Consider a recent news report: “Anthropic is in advanced talks to raise $2 billion in a deal that would value it at $60 billion, more than triple its valuation from a year ago” says the Wall Street Journal. Anthropic’s last venture capital (VC) funding round, less than two months ago, raised $2 billion, bringing its total funding to $9.6 billion.

OpenAI had raised $6.6 billion, last October just a year after it raised $10 billion from Microsoft the previous year and just months after it also raised funds on the debt and secondary markets. So that $6.6 billion round was the largest VC funding round ever, a little larger than the $6 billion that Elon Musk’s generative AI startup, xAI, raised in 2023. It nearly doubled OpenAI’s valuation to $157 billion, and brought its total VC funding to $21.9 billion.

Yet big losses continue

Meanwhile, losses for both startups are continuing to rise, which is one reason they keep raising money from VCs. Sam Altman has recently tweeted that OpenAI’s cumulative losses had reached about $8 billion at the end of 2024, sparking rumors of a bankruptcy last July, and its annual losses could reach $14 billion by 2026. Meanwhile its revenues were only $3.7 billion in 2024.

Anthropic’s situation appears worse, albeit on a smaller scale. It expected to spend over $2.7 billion in 2024, with revenue only a fifth to a tenth ($0.4 to $0.8 billion) of OpenAI’s while xAI has yet to make announcements about revenues, partly because it was only founded in 2023 and thus probably doesn’t have any. Meanwhile, self-driving startups such as Waymo, Nuro, and Cruise continue to need new funds because they generate very little revenue, and thus big losses.

Do losses matter?

Optimists will say, “Who cares about losses? AI is a once-in-a-century opportunity!”

Unfortunately, we have heard this story before. Just a few years ago, valuations for many loss-making startups boomed, the number of tech IPOs (initial public offering of shares) peaked (at 1,035 in 2021), and their stock market prices exploded before all three plummeted in 2022 and have continued to decline. During the peak, articles oozed with complimentary language about these startups, their genius founders, and the cities in which they were headquartered. Those cities that were missing out on the boom were heavily criticized.

As documented in my 2024 book, Startups, Unicorns, and Bubbles, many of those unicorn startups, particularly those valued at more than $1 billion before doing an IPO, had huge losses and subsequently went bankrupt. About 85% of publicly traded American unicorns (about 120) lost money in 2023. That doesn’t count the dozens that went bankrupt or were acquired for low prices in the few years before the book was published.

Furthermore, 26 such startups suffered cumulative losses greater than $3 billion, the amount that Amazon sustained the year it became profitable. Five have more than $10 billion in cumulative losses of the sort that caused WeWork to go bankrupt. Startups such as OpenAI and Anthropic may end up with the same big cumulative losses, albeit bankruptcy is a less likely outcome.

Will those losses lead to bankruptcies?

It is mostly the startups with smaller losses that have been gone bankrupt or been acquired for low prices in the last few years. This trend has continued since my book was published. In the last six months, for example, Fisker went bankrupt and 23andMe is nearing bankruptcy, as are many eVTOL, electric vehicle, and battery startups. Neither Better Therapeutics nor LumiraDX, biotech startups, have released income statements in the last year. Furthermore, activist investors are suspicious of the figures produced by Carvana and Hims & Hers Health, also once unicorns.

A more comprehensive assessment of digital health startups predicts “massive public bankruptcies” in the next few years. The report says VCs “dumped lemons into public markets, exploiting fear during the pandemic.” For AI transcription, the report refers to another study of “92(!) AI medical scribe apps, and 89 of them are blatant ripoffs.”

The statistics for startup failures is equally frightening. “Startup failures in the U.S. have surged by 60% in the past year: 254 venture-backed clients went bankrupt in the first quarter of 2024.”

But is the past the future?

Back to Anthropic and OpenAI. The problems with previous startups don’t doom today’s AI startups. They merely teach us to be skeptical. After all, a few profitable startups, some with high market capitalizations even before the current AI bubble, have emerged, including Uber, Palantir, and Moderna.

What we shouldn’t do is think that high valuations mean that a given startup is successful. VCs and founders have been playing this pump and dump game for 10 years, with the peak reached in 2021. They claim a startup is great because a few insiders gave it a high valuation when the high valuation is merely used to justify the large funding round, which is needed to cover the startup losses. If a startup is truly doing well, it doesn’t need more funding or a high valuation; it merely does an IPO.

Pump and Dump Strategy

The VCs and founders need to keep funding the startups and giving them high valuations until the public is ready for IPOs and then the dump phase of the pump and dump will be implemented. Are public investors ready yet? The $10 trillion that was added to the market capitalization of the magnificent seven (Apple, Microsoft, Google parent Alphabet, Amazon.com, Nvidia, Meta Platforms and Tesla) since January 2023 suggests to some that it might be here.

The number of IPOs has not significantly increased because many investors seem to believe the magnificent seven already has the AI market locked up. Thus there is only a small chance that today’s AI startups might succeed and thus be warmly welcomed by public markets, at least enough to justify the opening of their books. Going public requires startups to release details about revenues and costs and many startups would like to hide those details as long as they can. Even SpaceX and Stripe remain private, likely for that reason.

VCs and founders are getting nervous, however. Generative AI is very expensive; it requires huge training costs and to be honest, the revenues aren’t rising that fast. Many startups have far smaller revenues than does OpenAI and some of those startups are for the most part gone. Inflection AI’s employees and software were acquired for $650 million by Microsoft last year to satisfy impatient investors. AI unicorn Character.AI was acquired by Google and Adept was acquired by Amazon.

The magnificent seven are also getting nervous because their AI revenues are not rising very fast, something Gary Smith and I have written about many times. In one article, we demonstrate that the dotcom bubble of 2000, which also required huge investments in infrastructure, was generating about $1.5 trillion in revenues the year it popped, one hundred times the revenues that are now realized for generative AI.

Amidst the hype and irrational exuberance, VCs have no option other than to keep funding startups and announcing huge valuations for them. It is their only chance of survival. But can we say that those startups are successful? The future will tell.