Startups & Funding

Gartner Warns $3.5 Trillion in AI Investment Could Turn Worthless by 2030

Only 13% of surveyed IT directors report significant benefits from AI projects, and Gartner projects up to $3.5 trillion in AI infrastructure could lose its value by the end of 2030.

What Will Remain After AI? Where Value Survives
What Will Remain After AI? Where Value Survivesblickpixel / pixabay
By Marcus Bennett6 min read

Why it matters

  • Gartner estimates that by the end of 2030 more than $3.5 trillion poured into AI infrastructure could lose its value at current investment rates.
  • Around 40% of AI startups founded in 2024 have already shut down, according to estimates cited in the analysis.
  • Only 13% of more than 2,000 surveyed IT directors said their companies reported significant benefits from AI projects.
  • Leopold Aschenbrenner's Situational Awareness fund fell from $45 billion to roughly $10 billion within two months in the summer of 2026.
  • AI21 Labs cut more than 60% of its workforce this year despite prior investment from Google and Nvidia.

Gartner estimates that, at current investment rates, more than $3.5 trillion poured into AI infrastructure could lose its value by the end of 2030. That figure anchors a growing debate over whether the AI market has entered the same territory as the late-1990s dot-com boom — a debate that in October drew in two of the most influential names in global finance.

Billionaire Ray Dalio, founder of Bridgewater Associates, described the situation as a 'classic bubble,' citing companies that finance AI infrastructure through debt while interest rates rise. Ken Griffin, founder and CEO of hedge fund Citadel LLC, followed with public doubts that the massive investment in AI would ever pay off.

More money still flows into AI than the technology earns back. Some deals have become circular: one company invests in another, which then spends that money on the investor's own products. The revenue, in effect, loops back to its origin.

Why the timing question misses the point

An overinflated market behaves like a balloon, as the Unite.AI analysis puts it. It grows while money pumps in. The more it inflates without real returns, the thinner its shell becomes. Predicting the exact moment of rupture is, on this view, pointless. The productive question is different: where is the shell thinnest, what gives way first, and what opportunities open up afterward.

The stakes are considerable. If the correction arrives, it will redistribute capital, talent and physical infrastructure across the entire technology sector — and the companies that survive will be those whose value does not depend on borrowed foundations.

What is already collapsing?

Around 40 percent of AI startups founded in 2024 have already shut down, according to estimates cited in the analysis. Workflow automation tools such as Relay are closing for a specific reason: giants like Google and OpenAI are folding their features into their own products. The standalone services lose their reason to exist.

The pattern has precedent. When operating systems began opening PDFs and compressing files natively, hundreds of separate utilities became unnecessary overnight. AI accelerates this dynamic because it leaves even less room for add-on products built on third-party technology.

The survival test, according to the analysis, is structural rather than qualitative. It is not necessarily the tallest buildings that collapse in an earthquake, but those not built to withstand the shaking. For AI companies, the foundation consists of proprietary data and accumulated expertise. Companies relying on third-party models — models owned by the very platforms that can launch competing products — risk losing their competitive edge the moment that happens.

The practical evaluation question the analysis poses: not whether people need a product today, but whether they will still need it once a major platform builds the same function into its own offering.

Where do downturns create value?

If the AI bubble bursts, the market will take the hit. Not everyone will suffer. Crises destroy businesses, and they also create three specific categories of opportunity, the analysis argues.

Talent. A downturn returns experienced specialists to the market, people previously out of reach for most companies. The case in point is Israeli AI company AI21 Labs, one of the few independent developers of its own language models, which had previously received investment from Google and Nvidia. It still cut more than 60 percent of its workforce this year, including researchers and developers. Those specialists are now available to other employers.

New markets. The Kauffman Foundation estimates that more than half of the companies on the 2009 Fortune 500 list were founded during a recession or economic downturn. Airbnb launched in 2008. Uber followed in 2009. Neither waited for the economy to recover; both moved into markets that would have stayed occupied in better times. When incumbent players pull back and cut costs, they vacate space.

Cheap resources. When supply exceeds demand, prices fall. The fiber-optic market after 2000 supplies the template. Global Crossing built a network of roughly 100,000 miles of cable before going bankrupt in 2002. Its network remained in use. In AI, data centers and power capacity could similarly find different applications after a correction, priced far below construction cost.

What does the Aschenbrenner fund collapse show?

The cautionary tale about single-scenario investing comes from Leopold Aschenbrenner, a former OpenAI researcher who founded a hedge fund called Situational Awareness in 2024. The fund invested in AI infrastructure and simultaneously bet against software companies that might struggle to compete with AI — a hedge that looked diversified.

It was not. Both positions relied on the same underlying assumption: that the AI boom would keep accelerating.

Within two years, the fund grew to $45 billion. Then, in the summer of 2026, investors started doubting whether the enormous investments in AI would pay off quickly. It took just two months for Situational Awareness's assets to fall to around $10 billion — a decline of roughly 78 percent.

Aschenbrenner was not wrong about AI's potential to change the world, the analysis notes. The problem was that every one of the fund's investments depended on the same scenario playing out. Hedging against one risk while doubling exposure to another is not hedging at all.

Who actually survives?

The analysis draws a sharp line between what companies expect and what the data shows. Enterprise AI adoption is expanding and pilots are getting off the ground. The figures tell a different story:

  • On average, only one in five AI projects succeeds.
  • Of more than 2,000 IT directors surveyed, just 13 percent said their companies reported significant benefits from AI.

The greatest risk, the analysis argues, falls on companies that have not learned two disciplines: how to calculate the return on an AI investment before starting a project, and when to stop a project that delivers no results.

For companies asking whether they would survive a burst bubble, the analysis proposes four diagnostic questions:

  • Will customers keep paying for your product when an AI model provider launches the same functions tomorrow?
  • Is your product really built on your own knowledge, data and experience, or is it using someone else's model?
  • Does your company's growth come from its own revenue or from borrowed money?
  • What keeps the business moving forward — a couple of key people and a bit of luck, or a solid system that works regardless of who is involved?

What is the durable advantage?

The analysis closes with a judgment about what constitutes resilience. A few talented people on a team, or a run of luck, give a company only a temporary advantage — one that can disappear as quickly as it arrived. What makes a business durable is a system of established processes plus unique data and expertise built up over time.

When a crisis hits, the add-ons fall apart first. The systems endure, because those are the things no competitor can copy from someone else's model or borrow from someone else's foundation. In a market where 40 percent of a single founding cohort has already vanished and 87 percent of enterprises report no significant AI benefit, that distinction between add-on and system is likely to decide which companies remain standing on the other side of the cycle.

Source: unite.ai

Share this article:

More from Marcus Bennett

Marcus Bennett

Show full bio

Senior reporter covering companies, deals and strategy in artificial intelligence and machine learning at AI In Context.

718 articles

Related articles

  1. Forrester: AI bubble fears premature, but math is getting harder
  2. Michael Burry Says Nvidia's Real Problem Is a $176 Billion Accounting Gap
  3. UK Committee Chair Warns AI Boom Echoes Dot-Com Debt Spiral
  4. Santander Counts €194 Million From AI as Earnings Season Puts Numbers on the Hype
  5. Nasdaq 100's 5.5% Drop Puts $1 Trillion AI Bet on Trial

« Previous articleNext article »