Nobody needs $297 billion

Q1 2026 global venture funding hit $297 billion. AI startups absorbed $242 billion of that, roughly 81% of all capital deployed. Four deals alone, OpenAI at $122 billion, Anthropic at $30 billion, xAI at $20 billion, and Waymo at $16 billion, totaled $188 billion. That $188 billion across four companies is more than all global venture funding in some recent full years. It's more than ten times the peak quarterly venture investment during the dot-com boom. And it raises a question that the industry seems allergic to asking: is this rational?

The numbers in context

To appreciate how unusual Q1 2026 is, you need a baseline. In 2024, total global venture funding was approximately $314 billion to $368 billion for the entire year, depending on whose data you trust. AI captured about 37% of that, already a record share. By Q4 2024, AI crossed the 50% threshold for the first time. Then 2025 happened. AI startups pulled in roughly $259 billion globally, representing 61% of all venture capital. The concentration was accelerating, but at least it was spread across thousands of companies. Q1 2026 is different. Not just because AI's share jumped to 81%, but because of where the money went. Crunchbase data shows that the vast majority of venture dollars went to a select few companies, almost all based in the US. The top four rounds alone accounted for 63% of the quarter's entire global total. We're not funding an industry. We're funding a handful of bets.

The dot-com parallel everyone keeps making

The comparison to the dot-com bubble is unavoidable, but the details matter more than the headline. During the dot-com peak in 2000, quarterly venture investment hit approximately $28 billion. Q1 2026's $297 billion is more than ten times that figure. But the 2000 boom involved thousands of smaller rounds across many companies. The capital was distributed. The optimism was widespread. Everyone was getting funded. 2026 looks nothing like that. The capital is extraordinarily concentrated. A small number of companies are absorbing most of the money, and most of them are building foundation models, the same product category with the same basic bet on scale. During the dot-com era, the technology was real. The internet did change everything. But the valuations were disconnected from fundamentals. Companies with no revenue and no clear path to revenue commanded billions in market cap. The crash wasn't a repudiation of the internet. It was a repricing of irrational expectations. Today's AI valuations tell a similar story. OpenAI closed its round at an $852 billion valuation. It's generating about $2 billion per month, roughly $24 billion annualized, against what multiple reports describe as enormous and growing costs. The technology sector is trading at approximately 35 to 40 times earnings, stretched but not yet at the dot-com's 80x extremes. The question is whether we're on the way there.

The Jevons paradox angle

There's a popular argument that cheaper AI will lead to more AI spending, not less. This is Jevons paradox, the 19th-century observation that making coal engines more efficient didn't reduce coal consumption but increased it. Lower cost per unit of work meant more work got done. Applied to AI, the logic goes: as inference costs drop and models get cheaper to run, new use cases become viable. Demand expands. Total spending grows. It's a compelling framework, and there's real evidence for it. Every efficiency gain in AI chips or model architecture lowers the cost per unit of intelligence, making it viable for industries that couldn't previously afford to integrate AI. But Jevons paradox describes consumption of a resource, not capital allocation into companies producing that resource. More people using coal didn't mean more investors should have piled into coal mining companies at inflated valuations. The paradox explains demand growth. It doesn't validate every investment thesis attached to that demand. The fact that AI usage will grow enormously doesn't mean that four companies need $188 billion in a single quarter. Usage growth and capital efficiency are different questions.

What happens to everyone else

When 81% of all venture capital goes to one sector, the rest of the startup ecosystem operates on scraps. Non-AI startup investment fell to $41.8 billion in Q2 2025, the lowest in over seven years. Founders building outside AI face a stark reality: they're competing for a shrinking share of an increasingly distorted pool. This isn't just a fundraising inconvenience. It's a structural problem. Venture capital is supposed to fund a diverse range of bets across industries and technologies. When the overwhelming majority of capital chases a single thesis, the ecosystem loses its ability to discover and fund the next category of innovation. Climate tech, biotech, fintech, developer tools, there are real companies solving real problems in all of these spaces. But when every pitch deck needs "AI" in the title to get a meeting, the market's price signal breaks down. Capital stops flowing to the best ideas and starts flowing to the best narratives. Some observers argue this creates opportunity for contrarian investors. If everyone is zigging toward AI, the best deals might be among the companies zagging away from it. Maybe. But that's cold comfort for founders who need capital now.

The return problem

Here's the number that should make everyone uncomfortable: according to an MIT study, 95% of generative AI initiatives are generating zero quantifiable return. NPR has reported that only about 3% of customers currently pay for AI services. Let that sit alongside the $297 billion quarter. We're in a market where the vast majority of AI deployments produce no measurable value, yet capital keeps accelerating into the space. The gap between investment and returns isn't closing, it's widening. OpenAI is generating $2 billion per month, which sounds impressive until you consider its $852 billion valuation and the infrastructure costs required to sustain its operations. Sam Altman has publicly acknowledged the existence of an AI bubble. Ray Dalio has compared current AI investment levels to the dot-com era. Bill Gurley has warned of an inevitable "AI reset." When the people making the bets are acknowledging the froth, it's worth paying attention.

The counter-argument, and why it's partially right

None of this means the technology isn't real. AI is already transformative in specific applications. Code generation, drug discovery, content creation, customer service, there are genuine productivity gains in all of these areas. The infrastructure being built, data centers, custom chips, fiber networks, has tangible value that will persist regardless of which companies survive. The strongest case for current spending levels is that AI represents a platform shift comparable to the internet itself, and that under-investing in such a shift is more dangerous than over-investing. If AI really does automate a meaningful fraction of cognitive work, the companies that built the infrastructure early will capture enormous value. This argument has merit. But it also applied to the dot-com era. The internet was a platform shift. The companies that survived, Amazon, Google, eBay, did capture enormous value. The problem was that for every Amazon, there were a hundred Pets.coms. The technology being real doesn't make every valuation rational.

How many foundation model companies does the world need?

Most of Q1 2026's mega-rounds went to companies building foundation models: OpenAI, Anthropic, xAI. These are companies competing to build essentially the same product, general-purpose AI models, at massive scale. History suggests that platform markets tend to consolidate quickly. There might be room for two or three major foundation model providers, much like there are two or three major cloud providers. Funding five or six at enormous valuations implies that investors expect all of them to win. That's not how markets work. The more likely outcome is that some of these companies will fail or be absorbed, and the capital invested in them will be destroyed. Not because the technology was bad, but because there wasn't room in the market for all of them at the scale their valuations implied.

The belief system

At $297 billion in a single quarter, with 81% going to a single sector and 63% going to four companies, this isn't investment in the traditional sense. It's something closer to a belief system, a conviction so strong that it overwhelms normal due diligence, normal diversification, and normal risk management. The technology is real. The transformation is real. But $297 billion in a quarter, concentrated in a handful of companies building the same category of product, with 95% of deployments generating no measurable return? That's not a calibrated response to opportunity. That's a stampede. The dot-com era taught us that the right technology with the wrong capital allocation still ends in tears. The internet survived the crash and eventually exceeded even the wildest dot-com projections. But investors who bought at the peak waited a decade to break even, and many never did. Nobody needs $297 billion in a quarter. Not even AI.