Alphabet, the parent of Google, set out plans to raise about $80bn in fresh equity to bankroll the data centres and chips behind its artificial-intelligence ambitions, one of the largest capital-raising efforts ever undertaken by a technology company. The package combines underwritten share and convertible offerings, a direct sale to Berkshire Hathaway and a programme to sell stock gradually into the open market.

The structure pairs roughly $30bn in underwritten offerings of common shares and mandatory convertible preferred stock with a $10bn private placement taken up by Warren Buffett's Berkshire Hathaway, an unusual endorsement from an investor long sceptical of richly valued technology stocks. A further $40bn would be raised through an at-the-market programme, under which Alphabet can sell shares directly into the market beginning in the third quarter.

Alphabet said the proceeds would go toward general corporate purposes, including the capital expenditure needed to scale its global compute capacity and meet what it described as unprecedented customer demand for AI services. The company has guided to capital spending of between $180bn and $190bn this year, with a further significant increase expected in 2027.

The decision to tap equity markets so heavily marks a shift for a company that has historically funded its expansion from its own vast cash flows. It reflects how the race to build AI infrastructure has pushed even the most profitable technology firms toward outside financing, as the cost of graphics processors, power and real estate for data centres climbs into the hundreds of billions.

For investors, the size of the raise sharpened a debate that has run through the AI boom: whether the enormous sums being committed to computing capacity will generate commensurate returns, or whether the industry is overbuilding ahead of demand. Berkshire's participation lent weight to the bullish case, even as some analysts questioned the dilution to existing shareholders.

The announcement positioned Alphabet alongside rivals racing to lock in scarce chips and energy for the next generation of AI models. How quickly the spending translates into revenue, rather than depreciation, will shape the verdict on whether the build-out was prescient or excessive.