Google Boss: AI Boom Has 'Irrationality' and Europe Should Pay Attention

Pichai's comments echo a concern that has been building in parts of the European investment community. The European Investment Bank's annual investment survey has repeatedly flagged a gap between AI capital flows and demonstrable productivity gains in most sectors. Meanwhile, Margrethe Vestager, European Commission Executive Vice-President and the EU's long-serving competition chief, has made clear that regulatory scrutiny will follow wherever market concentration and inflated valuations intersect. Both signals point in the same direction: the current pace of AI valuation growth is not supported uniformly by underlying commercial fundamentals.

The dot-com parallel that Pichai invokes is instructive. That crash produced lasting economic damage across firms with entirely legitimate businesses alongside the purely speculative ones. European technology markets felt the reverberations for years. However, the analogy has limits. AI infrastructure investment, unlike many dot-com era assets, produces tangible capacity: data centres, specialised chips, and operational capability that retains value even if specific firms collapse. Any AI correction would therefore look different from a pure bubble pop, more a thinning of the herd than a wholesale wipeout.

The specific indicators of bubble dynamics are, however, hard to dismiss. Valuations of AI companies have risen faster than corresponding revenue growth in a significant number of cases. Investment flows to AI startups have reached levels that historically correlate with correction phases. And some capability claims have failed to survive independent evaluation, a pattern that Yoshua Bengio, Scientific Director of Mila and a leading voice on AI safety and governance, has warned about in the context of over-reliance on benchmark performance rather than real-world utility.

Google's full-stack defence and what it means for European buyers

Despite his cautious framing, Pichai is not predicting a crash; he is arguing that Google is better positioned than most to survive one. The company's strategy spans custom silicon through its TPU programme, proprietary model development via Gemini, frontier research through Google DeepMind based in London, and vast proprietary data assets. This vertical integration, built over the better part of a decade, provides a degree of insulation that narrowly specialised competitors lack.

European technology firms and enterprise buyers should take note of the structural logic. Companies with capability across multiple layers of the AI stack, from silicon and infrastructure through to models and applications, have more options during market turbulence than those that depend on a single external provider. In Europe, ASML, the Dutch semiconductor equipment maker whose lithography machines underpin virtually every advanced chip produced globally, occupies a similarly defensive position: indispensable to the supply chain regardless of which AI firm wins the application layer. Mistral AI, the Paris-based frontier model developer backed by European and international investors, is pursuing a comparably integrated approach at the model and application layer, aiming to reduce European dependence on American hyperscalers.

For European enterprise customers, the practical implication is straightforward: multi-vendor AI strategies reduce exposure to vendor-specific instability. Over-committing to a single provider during a period of uncertain valuations can leave buyers exposed if that vendor faces commercial stress. Flexibility in infrastructure contracts and a focus on AI applications with demonstrated, measurable value rather than speculative capability should be the operating principle for procurement teams right now.

The sectors most exposed to correction

Not all parts of the AI market carry equal risk. The sectors most vulnerable to a correction include AI application companies without strong commercial traction, infrastructure providers whose capacity has outrun likely near-term demand, and firms whose valuations rest primarily on narrative rather than current revenue. Companies with strong recurring enterprise revenue, clear paths to profitability, and valuation-to-revenue ratios that would survive a scrutiny round are materially less exposed.

European AI investment patterns differ from those in North America in ways that provide some structural cushion. Investment across the continent has generally been more fragmented, spread across a larger number of smaller rounds rather than concentrated in a handful of mega-valuations. That dispersion reduces the systemic risk of a single high-profile failure cascading through the broader ecosystem. Regulatory discipline imposed by the EU AI Act, which entered into force in 2024, is also pushing companies toward documented performance claims and auditable systems, conditions that make purely story-driven valuations harder to sustain.

Historical parallels and the infrastructure lesson

Beyond the dot-com comparison, the 19th-century railway bubbles offer a second, arguably more precise, analogy. Speculative investment in railway companies destroyed enormous amounts of investor capital. The physical rail infrastructure those investments built continued producing economic value for generations, even after the companies that funded it failed. The distinction between technology value and specific firm value is the key analytical task for anyone allocating capital in the AI sector today.

AI may follow the same pattern. Specific AI companies will likely face correction even as AI capability continues improving and embedding itself in the productive economy. Data centres, chips, and specialised AI capacity are likely to retain long-term value regardless of which firms funded them. European investors and enterprise strategists should price this distinction into their planning. Backing durable infrastructure and applications with demonstrated returns is a fundamentally different bet from backing a valuation story, even when both look superficially like AI investment.

What Pichai's candour actually signals

Most chief executives avoid commentary that could be read as bearish on their own sector. Pichai's willingness to use the word irrationality in public sits outside normal corporate communication norms. Two explanations are plausible and not mutually exclusive. First, he may be expressing genuine concern shared by executives who lived through the dot-com crash and recognise the pattern. Second, the framing benefits Google specifically: by acknowledging market risk while emphasising Google's defensive full-stack positioning, Pichai signals to investors that Google is built to benefit from consolidation that would harm weaker competitors.

For European observers, both readings point to the same conclusion. The technology is real and the value creation is real in specific applications. But specific valuations, in parts of the market, are running ahead of what the underlying commercial fundamentals can currently support. The correction, if and when it arrives, will not erase AI as a productive force; it will thin the field and reward firms, whether American, European, or otherwise, that built on substance rather than story. Distinguishing between the two is the central analytical challenge for anyone deploying capital or building strategy in the sector right now.