AI Valuations Soar as Skeptics Call Bubble
AI startup valuations have raced into what some industry veterans now bluntly call bubble territory. Sridhar Vembu, founder of business software giant Zoho and currently its Chief Scientist, has framed today’s AI funding boom as “clearly an investment bubble.” His view is nuanced: every major tech wave, from railways to the early internet, produced financial excesses without invalidating the underlying technology. What worries him is not whether AI is transformative, but whether investors can navigate the financial cycle “without losing one’s shirt.” Capital is pouring into a small set of model builders and infrastructure providers, often at eye‑watering valuations that run far ahead of proven, durable profits. As money chases the next big platform, the debate is shifting from whether AI is important, to whether current AI startup valuations reflect real demand or a speculative frenzy.
The Cloud Credit Loop Inflating AI Startup Valuations
Beneath headline AI valuations lies a structural concern: much of the apparent revenue may be little more than a circular flow of cloud credits. Large tech platforms invest in AI startups, but instead of cash, a big slice arrives as credits that can only be spent on the investor’s own cloud. The startup burns those credits to train models; the cloud provider then books that activity as fresh customer revenue. Microsoft’s multibillion‑dollar commitment to OpenAI reportedly followed this pattern, with Azure credits fuelling a cloud bill that has reached well beyond OpenAI’s actual revenue. Anthropic operates under a similar structure with Amazon Web Services, while Google also owns a significant stake. The result is blurred lines between investor, supplier, and customer. AI revenue that looks like broad market traction can, in reality, be the same money cycling between a handful of deeply intertwined firms.
Paper Profits and Concentrated Risk at the Top
The feedback loop does not stop at cloud revenue. Tech giants are also reporting sizeable paper profits as their AI holdings are marked up on balance sheets. When an AI startup raises a new round at a higher valuation, prior investors revalue their stakes and recognise those unrealised gains as profit. Alphabet recently reported record quarterly profit, with a substantial portion derived from an accounting gain tied to its Anthropic investment rather than operational performance. Amazon similarly logged a multibillion‑dollar profit boost from Anthropic even as its free cash flow plunged and it spent heavily on data centres. Meanwhile, Anthropic’s valuation has leapt to towering levels amid questions about how its revenue is booked, including the impact of gross reseller accounting. These dynamics concentrate financial risk in a few AI bets whose true, cash‑generating power is hard for outsiders to gauge.
Echoes of the Dot‑Com Era and the Risk of AI Market Correction
Critics hear echoes of the dot‑com era, when telecom firms inflated results by swapping network capacity and recording each swap as revenue. Those deals were ultimately deemed fictitious and, in some cases, illegal. Today’s AI cloud‑credit loop is different in one crucial respect: it appears to comply with current accounting rules. That legality may delay any reckoning, but it does not remove the economic fragility. Inflated revenues support rich AI startup valuations, which then feed major stock indices, funneling more passive capital into the same names. If sentiment turns or growth stalls, an AI market correction could ripple through cloud providers, pension funds, and late‑stage investors. History suggests that transformative technologies can survive brutal valuation resets. The open question is who will be left holding stakes priced for perfection if today’s tech investment bubble deflates.
What a Bubble Means for Startups, Investors, and the Tech Ecosystem
For startups, the current AI valuation bubble is a double‑edged sword. Easy capital and generous cloud credits let young companies train frontier models and scale quickly, but they also encourage business models that depend on subsidised infrastructure and optimistic revenue definitions. If investors demand hard profitability or if credit‑fuelled spending slows, many AI firms could discover their growth was built on unstable ground. For investors, the challenge is distinguishing genuine product‑market fit from numbers inflated by circular spending. Overexposure to a handful of richly valued AI names may amplify portfolio volatility. Meanwhile, the broader tech ecosystem risks distortion as talent and budget chase fashionable AI initiatives at the expense of less hyped, but more durable, innovations. As Vembu argues, the task now is not to dismiss AI, but to participate in its promise without mistaking accounting mirages for sustainable value.