AI Infrastructure Spending Reaches Unprecedented Scale at Google and OpenAI

Major AI companies are committing to infrastructure at a scale that is now straining their own finances. Google reported its first-ever negative free cash flow quarter in Q2 2026, with full-year capex projected at $205 billion, despite record revenues and 40% year-over-year operating cash flow growth [1][2]. Simultaneously, OpenAI announced Project Camellia — a $20 billion data center campus in Effingham County, Georgia, backed by a 3.2-gigawatt power agreement with Georgia Power [6]. Industry-wide, hyperscalers are projected to spend at least $625 billion on AI infrastructure in 2026 [13], with IDC forecasting annual spending above $1 trillion by 2029 [12].

For the first time, AI infrastructure spending is outrunning the cash generation of even the most profitable technology companies. Google's negative free cash flow — achieved despite record revenues — shows how quickly capital commitments can overwhelm near-term financial metrics. Whether Cloud revenue growth will eventually close that gap, or whether the build-out is running ahead of real enterprise demand, is the core question for the sector's financial health.

Google's Q2 2026 earnings revealed a first in the company's history: negative free cash flow, driven entirely by AI infrastructure spending that now outpaces even its substantial operating cash generation [1]. Total Q2 revenue reached $119.8 billion, beating analyst expectations, and operating cash flow grew approximately 40% year-over-year to around $39.1 billion. Google Cloud alone generated $24.8 billion — up 23.8% from Q1 2026 — confirming that demand for AI services is real and expanding [1]. Yet capital expenditures exceeded all of it. Google's full-year 2026 capex is projected at $205 billion [2], a trajectory that Evercore analysts had flagged as a potential 'red flag' as early as February 2026 [3], and that Semianalysis CEO Dylan Patel had predicted would push Google's free cash flow to zero by April of that year [4][5]. Both forecasts proved accurate sooner than expected.

On the same day Google's results became public, OpenAI disclosed Project Camellia: a $20 billion data center campus planned for Effingham County, Georgia, near Savannah [6][7][8]. The project has secured a 3.2-gigawatt power agreement with Georgia Power Company, to be delivered in phases between 2028 and 2032 [6][9]. OpenAI announced $80 million in community benefits directed toward schools, public safety, healthcare, and workforce training, plus up to $71 million in Codex credits for eligible Georgia college and technical school students [6]. The company committed to closed-loop water cooling and cited Georgia Public Service Commission rules as prohibiting the pass-through of infrastructure costs to existing residential ratepayers [6][10]. An independent annual public audit is promised to verify adherence to these commitments [6].

The community-first framing appears calibrated to defuse the backlash that has accompanied other large data center projects — concerns about electricity costs, water use, and traffic [11]. All disclosed details are pre-groundbreaking, and the commitments have not yet been tested against construction and operational realities.

Zooming out, the spending is not limited to Google and OpenAI. IDC estimates that AI infrastructure spending in Q4 2025 alone reached approximately $90 billion, and projects annual spending will exceed $1 trillion by 2029 [12]. Hyperscalers as a group are expected to commit at least $625 billion to AI infrastructure in 2026 [13]. Third-party operators are also scaling: Vantage Data Centers announced 'Frontier,' a $25 billion AI campus in Texas [14]. Goldman Sachs has examined the assumptions behind these projections, focusing on whether enterprise AI adoption will grow into the capacity being created [15].