Apple unseated Nvidia as the world’s most valuable company on 17 July, closing the week with a market capitalisation of $4.88 trillion against Nvidia’s $4.86 trillion after a 3.5 percent daily decline in the chipmaker’s shares. The Philadelphia Semiconductor Index entered bear-market territory, down more than 20 percent from its June peak, as a broader sell-off in technology stocks wiped out a portion of the year’s earlier gains. The shift marks a tangible reassessment of the artificial-intelligence trade, with capital rotating from the primary hardware beneficiaries toward firms seen as better positioned to monetise AI through services and installed bases.

The immediate trigger was the emergence of Kimi K3, an open-source model from Chinese start-up Moonshot AI that claims performance parity with leading Western systems at a fraction of the cost. The development, viewed from trading desks in Frankfurt and Tokyo, threatens the pricing power of proprietary models from OpenAI and Anthropic, both of which are preparing blockbuster public listings. Simultaneously, investors are questioning whether the estimated $5.3 trillion in projected AI infrastructure spending can generate commensurate returns. Analysts at Bank of America described the semiconductor retreat as a “summer reset” rather than a fundamental reversal, noting that the sector’s five-year return of over 250 percent dwarfs the broader market. Yet the unease is spreading: credit spreads on AI-linked corporate debt have widened, and the combined backlogs of three major US construction contractors have risen sixfold, exposing a severe shortage of skilled electricians and HVAC installers that is delaying data-centre projects and driving up local labour costs in Texas and Arizona.

Central bankers in Washington and Frankfurt are now grappling with the inflationary side-effects of the build-out. The Federal Reserve’s June minutes cited AI-related demand for chips, specialised labour, and energy as a source of price pressure, even as Chair Kevin Warsh publicly minimised the risk. Economists in São Paulo and London point to a paradox: the capital expenditure boom is stoking inflation today, while the expected productivity gains that could dampen prices remain years away. This “K-shaped” dynamic—where tech-linked sectors overheat while others stagnate—complicates monetary policy, leaving the Fed with a single instrument to manage a bifurcated economy.

Governments are responding with a wave of regulatory initiatives. Australia’s Labor government announced a framework to govern automated decision-making in federal agencies, a digital duty of care that will oblige AI firms to conduct risk assessments, and a world-first standard for data-centre siting to limit strain on energy grids and water supplies. In the United States, the Federal Trade Commission proposed a policy statement warning that AI systems steering consumers toward undisclosed sponsored results may constitute deceptive practices. These moves reflect a broader push, from Canberra to Washington, to impose guardrails before the technology’s labour-market and fiscal consequences deepen. A letter signed by 16 Nobel laureates and over 200 economists urged governments to act now, warning that AI could trigger a transformation larger than the Industrial Revolution but compressed into a single decade.

The next factual milestone arrives on 22 July, when Alphabet and Tesla report quarterly earnings. Their results will provide the first concrete read on whether the AI spending cycle is translating into revenue growth beyond the chipmakers, and whether the recent rotation has further to run.