We don't need him, your intelligence - ForkLog

# We don't need him, your intelligence

Construction of nearly half of the data centers in the United States has been put on hold

The artificial intelligence industry has hit a barrier that cannot be overcome with money or the release of new chip generations. A global shortage of power grids, a lack of components, and resistance from local communities have turned the construction of data centers (DCs) into one of the most difficult logistical and political challenges for the tech sector.

What is fundamentally different about AI data-center deployments from traditional data centers, why the industry has become even more dependent on China, and how people are overthrowing local authorities in attempts to protect themselves from noise and an environmental catastrophe—this is covered in a new ForkLog piece.

Specifics of the architecture

Traditional data centers that served the internet economy over the past twenty years differ fundamentally from the architecture required to run large language models.

A traditional data center is CPU-focused and consumes an average of 5 to 10 kW of electricity per server rack, whereas AI tasks on GPUs require roughly 10 times more power. Machine-learning racks with accelerators like the Nvidia H100 or B200 require 40 to 120 kW per unit. The difference in power density affects the basic physics of these facilities.

A cluster of tens of thousands of GPUs at peak load consumes electricity on the scale of a small industrial city. The problem is that distribution networks and substations are usually not designed to handle such consumption spikes in isolated areas.

The appetites of leaders in the AI industry have depleted inventories of critical components for bringing in power: high-voltage transformers, generators, and batteries for uninterruptible power supply systems. Manufacturing capacity in the United States and Europe cannot keep up with order volumes. As a result, the queue for industrial transformers—mostly from China—has grown from one or two to three to five years

Growing dependence of the U.S. market on Chinese components amid the opposite political course. Source: Bloomberg. Another problem is heat dissipation. Air cooling cannot cope with the density of AI servers. The industry is forced to shift to liquid cooling systems such as Direct-to-Chip or immersion baths. They require millions of liters of purified water and pose a threat to regions—especially in arid climate zones.

According to figures provided by Dutch specialist Alex de Vries-Gao, in 2025 AI systems worldwide consumed about 765 billion liters of water. To save natural resources, developers are refining the mechanisms. Instead of traditional cooling towers, where water evaporates into the atmosphere, new data centers increasingly come with closed-loop systems. In them, water circulates through pipes, absorbs heat, cools in radiators, and returns to the servers with virtually no volume loss. However, the pace of adoption of this technology lags far behind the speed at which new data centers are being built.

Almost like Stargates

An unprecedented budget, support from top state officials, and the status of the main AI alliance of the decade. At the start, the Stargate project had everything—at least in words

Stargate is an ambitious initiative worth $500 billion, announced by U.S. President Donald Trump in January 2025 as part of a national campaign to maintain technological dominance. A joint venture of OpenAI, SoftBank, and Oracle was supposed to become the main driver of expanding AI data-center infrastructure.

One year after the high-profile announcement, the joint venture still had not assembled a full team and had not signed a single major construction deal on its own behalf.

The situation worsened in financial markets as well. JPMorgan Chase, which was expected to arrange the placement of Stargate’s $38 billion debt obligations, faced investors’ doubts about the project’s profitability.

OpenAI CEO Sam Altman and SoftBank founder Masayoshi Son broke ranks on fundamental issues: exactly where to build facilities and who would control them. From September to October 2025, Stargate’s top executives repeatedly flew to Tokyo for tough negotiations with Son, but still could not decide who would own the platform for the flagship campus in Abilene, Texas.

The flagship site in Abilene with eight plots and planned capacity of 2.1 GW. Source: Bloomberg. OpenAI considered implementing the project on its own, but the lenders refused to provide funds to the company without a clear plan to reach profitability.

Stargate gave up on its ambitious targets and freed up the site for 900 MW, while keeping baseline capacity in Texas aimed at 1.2 GW in the future. The partners shifted their expansion focus, and in April 2026 developer Related Digital, together with Oracle, raised $16 billion in debt and equity financing to build a new mega data center in Michigan for OpenAI’s needs.

Meanwhile, competitors took advantage of Stargate’s uncertainty. At the end of March 2026, the vacant 900 MW was taken over by Microsoft, becoming a new partner of Crusoe Energy to expand the Abilene campus. The modernization will increase the total capacity of this site to 2.1 GW by mid-2027 using Nvidia GPUs.

People vs.

Data centers are no longer seen as an unambiguous driver of the economy—they create at least some jobs only after construction is finished, but at the same time they burden networks, consume water, and generate constant noise.

In April 2026, residents of Festus, Missouri, staged a protest against the construction of a $6 billion data center. Local residents managed to remove four of the eight members of the city council, and also launched a petition to fire the remaining officials, including the mayor.

On April 9, residents filed a lawsuit against the city, claiming that the authorities of Festus did not provide the public with enough time to review the proposal before making a decision and made illegal zoning-change decisions for this project. The lawsuit also states that the city held private meetings regarding the project instead of public ones.

The approved project for an unnamed developer is set to occupy 360 acres of land.

