The AI Bet

Silicon Valley has built a vast economic edifice on an assumption that remains surprisingly untested: that once artificial intelligence becomes sufficiently capable, almost everyone will use it as intensively as software engineers do. The assumption matters because extraordinary sums are being committed to make it possible. Technology companies are spending hundreds of billions of dollars on data centres, specialised chips and the electricity needed to run them. The investment is premised not merely on AI becoming useful, but on it becoming an indispensable and heavily consumed utility across the economy. So far, software developers offer the strongest proof that this can happen. AI coding assistants have moved rapidly from novelty to everyday tools. Engineers increasingly use AI to generate, debug and rewrite code, allowing a single worker to accomplish tasks that once required considerably more time. For technology companies, programmers are therefore the model customer: digitally sophisticated, frequent users whose work generates an obvious and measurable demand for computing power. But coders can supervise an AI system without surrendering professional responsibility for its mistakes. Advertisement Many other occupations do not offer coding's advantages. A lawyer cannot safely approve a contract merely because an AI system has produced convincing prose. A doctor cannot delegate clinical judgement to a chatbot. A corporate executive cannot blame an algorithm after an important negotiation goes wrong. In fields where mistakes carry legal, financial or human consequences, somebody must remain accountable. This may prove to be the central economic constraint on the AI revolution. The problem is no longer simply whether machines can perform tasks. Advertisement Increasingly, they can. The harder question is whether institutions are willing to reorganise work around systems that remain imperfect and, in many cases, difficult to audit. That distinction should concern investors as much as workers. History offers ample warnings. The railways, electricity and the internet, all transformed economies but turned out to be spectacularly bad investment decisions. Being right about the future does not guarantee that today's valuations are right. The same may hold for AI. There is little reason to doubt that artificial intelligence will spread far beyond software engineering. It is already entering law, medicine, finance, marketing and administration. But widespread adoption is not the same as intensive use. An accountant who consults an AI assistant several times a day does not necessarily generate anything like the computing demand created by thousands of programmers constantly running sophisticated coding agents. The danger, therefore, is not an AI winter caused by technological failure. It is a mismatch between infrastructure built for explosive demand and an economy that adopts the technology more gradually. The great question confronting the AI industry is not whether people will use artificial intelligence. They undoubtedly will. It is whether enough of them will use enough of it, often enough, to pay for the colossal machinery now being built. That answer may determine whether the AI boom becomes an industrial revolution ~ or an infrastructure bubble.