What’s Behind Google’s Rising Capex Hike to US$205bn?
Alphabet’s latest financial report delivered glittering figures for Google Cloud, driven by AI-hungry enterprises that pushed its second-quarter revenue up 82% to US$24.8bn.
The average expectation among analysts for the quarter was just a 64% increase, according to data compiled by LSEG.
The consensus estimate of US$116.9bn for the quarter’s total revenue was also overshadowed by the actual figure that came around to US$119.8bn.
On the other hand, the advertising revenue at US$81.6bn remained closer to its estimate of US$81.1bn.
The cost of supporting the AI boom brings in a much bigger number that Anat Ashkenazi, the company’s Chief Financial Officer, expects to be between US$195bn and US$205bn in capital expenditures (capex).
This forecasted hike in the 2026 capex caused much anxiety in Wall Street, dipping Alphabet’s shares by about 3% in extended trading.
In the last quarter, the projection stood between US$180bn and US$190bn, which was already a US$5bn bump from what Anat had announced the quarter before.
- Google Cloud revenue grew by 24% quarter-over-quarter, rising from US$20bn in Q1 to US$24.8bn in Q2
- Alphabet’s US$44.9bn quarterly outlay exceeded operating cash flow, creating a US$5.9bn negative free cash flow.
Sinking into the negatives
For the first time ever, Alphabet reported negative free cash flow, burning US$5.9bn as the spending hit US$44.9bn in a single quarter.
Thomas Monteiro, Senior Analyst at Investing.com, spoke to Reuters: “After a negative cash flow quarter, the new raise in capex does not sit well for Alphabet. The market’s most reliable cash generators are now spending more than they bring in.
“As long as revenue keeps accelerating, investors will tolerate it. But capital has a real cost again and the room for error is shrinking every quarter.”
The earnings call also highlighted a strategic challenge inside Google’s core AI division: model uncertainty and trailing competitors in frontier development.
Much concern now remains in the street of bulls and bears regarding Google’s consumer and enterprise AI models.
A costly delay
Alphabet’s AI momentum slowed earlier this year after it delayed the scheduled launch of its flagship model, Gemini 3.5 Pro.
The shortfall left Google trailing in the high-stakes market for AI coding and developer tools, where rivals OpenAI and Anthropic aggressively captured the sector with their respective model releases.
During the earnings call, analysts pressed CEO Sundar Pichai on whether Google could maintain its position at the frontier of AI development.
He said: “There are many attributes on which we are still at the frontier. There are areas where we’ve acknowledged we need to improve; coding and agentic coding is an example of that.”
The top executive then added that instead of focusing all internal supercomputing power on pushing out Gemini 3.5 Pro, Google continued testing.
It had also started training Gemini 4 and was ‘applying a lot of our compute and effort in that direction' to remain competitive.
By pouring resources into next-generation frontier training, Google is strategically reallocating its compute to jump ahead with a generational leap instead of fighting a defensive battle with a delayed update.
However, skipping or delaying a model step comes with its own risks as developer habits are often considered to be sticky in the software ecosystem.
For Google to stay ahead in the AI race, Gemini 4 needs to arrive before coders and enterprises standardise their workflows around rival models that came in early.
