“I think no company is going to be immune, including us.” The quip from Sundar Pichai, uttered from Google’s California headquarters, fell as both a warning and a dare on a market drunk on artificial intelligence. Alphabet’s valuation has surged to $3.5 trillion doubling in seven monthsyet its chief executive confesses to “elements of irrationality” redolent of the late‑1990s dot‑com boom.
The parallels are not hard to draw. Valuations then, as now, were pushed far from fundamentals by exuberance, with the collapse of early internet firms when capital dried up. Today, analysts question whether the market pricing of AI is sustainable, with a tortured web of $1.4 trillion in deals around OpenAI contrasting with revenues of less than one-thousandth that figure. Pichai’s warning echoes Alan Greenspan’s 1996 call against “irrational exuberance,” underlining how even dominant players might get caught in a correction.
Owning its “full stack” of AI technology from proprietary models and YouTube’s vast datasets to frontier science and in‑house semiconductors is how Google will weather such turbulence. It has designed custom AI superchips that go toe to toe with accelerators from Nvidia, whose market dominance has propelled the company to a $5 trillion valuation. Chips integrating high‑bandwidth memory, advanced interconnects, and optimized tensor processing units can cut latency in large‑scale training and inference workloads and provide advantages in cost relative to third‑party GPUs.
Yet the hardware race is inextricably linked to the energy challenge: AI already represents 1.5% of global electricity consumption, with the demand from accelerated servers expected to surge 30% every year. Hyperscale AI data centres some devouring as much power as a large city are stressing grids designed for much more modest load growth. According to the International Energy Agency, data centres could hit as high as 945 TWh of annual consumption by 2030 a nearly 3% share of global demand. Utilities face multi‑year lead times for generation and transmission projects; meanwhile, data centres can be built in less than two years, which could create a mismatch risking outages and price spikes.
Google’s £5bn UK expansion illustrates both the opportunity and the strain. The investment includes infrastructure, research, and the Waltham Cross data centre, which uses advanced air‑cooling and heat‑recovery systems to minimize its environmental impact. Deals with ENGIE for offshore wind and with Shell Energy Europe for battery storage highlight its aim for 95% carbon‑free energy for all UK operations by 2026.
But Pichai, chief executive of Alphabet, concedes that climate targets are slipping under the weight of AI’s “immense” energy needs, even as the parent maintains a 2030 net‑zero goal. The geopolitical dimension is no less stark. Nations such as the US, China, and Gulf states are investing hundreds of billions in AI infrastructure, locking in domestic compute capacity and clean‑energy baseloads.
The UK, despite ambitions to be the world’s number‑three AI superpower, holds only about 3% of global computing power much of it not optimised for AI and pays industrial electricity prices 46% above the IEA median. Absent reforms to planning, grid connections, and energy strategy, it is in danger of becoming dependent on foreign infrastructure for critical AI workloads. Workforce impacts add yet another layer of complexity.
Goldman Sachs estimates that AI could raise productivity in developed markets by 15% when fully adopted, temporarily lifting unemployment by 0.5 percentage points during the transition. Displacement rates may range from 3% to 14%, with roles in programming, accounting, and customer service among the most exposed. But adoption remains uneven, and historical patterns indicate that widespread disruption will unfold over decades, not months.
According to IMF analysis, 60% of jobs in advanced economies could be affected half benefiting from augmentation, half at risk of automation potentially exacerbating inequality unless retraining and safety nets are strengthened. To investors and industry strategists, the message is clear: AI’s rise is as much about physical and energy infrastructure as about algorithms and valuations.
Pichai’s twin recognition that AI is “the most profound technology” humanity has worked on, and that the current market may be overshooting frames the paradox of the sector. There will be winners in connecting semiconductor innovation, grid resilience, and workforce adaptation to capital cycle realities, thereby sidestepping the fate of the previous bubbles in harnessing the transformative power of AI.
