Version 4

AI infrastructure spending is reshaping capital deployment well beyond the technology sector. Google projected $205 billion in 2026 capex [2] and reported its first-ever negative free cash flow quarter in Q2 2026 [1], with Alphabet's total forward spending commitments reaching $811 billion [3]. OpenAI is building a $20 billion data center campus in Georgia with a 3.2-gigawatt power agreement [9]. Verizon signed a $1 billion-plus dark fiber deal with Google and is converting legacy central offices into edge data centers for AI inference [19]. The Trump EPA is advancing proposals to allow construction before air permits are issued and give states authority to limit public participation in air pollution reviews [11][13].

The build-out is now pulling in telecom infrastructure — dark fiber routes and repurposed central offices — alongside the cloud and power investments already in motion. The EPA permitting changes, if finalized, would compress the community review window precisely as construction accelerates across multiple states.

The scale of AI infrastructure spending has reached a point where even the most profitable technology companies cannot fund it from operating cash flows. Google reported record Q2 2026 revenue of $119.8 billion and Cloud revenue of $24.8 billion — up 23.8% from Q1 — yet capital expenditures exceeded all of it, producing the company's first-ever negative free cash flow quarter [1]. Google's full-year 2026 capex is projected at $205 billion [2], and Alphabet's total forward spending commitments reached $811 billion [3]. Evercore analysts warned in February 2026 that hyperscaler capex could exceed operating cash generation before revenue caught up [4]; Semianalysis CEO Dylan Patel predicted in April 2026 that Google's free cash flow would hit zero [5][6]. Both forecasts proved accurate. Across the industry, hyperscalers are expected to commit at least $625 billion to AI infrastructure in 2026 [7], and IDC projects annual spending will exceed $1 trillion by 2029 [8].

On July 22, 2026, OpenAI disclosed Project Camellia, a $20 billion data center campus planned for Effingham County, Georgia, near Savannah [9]. The project has secured a 3.2-gigawatt power agreement with Georgia Power, to be delivered in phases between 2028 and 2032 [9][10]. OpenAI pledged $80 million in community benefits directed toward schools, public safety, healthcare, and workforce training, and cited Georgia Public Service Commission rules as prohibiting the pass-through of infrastructure costs to existing residential ratepayers [9]. An independent annual public audit is promised to verify these commitments [9]. All disclosed details precede groundbreaking and remain untested against construction or operational realities.

The Trump EPA is advancing regulatory changes that would alter how data center and power plant development intersects with air pollution permitting. One proposal would give states authority to determine the extent of public participation in air pollution permitting reviews, reducing or eliminating public notice requirements for facilities such as gas plants and diesel generators that support data center operations [11][12]. A companion proposal reported by E&E News would allow construction to begin before air permits are issued [13][14]. A law firm analysis frames the agency as stepping back from federal data center regulation and handing authority to states [15], while the Environmental Data and Governance Initiative has documented that communities near EPA-regulated data centers already face elevated air pollution [16]. The Environmental Defense Fund and Sierra Club, joined by other health and environmental groups, formally opposed the pre-permit construction proposal, arguing it removes the practical window in which community and regulatory review can affect a project's design or siting [17][18].

The build-out is beginning to pull in telecom infrastructure. Verizon signed a deal worth more than $1 billion with Google to connect data centers using its existing dark fiber network [19]. The company is also stripping copper wiring from legacy central offices and repurposing them as small edge data centers optimized for AI inference workloads, targeting use cases such as robotics, remote surgery, and autonomous driving that require ultra-low latency [19]. Verizon's CEO expects additional AI-related deals worth multiple billions of dollars to be announced by year end [19]. The arrangement shows how AI capital deployment is extending into sectors with existing physical infrastructure — fiber routes and urban real estate — that can be repositioned without greenfield construction.