The world’s largest technology companies are preparing to spend more than 1.3 trillion dollars on artificial intelligence infrastructure in 2027, a figure that now outruns the cash those companies generate. S&P Global Ratings projects combined spending by six hyperscalers, Alphabet, Amazon, Microsoft, Meta, Oracle and SpaceX, will reach 1.3 trillion in 2027, up about 50 percent from the 870 billion dollars expected in 2026, and that all six will produce negative free operating cash flow in both years, with recovery not projected until 2029.
Goldman Sachs reaches a similar total by a narrower route, forecasting that five United States hyperscalers will spend 1.2 trillion dollars in 2027, a 54 percent increase, funded in part by about 400 billion dollars in investment grade bond issuance. Goldman estimates the group needs roughly 300 billion dollars in annual AI revenue to break even on the buildout, a figure that highlights the gap between technical adoption and economic return. Morgan Stanley analysts have separately said Alphabet, Amazon, Meta and Microsoft are expected to raise total capital expenditure by more than 50 percent in 2027 compared with 2026.
Company disclosures show the momentum is already strong. Amazon recorded about 44.2 billion dollars in quarterly capital expenditure, Alphabet 35.67 billion dollars, and Microsoft 30.88 billion dollars, while Meta raised its full year guidance to between 125 and 145 billion dollars, citing components and data centre costs. Microsoft’s fiscal third quarter capital expenditure jumped 84 percent year over year, while its AI revenue surpassed a 37 billion dollar annualised run rate.
Only one company in positive territory
S&P’s company level projections sharpen the story. Alphabet is projected to spend 357 billion dollars in 2027 with negative free cash flow of 82.7 billion dollars, Amazon 319.1 billion dollars with negative 60.1 billion dollars, and Meta 164 billion dollars with negative 3.5 billion dollars. Only Microsoft, at 189 billion dollars of spending, is projected to have positive free cash flow next year, at 33.6 billion dollars. Microsoft also carries more than 329 billion dollars in future lease obligations, ahead of Meta at 279 billion dollars, a reminder that headline capital expenditure understates total commitments.
The rating agency expects 2028 to mark an inflection point, when AI related revenue accelerates and capital expenditure growth begins to moderate. That forecast is the central bet of the whole buildout: that data centres, chips, networking and power infrastructure bought today will earn their keep within a few years, even if they burn cash in the meantime.
What it means for digital business
For smaller digital businesses, the hyperscaler race cuts both ways. Cheaper and more powerful AI infrastructure, rented from the same companies doing the spending, can lower the cost of building AI products. At the same time, the scale of the buildout raises costs for power, components and talent, and it concentrates bargaining power in a handful of suppliers. Firms that plan around rented capacity, rather than owned infrastructure, are less exposed to the financing risk, but more exposed to price and access decisions made by the hyperscalers themselves.
The question for investors, and for the wider digital economy, is no longer whether demand for AI exists. It is whether revenue, power supply, data centre capacity and debt markets can scale fast enough to earn an adequate return on more than a trillion dollars a year. The 2027 spending plans say the companies believe they can. Their cash flow projections say the proof will take until at least 2029.
