Will AI turn out to be a bubble?

Op-Ed written by Lars Gumede for Business Day on whether the AI boom is in fact a bubble. Available at:

https://www.businessday.co.za/opinion/2026-03-11-lars-gumede-will-ai-turn-out-to-be-a-bubble/

FULL TEXT

By Lars Gumede

With the AI boom in full swing with record valuations and hundred billion dollar deals being regularly signed, many are starting to question whether the entire AI boom is in fact a bubble.

What the experts are saying

The AI mania is causing industry leaders to sound the alarm. Goldman Sachs CEO, David Solomon, warned that a lot of the capital flowing into AI will not deliver returns, Amazon founder Jeff Bezos has stated that although AI will bring great benefits, we are likely in an industrial bubble, and OpenAI’s CEO, Sam Altman, has publicly stated that we are in a bubble and that “someone is going to lose a phenomenal amount of money”.

Google’s CEO Sundar Pichai has stated that the AI investment boom has an element of irrationality in his view. BlackRock boss Larry Fink has said that he does not believe there is a bubble since most of the money is being spent on “cloud and the power of the cloud” which will likely be well-spent. However, he added that along the way, “we’re going to have some big winners and we’re going to have some big losers”.

The Bank of England recently warned that the entire AI boom could unravel due to the debt-fuelled nature of spending on AI infrastructure. On top of that, a report by the Nobel prize winning MIT economist Daron Acemoglu, showed that in 95% of cases corporate AI pilot programs have shown no measurable positive impact. So what are the actual concerns?

State of AI valuations

The top AI-related companies are trading at enormous valuations — far above ordinary technology valuations, with record revenue multiples.

Driven by the demand for AI-capable chips, Nvidia reached a valuation of $5 trillion briefly in October 2025 — larger than the GDP of every country except the US and China. OpenAI has an estimated annual revenue of $10 billion (although they claim it is higher). They are now in talks to IPO at a $1 trillion valuation — representing a valuation of 100 times revenue (100x revenue multiple). Palantir, the AI data integration and analysis company, in 2025, briefly traded at a $600 billion market capitalisation — a whopping 600x price to earnings ratio. Anthropic, as of Jan 2026, has raised over $20 billion; having a new funding round every few months, doubling its valuation every time to a current valuation of $350 billion off $9 billion revenue. Perplexity, an AI search engine, as of September 2025, was raising at a $20 billion valuation despite annualised revenues of just over $120 million; 120x to 180x multiple on revenue. These revenue multiples are far above the ordinary listed tech company median of approximately 7—10x.

But it is not just the top companies, AI startups are raising at record levels too. We are in the era of AI startups. In 2025, AI startups raised a record $150 billion globally, representing roughly half of all global venture funding. In the US alone over 50 AI startups raised over $100 million in 2025. AI startups around the world are raising incredible amounts of money and at incredible valuations (many with no revenue at all). In fact there are startups raising hundreds of millions of dollars with no revenue, no product, just an idea and a promise.

Safe Superintelligence Inc was founded by Ilya Sutskever (former senior scientist at OpenAI). Ilya raised $2 billion in funding and Meta immediately offered to buy SSI for $32bn. SSI has no product, no plan, no monetisation strategy, just one vague goal; ‘build a safe super-intelligence before someone else builds a dangerous one’. Thinking Machines Lab was founded by Mira Murati (former CTO at OpenAI) after a fallout at OpenAI. Murati secured $2 billion in funding at a $12bn valuation with no product, no revenue and no customers. Flapping Airplanes is the latest AI startup founded in January 2026 with no product and no revenue that raised $180m at a $1.5bn valuation with the goal to “rethink model training”.

Circular deals by top AI companies

Another issue being raised is that of circular deals being done by the top AI giants. For example, Nvidia agreed to invest up to $100 billion in OpenAI to help fund a data center, OpenAI then committed to outfitting those datacenters with millions of Nvidia chips.

OpenAI signed deal with Advanced Micro Devices (AMD) to get tens of billions of dollars’ worth of its chips. As part of the arrangement, OpenAI is to become one of AMD’s largest shareholders.

OpenAI also struck a separate $300 billion deal with Oracle to build datacenters across the US. Oracle then buys billions worth of Nvidia chips for those facilities, sending the money back to Nvidia, which then continues to invest in OpenAI. On top of that, OpenAI does not have $300 billion for the chips, and Oracle does not have $300 billion worth of chips to sell.

This type of dealmaking has lead some analysts to say that while billions of dollars of deals are being signed, in reality the same money is moving back and forth between the top AI companies.

Paulo Carvao, a senior fellow at the Harvard Kennedy School who researches AI policy and who worked in tech in the late 1990s, has said that these types of circular deals were a large factor during the dot-com crash as well, and that while these companies do have tangible products and customers, their spending far outpaces their monetisation.

There are two primary fears in relation to this type of circular deal making. First, these deals give the spectre of high growth and will artificially inflate companies’ revenue numbers, which inflates the true value of these companies. Second, these deals tie the fortunes of all of these companies together increasing systematic risk and making any potential crash much worse.

When do we make money?

The hype around AI came because this technology was so novel and exciting that investors thought they needed to get in at any price. However, as time goes on investors are increasingly starting to ask when these top AI companies are actually going to make money.

A recent HSBC report showed that OpenAI likely will not make a profit until 2030, and that it needs to raise at least $200 billion plus in order just to not collapse. Elon Musk’s XAI is burning $1 billion a month on compute, allowing free usage to consumers. The issue is many AI companies do not have profitable business models, although some like Anthropic do. Anthropic is expected to overtake OpenAI on revenue this year and hopes to become cashflow positive in 2028.

Now many argue that it does not even matter if these companies have a profitable business model as the revenue growth and hype alone is enough to deliver investor returns. But this becomes a real risk given the trillions being spent on AI infrastructure which is heavily debt-fuelled. If the bet does not pay off, it could leave these companies in ruin, leading to OpenAI’s overtures to the Trump Administration for loan guarantees. This, in turn, lead the White House AI-czar David Sacks to declare that “there will be no AI bailouts.”

Problem is misallocation of funds

Now AI does have the potential to completely reshape the future for the better and the results of the AI craze will likely be positive. In fact, there is still massive upside for AI investments given its potential. The issue is not whether there is a bubble or not, but one of misallocation of funds within the AI field. Investors are chasing returns and top-blasting a few large companies when those investments should be going into smaller, promising ventures with profitable business models and that can show a tangible benefit to investors and society. Investors top-blasting a few large names will likely loose massively in any future correction but investors betting on smaller ventures building real, profitable and customer-facing businesses will likely win big in the long term.