A dramatic disconnect between public sentiment and resource consumption data is threatening to reshape the timeline for America’s AI infrastructure boom. A new Gallup poll reveals that nearly 70% of Americans oppose the construction of data centers in their local communities, driven by fears over soaring electricity costs, the conversion of agricultural land, and the strain on local water supplies. The backlash, now a nationwide phenomenon, comes as hyperscale technology companies prepare to deploy an estimated $700 billion in capital expenditures this year alone to expand their AI infrastructure.
Yet a closer examination of water usage data tells a strikingly different story. According to an analysis by researcher Andy Masley, cited by tech publication Pirate Wires, even if total water consumption by U.S. data centers triples by 2030, it would still amount to just 8% of the water used by the nation’s golf courses. That means golf courses currently consume more than 12 times the water that data centers do. The agricultural disparity is even more extreme: U.S. almond farming alone uses 80 times more water than every data center in the country combined.
The economic efficiency gap is equally stark. A model shared by X user Smirkley calculates that data centers generate approximately $132.28 in economic output for every five gallons of water consumed, based on $529.1 billion in industry revenue divided by an estimated 20 billion gallons of annual water use. California almond farming, by contrast, produces roughly 1.78 cents per five gallons of water, based on $5.66 billion in output against 1.59 trillion gallons of annual consumption. The difference: a factor of roughly 7,400 times.
The public opposition, however, is already having tangible consequences. Regulatory hurdles and rejected permit applications threaten to delay or derail nearly half of all planned data center projects this year, according to market estimates. This pressure comes at a critical moment for the technology industry, which is simultaneously navigating a wave of white-collar layoffs driven by AI automation. Meta, for instance, recently announced a fresh round of job cuts as it deepens its investment in artificial intelligence.
The $2 Trillion IPO and the AI Arms Race
The tension over physical infrastructure is unfolding against a backdrop of staggering financial ambition in the AI sector. SpaceX, the aerospace company controlled by Elon Musk, publicly filed its S-1 registration statement with the Securities and Exchange Commission on Wednesday, setting the stage for what could be the largest initial public offering in history. The company is targeting a valuation of up to $2 trillion and plans to list on the Nasdaq under the ticker symbol “SPCX.”
The filing reveals a company burning through cash at an extraordinary rate. SpaceX reported revenue of $11.39 billion for 2025, but posted an operating loss of $4.9 billion. In the first quarter alone, the company generated $4.69 billion in revenue against an operating loss of $1.94 billion. Capital expenditures surged to more than $20 billion in 2025, up from $11.2 billion the prior year, as Musk’s company pours money into AI development and its ambitious plan to build orbital data centers.
Musk’s voting power after the IPO will stand at 85.1%, secured through ownership of 12.3% of Class A shares and 93.6% of Class B shares, which carry 10 votes each. Forbes estimates Musk’s net worth at $807.7 billion as of May 20, already making him the world’s richest person by a wide margin. A successful IPO at the targeted valuation could push him past the trillion-dollar threshold.
SpaceX is not alone in its pursuit of public market capital. OpenAI, the company behind ChatGPT, is preparing to confidentially file for an IPO as soon as this week, working with Goldman Sachs and Morgan Stanley, according to a source familiar with the matter. The company is targeting a valuation of roughly $1 trillion and aims to debut sometime in the fall. The move follows a jury verdict earlier this week that rejected Musk’s lawsuit against OpenAI and co-founder Sam Altman, removing a significant legal overhang from the company’s path to going public.
“We regularly evaluate a range of strategic options. Our focus remains on execution,” OpenAI said in a statement. The company recently completed a $122 billion funding round at an $852 billion valuation and has committed to spending more than $1.4 trillion on physical infrastructure in the coming years.
The capital demands of the AI industry were underscored by Nvidia’s latest earnings report. The company posted record revenue of $81.6 billion for the February-April quarter, up 85% year-over-year, with net profit surging 211% to $58.3 billion. The results came despite the company shipping no data center chips to China, highlighting the immense demand from the rest of the world.
Google’s Efficiency Play and China’s Chip Surge
At its annual I/O developer conference this week, Google CEO Sundar Pichai unveiled the Gemini 3.5 Flash model, positioning it as a cost-effective alternative for enterprise customers. Pichai said the model could help top Google Cloud clients “save over $1 billion annually” if they shift 80% of their workloads from other frontier models. Google also disclosed that its Gemini AI application now has 900 million monthly active users, doubling over the past year, with daily request volumes growing more than sevenfold.
The search giant is also deepening its infrastructure commitments. Google is partnering with Blackstone on a joint venture to provide cloud services powered by its in-house TPU AI chips, putting it in direct competition with Nvidia’s dominant position in the data center chip market.
Meanwhile, China’s push to localize critical semiconductor production is yielding dramatic results. ChangXin Memory Technologies (CXMT), the country’s top memory chipmaker, disclosed in an IPO prospectus filed with the Shanghai Stock Exchange that net profit skyrocketed more than 1,688% year-over-year in the January-March period, while revenue surged over 719%. The company is undertaking its most aggressive expansion ever in Shanghai, aiming to multiply production capacity and develop high-bandwidth memory, the critical component used in AI data center servers.
The Orbital Data Center Frontier
SpaceX’s IPO prospectus notably excludes China and Russia from its total addressable market estimation, while warning investors that Beijing represents one of the company’s biggest competitive threats due to heavy state investment in its domestic space industry. The filing marks a sharp contrast with Tesla, another Musk-owned company that built much of its success on the Chinese market.
Musk’s ambitions extend beyond terrestrial infrastructure. SpaceX is developing an AI supercomputer known as “Colossus,” and AI competitor Anthropic recently announced a deal to access the multi-billion-dollar machine. Anthropic has also expressed interest in partnering with SpaceX to build orbital data centers, a concept Musk claims can provide “near-infinite sustainable power while minimizing impact on Earth.” Google is reportedly in discussions with SpaceX about rocket launch contracts as it eyes a potential entry into orbital data centers, which some experts believe could become feasible in the 2030s.
The convergence of these forces—public resistance to ground-based data centers, the staggering capital requirements of AI infrastructure, and the technological push toward orbital alternatives—is reshaping the landscape for investors. The information asymmetry between public perception and actual resource consumption data is not merely an academic curiosity; it directly influences policy decisions, site selection strategies, and construction timelines, injecting a new layer of uncertainty into the already complex calculus of AI infrastructure investment.
As hyperscale companies race to deploy $700 billion in capital this year, the gap between public sentiment and water-usage reality may prove to be one of the most consequential variables determining how quickly—and at what cost—the next generation of AI infrastructure gets built.