profitable customers.
AI very likely will change the world, but as with many other revolutionary technologies, that transformation may take a different form — and arrive more slowly — than we expect. As an industry, AI companies would be far better served by less hype and more realism about what people will actually pay for. Until that happens, the capital markets would be wise to end the groupthink — and perhaps hold off for a bit on setting the next new funding record.
A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out some of our webcasts.
Bloomberg News provided this article. For more articles like this please visit
bloomberg.com.
Read more articles by Gautam Mukunda
