Amazon, Microsoft, Alphabet and Meta have together committed more than $1.1 trillion to AI infrastructure since the boom began in 2023. The scale of that spending marks a decisive turn away from the asset-light software model that defined the tech sector for two decades.
Scale of AI spending
The four hyperscalers plan to invest another $745 billion this year, making capital intensity a core cost of competing in the AI race. Historically, investors rewarded companies that required little capital and generated high margins. Today those same firms are spending at a level the sector has never seen.
From models to finance
As large language models become more interchangeable, the advantage shifts from owning a unique model to financing and operating the supporting infrastructure at the lowest cost. Satya Nadella recently remarked that "every model is substitutable", while Andy Jassy predicted there will soon be "at least half a dozen" comparable AI models.
"Every model is substitutable," said Satya Nadella.
"There will be at least half a dozen comparable AI models," said Andy Jassy.
This change means hyperscalers are building flexible data centres capable of running many models, rather than betting on a single breakthrough. Financing and scale now matter more than the frontier model itself.
Financing the AI engine
Companies such as Nvidia are partnering with firms like Apollo, Blackstone and Goldman Sachs to mobilise over $500 billion of additional capital for AI infrastructure. Google has created a $200 billion financing structure with Broadcom, Apollo, Blackstone and Morgan Stanley to fund Anthropic's chips and data centres.
These arrangements underline how the competition has expanded into the finance sector. The biggest balance sheets, Microsoft, Amazon and Google, can secure cheap capital and generate revenue from the same data centres used to train AI, giving them a durable edge even if models become commoditised.
Implications for the market
While the infrastructure spend is massive, the profitability of the models themselves remains uncertain. Both OpenAI and Anthropic are still loss-making, yet capital continues to flow. The shift is already affecting other tech firms. IBM saw a 25% one-day share price drop in July after customers delayed software purchases to secure AI infrastructure.
Big tech earnings show cloud revenues are rising sharply: Microsoft's cloud business grew 32% to $39.3 billion, Amazon Web Services rose 37% to $42.2 billion, and Google's cloud segment surged 82% to $24.8 billion. These gains have lifted market valuations, with Microsoft adding a record $450 billion in a single day.
In contrast, Apple has restrained its AI spend, and investors briefly rewarded that discipline with a $5 trillion valuation, positioning the iPhone maker as a control group for the sector.
What comes next?
Investors now face two divergent bets: backing firms that pour capital into AI infrastructure, or supporting those that maintain financial discipline. New entrants such as SpaceX and sovereign wealth funds like Saudi Arabia's PIF and Abu Dhabi's MGX could challenge the dominance of the current hyperscalers.
Regardless of who ultimately wins, the rules of competition have already changed. Companies with the strongest balance sheets, cheapest capital and highest utilisation of their infrastructure will have the edge as AI models converge in capability. In effect, AI is becoming a financial engineering business.