Cloud giants pour nearly $600B into capex as AI demand surges
Amazon and Google have both raised their capital expenditure forecasts for 2026, underlining how the AI boom continues to outstrip the enormous sums cloud providers are already investing in datacentre capacity. Alongside Microsoft, the two hyperscalers are collectively pouring close to $600 billion into infrastructure this year, driven by surging demand for the computing power needed to train and run AI models, even as their cloud divisions report accelerating revenue growth.
Amazon Web Services posted $42.2 billion in second-quarter revenue, up 36.7 per cent year-on-year and its fifth straight quarter of accelerating growth, putting it on a $169 billion annualised run rate. Amazon now expects to spend around $220 billion in cash capex in 2026, up from an earlier estimate of $200 billion, partly due to rising memory costs, with CEO Andy Jassy warning that demand will still outstrip capacity into 2027. Google similarly lifted its 2026 capex guidance to $195-205 billion, from $180-190 billion previously, as cloud revenue grew 82 per cent to $24.8 billion, with CFO Anat Ashkenazi saying the industry remains supply-constrained despite years of heavy investment.
- Amazon and Google both raised 2026 capex forecasts amid AI demand
- Amazon now expects to spend around $220 billion on infrastructure this year
- Both firms say demand still outpaces their capacity expansion
New here? Start with this
Amazon, Google and Microsoft are the world's biggest cloud computing providers, meaning they rent out the servers and data storage that power much of the internet. All three are now spending vast sums to build new datacentres packed with specialised computer chips, because running artificial intelligence systems like chatbots requires far more computing power than most existing technology needs.
This spending, known as capital expenditure or "capex", covers physical infrastructure such as buildings, servers and equipment. Companies keep raising their spending plans because demand for AI computing power has continued to grow faster than they can build capacity to meet it, even after several years of heavy investment.
This matters because these three companies' decisions shape much of the wider technology industry, from the chipmakers who supply them to the businesses that rely on their cloud services to run AI tools. Their spending is also watched closely by investors as a signal of how sustainable the current boom in AI investment might be.
Both sides, in good faith
The strongest fair case each way — we don't pick a winner.
The case for
Supporters of this scale of investment argue that the numbers speak for themselves: cloud revenue is accelerating even as capacity expands, and executives at both firms say they are still supply-constrained despite years of heavy spending. From this view, pouring capital into datacentres is simply rational business behaviour, meeting demonstrable customer demand rather than speculative excess, and firms that under-invest risk ceding a generational computing shift to rivals who build ahead of need.
The case against
Sceptics counter that committing close to $600 billion in a single year, on top of already enormous prior outlays, concentrates enormous risk on the assumption that AI demand keeps compounding indefinitely. They point to the history of infrastructure build-outs, such as the dot-com era's fibre glut, as a caution that today's growth in AI spending could outpace real economic returns, leaving shareholders exposed to under-utilised assets, rising input costs such as memory, and heavy depreciation if demand growth ever slows or plateaus.