Over the past several months, the United States has seen a string of similar events:

• In February 2026, the New Brunswick city council (New Jersey), under public pressure, rejected a data center construction deal and instead decided to use 27,000 square feet of property to build a public park;
• In the same month, a proposal to annex land into the city limits of Foristell (Missouri) met resistance from citizens due to fears that the land would be used to build a data center. In the end, the decision was changed and the area retained its status as agricultural land;
• In September 2025, Prince George’s County (Maryland) paused data center creation projects after local residents’ protests and formed a task force to study the risks;
• In St. Charles (Missouri), which is less than an hour’s drive from Festus, efforts are underway to permanently ban data center construction after a moratorium was introduced in August 2025.

A data center in New Brunswick was canceled tonight when hundreds of residents showed up. When fight big tech and private equity we win. pic.twitter.com/doZ63Pdwue

— Ben Dziobek (@BenDziobek) February 19, 2026

The growing trend of fighting the construction of computing farms has increased demand for transparency and access to real-time data. The team behind the “U.S. Data Center Moratorium Tracker” is looking for the companies behind the unnamed deal participants and is tracking all locations where authorities have officially imposed temporary bans on building new data centers.

According to the dashboard, as of April 14, 2026 there are 58 moratoriums in effect across the United States.

Source: “U.S. Data Center Moratorium Tracker”.## Well then, into space

Energy shortages, component delays, and public protests have led to stagnation in the industry.

According to Bloomberg, construction of about half of all planned data centers in the United States has either been postponed indefinitely or canceled entirely. Of the projected capacity, less than a third is in the active construction phase.

Expected data center delivery timeline as of March 9, 2026. Source: Bloomberg. Tech giants and independent developers are seeking alternative ways to deploy computing capacity—from the ocean floor to space.

Between 2014 and 2024, Microsoft investigated immersing hermetic server capsules underwater. The last major test of the Project Natick initiative was carried out off the coast of the Orkney Islands (Scotland) from 2018 to 2020. A capsule containing two racks of 864 servers was lowered to a depth of about 35 meters.

Over two years underwater, only six computing units failed. For comparison, in the control group on land, eight times more equipment broke. This was attributed to the inert nitrogen inside the capsule, the lack of temperature fluctuations, and the human factor—often a cause of failures.

Server capsule off the coast of the Orkney Islands (Scotland). Source: Microsoft. According to researchers, the ocean provided free and infinite heat dissipation, and contrary to fears, the data center “did not harm the ecosystem.” Moreover, an artificial reef formed around the capsule, attracting fish.

Despite the success, the project was shut down due to its inapplicability to the AI direction and logistical problems. For any physical intervention, ships must be repositioned, the multi-ton capsule must be lifted from the seabed, and then resealed.

So how will they handle this in space? In late 2025, experts from the 33FG research group estimated that by 2030, AI computing in orbit would be cheaper than on Earth.

In February, SpaceX filed a request with the U.S. Federal Communications Commission seeking permission to place a group of 1 million satellites into orbit for data centers. The project involves creating a network of data centers connected by laser channels

The logic of space data centers rests on two factors: access to 24/7 solar energy and low temperatures for ideal natural cooling.

However, the concept faces strict commercial and physical barriers. SpaceX leadership warned of the risks of such projects being unprofitable at the current stage.

Main difficulties of the concept:

• Launch costs. Despite the drop in price per kilogram of payload, sending heavy server racks with tungsten shielding against cosmic radiation remains extremely expensive;
• Signal latency. For inference (running a ready neural network in real time with users), milliseconds matter. Transmitting massive amounts of data from Earth to orbit and back introduces delays, making the system unsuitable for a number of tasks. Such data centers may only be suitable for asynchronous model training;
• Maintenance. Replacing a failed GPU in space is impossible. The service life of equipment is strictly limited by its radiation resistance.

Other projects are also actively participating in the space initiative: Google said it aims to create a satellite system in low Earth orbit to obtain solar energy and power data centers, and Nvidia announced plans to develop a computing platform for space data centers.

In 2026, California startup Aetherflux plans to launch solar mini-farms in the form of satellites to transmit energy from space to Earth using lasers

On April 27, 2026, Meta agreed to supply 1 GW from space to its data centers with another startup. According to Overview Energy, the developer of an extraterrestrial power plant, the first orbital system demonstration is expected in 2028, with commercial deliveries starting in 2030.

The development of AI infrastructure has run into physical and administrative constraints. The high energy consumption of new GPU clusters, the need for water resources for cooling, and the load on local power grids have led to a reassessment of data centers by the public and municipalities. As a result, scaling up terrestrial computing power has shifted from an issue of affordable capital to a complex logistical and social problem.

Initiatives to create orbital data centers, despite their current high cost and maintenance barriers, are becoming a pragmatic response to the terrestrial infrastructure crisis. In the coming years, companies’ ability to solve the problem of physically placing equipment will determine the pace of further development of computing systems.