$242 billion bought nothing new

Q1 2026 saw $242 billion pour into AI startups, roughly 81% of all global venture capital. Four of the five largest funding rounds in history closed in a single quarter. OpenAI raised $122 billion at an $852 billion valuation. Anthropic took $30 billion. xAI, $20 billion. Waymo, $16 billion. Those four firms alone captured $188 billion, or about 65% of every venture dollar deployed on Earth. The numbers are staggering. But here's the question nobody funding these rounds seems eager to answer: what, exactly, did all this money buy that you use differently today?

The receipts

Let's be specific. Global venture investment hit roughly $300 billion in Q1 2026 across about 6,000 funded companies. AI swallowed the vast majority of it. Late-stage rounds totaled $246.6 billion across 584 deals, meaning the capital is concentrating in fewer and fewer hands at larger and larger checks. This isn't seed-stage experimentation. This is mature capital markets making enormous bets on a handful of companies. And the implicit promise behind every one of those checks is the same: these companies will build products that change how the world works. So, have they?

What the previous $100 billion bought

Benchmarks, mostly. The Stanford HAI 2026 AI Index tells a revealing story. Models are objectively getting better. Gemini Deep Think earned a gold medal at the International Mathematical Olympiad. The best models now correctly answer over 50% of questions on Humanity's Last Exam, up from 8.8% just a year ago. Claude Opus 4.6 and Gemini 3.1 Pro are pushing boundaries that seemed fixed twelve months ago. But here's where it gets awkward. The top-performing model that can solve competition-level math reads an analog clock correctly only 50.1% of the time. AI agents jumped from 12% to roughly 66% task success on OSWorld, which tests agents on real computer tasks, yet they still fail one in three attempts on structured benchmarks. Documented AI incidents rose to 362 in 2025, up from 233 the year before. Hallucination rates across 26 leading models range from 22% to 94%. Capabilities are advancing. Reliability is not keeping pace.

The adoption gap nobody talks about

The distance between "we're investing in AI" and "we're using AI in production" is widening, not shrinking. According to one widely cited dataset, 79% of enterprises say they've adopted AI agents in some form. But only about 11% have agents running in production. That's a 68-point gap between adoption and deployment. A Cisco survey found an even starker picture: 85% of enterprises have AI agent pilot programs underway, but only 5% moved those agents into production. The reason isn't technical capability. It's trust. As Cisco's Jeetu Patel put it, the difference between delegating tasks to agents and trusted delegating of tasks to agents is the difference between market dominance and bankruptcy. So we have a quarter-trillion dollars flowing into AI companies while the overwhelming majority of enterprises are still running pilots they don't trust enough to ship. That's a disconnect worth sitting with.

The Jevons paradox of AI capital

In 1865, William Stanley Jevons observed that as coal engines became more efficient, coal consumption didn't decrease. It increased. Efficiency lowered the cost per unit of work, which unlocked new applications, which drove aggregate demand higher than before. The same dynamic is playing out in AI, but with capital instead of coal. Every efficiency gain in compute, every drop in inference cost, every improvement in model architecture doesn't reduce spending. It accelerates it. DeepSeek showed the world that capable models could be built for less. The market's response was not to spend less. It was to spend more, faster, on more things. Goldman Sachs expects AI hyperscaler capex to exceed $500 billion in 2026. Gartner forecasts worldwide AI spending will hit $2.52 trillion, a 44% year-over-year increase. The IEA reports that capital expenditure by the five largest tech companies surged past $400 billion in 2025 and is set to increase by another 75% in 2026. The capital isn't scaling innovation. It's scaling compute. And cheaper compute is creating more demand for compute, not less.

Funding velocity is not value creation velocity

The dotcom parallel is imperfect but instructive. In the late 1990s, capital poured into internet companies at a pace that felt justified by the scale of the opportunity. And the opportunity was real. The internet did change everything. But funding velocity ran far ahead of value creation velocity, and the correction was brutal. In February 2000, Gartner predicted that B2B eCommerce would soar to $7.29 trillion by 2004, up from $145 million in 1999. Today, Gartner predicts AI will account for nearly all IT spending by 2030. The pattern of breathless extrapolation from real trends feels uncomfortably familiar. BCA Research recently noted four signs that the AI trade could be entering a late-stage "melt-up," a sharp rally echoing 1999. Reuters reports that a data center originally budgeted at $1 billion can easily inflate to $1.3 billion or more. Alphabet's return on invested capital is expected to fall from 51% last year to about 36% by 2030. Microsoft's is projected to drop from 95% in 2020 to 36% in the same period. The infrastructure is getting built. The returns on that infrastructure are compressing.

The bill is coming

The Stanford AI Index flags something else that deserves attention: the environmental costs. Data center electricity consumption surged 17% in 2025, outpacing global electricity demand growth of 3%. AI-focused data centers grew even faster. By some estimates, data center electricity use could approach 1,050 TWh by 2026, which would make data centers the fifth-largest energy consumer in the world, between Japan and Russia. Cornell researchers estimate that by 2030, the current rate of AI growth would annually put 24 to 44 million metric tons of CO2 into the atmosphere, the equivalent of adding 5 to 10 million cars to U.S. roadways. Water consumption is equally sobering: 731 to 1,125 million cubic meters per year, equal to the household water usage of 6 to 10 million Americans. AI systems may already have a carbon footprint equivalent to New York City's. These costs don't appear on anyone's cap table, but they're real.

The counterargument, and why it's partly right

None of this means the money is wasted. Infrastructure investment often looks foolish before it looks visionary. AWS burned cash for years before it became the profit engine that funds everything else Amazon does. The fiber optic cables laid during the dotcom boom, often by companies that went bankrupt, became the backbone of the modern internet. Some of this $242 billion is genuinely building the foundation for transformative applications. The models are getting meaningfully better. The tooling is maturing. Enterprise adoption, while slow, is real and growing. Over half of large organizations now have agents in production, and that number is climbing. The question isn't whether AI infrastructure will eventually generate enormous value. It almost certainly will. The question is whether the pace of capital deployment has decoupled from the pace of value creation, and if so, what happens when the gap becomes impossible to ignore.

What I keep coming back to

When I look at Q1 2026, I see a quarter where four companies raised $188 billion and the world barely blinked. Where the biggest funding round in history, $122 billion for a single company, was treated as expected. Where 81% of all venture capital went to one category of technology. I don't think this is irrational. I think AI is a genuinely transformative technology. But I also think there's a meaningful difference between "this technology will change the world" and "this specific allocation of capital, at this specific pace, will generate returns." The dotcom era proved that both statements can be true at the same time, and investors can still lose everything. $242 billion in 90 days. The question isn't whether AI is real. It's whether the capital markets have gotten ahead of the products. And right now, with most enterprises still unable to ship what they've built, the honest answer is: probably.