Alphabet burns cash for the first time since 2004 to build AI compute
Alphabet reported negative free cash flow of $5.9 billion for Q2 2026 — its first cash-negative quarter since the 2004 IPO — as capital expenditure doubled year-over-year to a record $44.9 billion. Operations generated $39.1 billion, less than what the company spent. CFO Anat Ashkenazi raised full-year 2026 capex guidance to $195–$205 billion. Trailing 12-month free cash flow stayed positive at $53.3 billion, and long-term debt reached $98.2 billion, up from $46.5 billion at the end of 2025.
About 60% of the quarter's infrastructure spending went into servers, with the rest split across data centers and networking. Google delivered its TPU systems to customers' data centers for the first time in Q2, though the CFO expects most of that revenue to land in 2027. Google Cloud grew 82% to $24.8 billion with a backlog of $514 billion. Anthropic's October 2025 agreement gives it access to up to one million TPUs and more than 1 GW of capacity.
The abundance angle is compute. This spending is the physical substrate of machine intelligence — chips, buildings, power, and networking. As it scales, and as custom silicon like the Ironwood TPU spreads, the cost per unit of useful AI work falls, making capable models cheaper to run and more widely available over time. It also commits enormous energy demand.
Caveats: the source is a member-exclusive analysis piece, and Q2 net income of $112.1 billion was dominated by $99.0 billion in largely unrealized equity gains rather than operations. The article text is cut off at the end.
Source: Tom's Hardware
MANY MINDED