bbc.com web signal

Alphabet burns $5.9bn cash as AI capex plan hits $205bn

TL;DR

  • Alphabet posted negative free cash flow of $5.9 billion last quarter, its first negative print in at least a decade.
  • Second-quarter capital expenditure hit about $45 billion, split roughly 60% into servers and 40% into data centres.
  • Alphabet raised its 2026 AI spending outlook to as much as $205 billion, up from a prior ceiling of $190 billion.

Alphabet's second-quarter numbers landed with a strange split screen. Revenue climbed to $119.8 billion, up 23% year on year, according to the BBC's write-up. But free cash flow came in at negative $5.9 billion, the first time the company has posted a negative print in at least a decade. The cause is not a business problem in the traditional sense. Capital expenditure hit roughly $45 billion in a single quarter, and essentially all of it went into AI infrastructure.

The eye-catching line is the updated guidance. Alphabet now expects to spend up to $205 billion on AI-related capex this year, an increase from $190 billion. CFO Anat Ashkanazi told analysts that 'the demand still outpaces that investment' and that the company 'will continue to invest' as long as attractive opportunities exist. Roughly 60% of the quarterly capex went to servers and 40% to data centres, so this is physical, long-lived hardware spend rather than software costs.

Why this matters if you are not a Google shareholder: the largest single buyer of AI infrastructure just moved its order book up again, and signalled that 2027 spending will step up as well. That is a demand signal for chip suppliers, power and cooling vendors, and data centre landlords. It is also the clearest statement yet that hyperscaler capex is not mean-reverting soon. Sundar Pichai framed the shift as 'early innings in a shift across multiple areas' and called the returns plan 'disciplined', which is roughly the phrasing investors were listening for.

The honest caveat is that the market did not fully buy it. Shares reportedly slipped around 4% in after-hours trading. What the reporting does not give you is a timeline for when AI revenue is expected to catch the spend, how much of that $205 billion is contractually locked versus flexible, or the split of the $44.9 billion quarterly capex between Nvidia GPUs and Google's own TPUs.

For anyone selling into the AI buildout, the takeaway is straightforward. The biggest customer just raised its budget, not trimmed it.

Shared on Bluesky by 2 AI experts