Finance · Technology · Markets
Is AI a bubble?
Verdict based on 8 sources. The answer is genuinely split — depends what you're asking.
Last updated Mar 19, 2026 · Fast-moving — review weekly
Background
By early 2026, global investment in AI had reached hundreds of billions of dollars annually, with the top technology companies committing unprecedented capital to AI infrastructure, chips, and models. Whether this represents rational investment in a transformative technology or a speculative bubble disconnected from current returns is the defining financial debate of the moment. This page examines what the evidence shows.
The numbers defining the debate
$690B
AI capex committed by Big Tech in 2026 — nearly doubling 2025
Futurumgroup, Feb 2026
1.1%
of US GDP growth in H1 2025 came from AI-related capex — AI became the primary growth driver
89%
of companies (6,000 CEOs) report no productivity impact from AI in 3 years
NBER, Feb 2026
$500B+
AI investment projected by Goldman Sachs for 2026 — driven by enterprise demand
The central tension: $690B in AI capex. Near-zero measurable productivity contribution. A handful of stocks carrying the entire market. When investment runs this far ahead of output, some investors will be disappointed — regardless of whether the technology ultimately succeeds.
The case on both sides
🫧 Signs of a bubble
📈 Signs it's not a bubble
📚 Historical parallels
Dot-com (1995–2001)
Revenue was fictional. Many companies had no business model. Technology worked — the internet transformed everything — but most investors lost money.
Railroads (1840s–50s)
Enormous overbuilding, widespread bankruptcies. But railroads transformed the economy. Yale: "AI is the railroad era, not dot-com." Technology real; valuations not.
Electricity (1890s–1920s)
100 years from discovery to productivity gains. Massive capex, gradual returns, then transformative impact. The SF Fed uses this as the AI baseline.
Key voices
"Artificial intelligence is advancing at startling speed. Its effect on output, not so much."
"AI is everywhere except in the incoming macroeconomic data. You don't see AI in the employment data, productivity data, or inflation data."
Torsten Slok, Apollo chief economist — Fortune, Jan 2026
How sources frame the question
The Economist
UK · centre-right
Investment running far ahead of measurable output. "Solow's productivity paradox, the sequel." Technology may be real; timing is uncertain.
Financial Times
UK · centre-right
S&P 500 companies all talking about AI, few showing it in results. Sober evidence: where it works it's visible; most companies can't explain the ROI.
Yale
Academic
Most structurally bearish. Bubble bursts "through disappointment, not fraud." Technology wins; most investors lose on timing. Railroad era analogy.
Goldman Sachs
Investment bank
Bullish. Enterprise demand accelerating. Revenue growth for AI infrastructure companies is real. Raising forecasts for 2026-2030.
JPMorgan
Investment bank
5-factor diagnostic: AI "does not meet the classic criteria for a financial bubble." Genuine structural utility, not purely speculative.
Fortune / NYT
US press
Both sides. Acknowledges signs of bubble-like behavior AND real underlying demand. Outcome depends on enterprise monetisation speed.
The most honest answer
The technology is probably not a bubble. The valuations may be.
The underlying technology is demonstrably real — revenue for AI infrastructure companies is genuine, enterprise adoption is accelerating, and JPMorgan and Goldman are right that this is not dot-com fiction.
But the investment-to-output gap is historically unusual. Yale's framing may be most accurate: "The bubble bursts through disappointment, not fraud." Like the railroad era: the technology transforms society, and most investors still lose money on the timing.
The underlying technology is demonstrably real — revenue for AI infrastructure companies is genuine, enterprise adoption is accelerating, and JPMorgan and Goldman are right that this is not dot-com fiction.
But the investment-to-output gap is historically unusual. Yale's framing may be most accurate: "The bubble bursts through disappointment, not fraud." Like the railroad era: the technology transforms society, and most investors still lose money on the timing.