The scale of the global AI build-out moved from striking to historic this week, as fresh disclosures crystallised just how much capital the largest US tech companies are now committing. According to BizNews Daybreak’s 4 May 2026 markets round-up, the latest results from Google, Meta and Amazon show AI infrastructure spend has roughly doubled year-on-year, with annualised capex across the hyperscaler cohort now running at around $72 billion. The same source flagged the figures as confirmation that AI has become the single largest capital cycle in technology history, dwarfing the late-1990s telecoms boom on both pace and concentration.
The composition of the spend is as significant as the headline number. Hyperscalers are no longer simply buying GPUs; they are committing multi-year capital to land, power purchase agreements, custom silicon, networking and bespoke data-centre designs optimised for training and inference at frontier scale. That shift has profound implications for grid operators, utilities and planning authorities in every market where the big three are expanding, including the UK and Ireland.
The corporate logic behind the capex is also reshaping workforces. Weekly AI trackers noted this week that Meta has cut thousands of jobs to free up cash to redirect into AI infrastructure, reinforcing an “AI or bust” posture now visible across the hyperscaler cohort. Investors have so far rewarded the trade-off, but the margin of safety is narrowing as depreciation from this year’s spend begins to bite into reported earnings.
For Silicon readers, the practical implication is competitive. With $72 billion a year chasing power, land and chips at hyperscale, smaller UK and Scottish operators face an increasingly tilted playing field on both compute pricing and grid access through 2026 and beyond.
