Hyperscalers forecast $805B capex in 2026

Data centers are the story here. Not apps, not chatbots, not vague “AI investment” talk — actual buildings, servers, chips, transformers, and power contracts. In late April, Meta, Alphabet, and Microsoft all pushed 2026 capital-spending plans higher, while Amazon and Oracle kept defending already huge buildouts. Put together, the five biggest cloud and AI infrastructure players now look set to spend something like three quarters of a trillion dollars this year, not $805 billion. The number in circulation seems overstated, but the underlying point is real: the buildout just got bigger. (investor.atmeta.com) ### Who actually raised spending? Meta did it explicitly. On April 30, it raised 2026 capex guidance to $125 billion to $145 billion, up from $115 billion to $135 billion, and said higher component pricing was part of the reason. Alphabet did the same on April 29, lifting full-year capex guidance to $180 (investor.atmeta.com)gher component prices. (investor.atmeta.com) ### What about Amazon and Oracle? Amazon hasn’t put the same kind of tight quarterly range around 2026 capex in its earnings release, but Andy Jassy has been blunt that the company expects to spend roughly $200 billion this year, mostly for AI infrastructure. Oracle is smaller than the other four but stil(investor.atmeta.com)t its capex around the mid-$30 billions. (cnbc.com) ### So where does the total land? If you take the midpoint of Meta’s range, the top end of Alphabet’s updated range, Microsoft’s stated figure, Amazon’s roughly $200 billion, and Oracle around $35 billion, you get about $750 billion. Use lower assumptions and it’s closer to $740 billion. That is still enormous. But it is not the same as a verified $805 billion forec(cnbc.com)xing company guidance with aggressive analyst extrapolation. (investor.atmeta.com) ### Why are they spending this hard? Because demand is still outrunning supply. Microsoft said it remains focused on delivering cloud and AI infrastructure at scale. Alphabet said 2027 capex should rise again from 2026 levels. Amazon is arguing that if customer demand is there, being conservative would be the bigger mistake. Basically, the hyperscalers think the scarce thing now is not AI demand — it’s capacity. (microsoft.com) ### Why do component prices matter so much? Because this isn’t just a chip order. A modern AI data center is a stack problem. You need GPUs or custom accelerators, high-bandwidth memory, networking gear, cooling, backup systems, switchgear, transformers, and a lot of electricity. Meta and Microsoft both directly flagged higher component pricing. Alp(microsoft.com)gy-related operating costs. (sec.gov) ### Does this spill beyond tech? Yes — especially into semis and power equipment. IDC now sees the global semiconductor market reaching $1.29 trillion in 2026, up 52.8%, with AI infrastructure and hyperscaler spending doing most of the work. That kind of demand can lift suppliers, but it also tightens supply chains for everyone else buying memory, networking silicon, and electrical equipment. (idc.com) ### What’s the bottom line? The real news is not a clean $805 billion number. It’s that the biggest cloud companies just used earnings week to confirm another step up in AI infrastructure spending — and some of them said costs are rising as they do it. That makes this less like a normal tech capex cycle and more like an industrial buildout with software economics attached. (investor.atmeta.com)