The Technology Analysis
Technology cycles have historically followed a pattern that is now well-documented: breakthrough discovery, rapid capability development, inflated expectations, capability plateau, disillusionment, and then the quiet period during which the technology is embedded into economic infrastructure in ways that eventually produce transformation more substantial than the initial hype suggested.
We are currently in different phases of this cycle simultaneously for different AI capability clusters. Large language models are in the late expectation inflation phase; the next 18 months will likely involve significant recalibration of near-term projections. Meanwhile, narrow AI applications in drug discovery, materials science, and computer vision are already delivering the quiet infrastructure transformation that will eventually be recognised as the period when the technology became genuinely economically significant.
The Structural Shifts
The labour market implications of current AI trajectories are being analysed through an inappropriate framework β the question "which jobs will AI replace?" assumes a stability of job structure that historical technology transitions do not support. The more historically accurate question is "how will the distribution of tasks within jobs change, and which new job categories will emerge from the complexity AI introduces?"
Historical evidence from electrification, computing, and the internet suggests that the net employment effects of transformative technology are typically positive over 20-year horizons but negative over 5-year horizons for specific skill clusters. The challenge is not the long-run equilibrium but the transition β specifically, whether skill retraining infrastructure can operate at the required pace and scale.
The Investor Perspective
The capital allocation patterns in AI infrastructure are revealing about which technology bets major players consider most durable. The sustained investment in electrical grid infrastructure and data centre construction β long-duration physical assets β suggests that sophisticated investors expect the AI compute demand cycle to extend well beyond what typical technology boom cycles would suggest. Physical infrastructure bets are not made by investors who expect a near-term contraction.