Whether it is the front page of the local newspaper or the cover of a national magazine (or just routine conversation), it is difficult to escape the fact that there has been relentless pressure to build more and more data centers, in large part to support the boom in development of Artificial Intelligence (“AI”). The pressure to build, of course, comes primarily from the firms committed to developing AI. The pushback comes from various quarters: people and institutions concerned with environmental impacts of the data centers together with people and institutions (including state and local governments) who fear the long-term consequences of the “AI Revolution.”
In speaking recently on the topic of AI data center development, I have been asked several times to put the tensions in historical perspective. In my more than half a century of private law practice, I have seen these same tensions occur in circumstances where major innovations are being proposed. It is worth recounting some of those circumstances as lessons in how the “for v. against” dynamic played out. The results were far from uniform, which calls for caution in predicting how the data center / AI debate will evolve. While some will react to the historical examples as “out of left field” and “fundamentally different” given the substance and sheer magnitude of the AI issue, it is nonetheless useful to look back on history and assess whether the lessons are instructive.
The first historical example occurred decades ago, in the early 1970s. For a variety of reasons, natural gas was then treated as a scarce commodity in the United States. The dominant domestic energy production was oil, and even so, the country was dependent on foreign sources for much of its oil supply, a fact made painfully obvious by the Arab Oil Embargo of 1973. As the situation that led to the embargo was evolving, the administration determined that increasing the U.S. supply of natural gas was important, even if it meant importing liquefied natural gas (“LNG”) from foreign sources. In 1972, the administration entered into a major trade agreement with what was then the Soviet Union, and among the other elements of that agreement was a commitment to develop two large LNG projects—one for delivery of LNG to the U.S. East Coast, and one to the U.S. West Coast. The East Coast project was named “North Star” and involved 22 LNG ships transiting LNG from the Yamal Peninsula in the Soviet Union (via Murmansk) to Philadelphia, Pennsylvania.
Despite strong industry and government support for North Star, public outcry arose over importing LNG from a country with discriminatory state practices. While the connection between a large LNG project and an abhorrent human rights issue in the source country seemed remote, it was not. As North Star was evolving, Senator “Scoop” Jackson (D-WA) and Congressman Charles Vanik (D-OH) introduced in 1974 what became known as the Jackson-Vanik Amendment, which barred access to Export–Import Bank of the United States financing for projects involving countries violating basic human rights norms. That amendment killed North Star (and the West Coast project as well), creating a setback for the U.S. policy of increased reliance on natural gas.
The second historical example is not unrelated to the first. With the blow to the effort to import LNG, the energy industry redoubled its efforts to produce and deliver natural gas in the United States. That effort was broad based, but it evolved into a situation where a handful of natural gas pipeline companies controlled the delivery of natural gas in the United States. In 1980, a group of natural gas distribution companies in the U.S. Northeast, facing a rapidly increasing demand for clean natural gas, decided to seek natural gas supplies from western Canada. Getting gas from western Canada to the U.S. Northeast was obviously a challenge, but Canada had developed a robust gas pipeline system—the TransCanada Pipeline—that interconnected with U.S. natural gas pipelines in the U.S. Midwest, and which became a feasible means of getting Western Canadian gas to the U.S. Northeast. But the U.S. interconnecting pipelines were the same entities imposing limits on U.S. natural gas deliveries, and those pipelines decided to impose limits on the transportation of Western Canadian gas. In 1984, the same utilities that had turned to Western Canada for natural gas supplies decided (against all odds) to propose a new gas pipeline—the Iroquois Gas Transmission System—connecting the TransCanada system directly to the U.S. Northeast.
The bold decision to build the Iroquois Pipeline met fierce opposition. The pipeline would follow a new “greenfield” route several hundred miles through New York and Connecticut and across Long Island Sound onto Long Island. Major opposition came from the landowners whose properties would be impacted by the pipeline. Major opposition also came from the heating oil industry in New York, which did not want the competition from new sources. And finally, opposition came from the U.S. pipeline industry, not in the form of “don’t build Iroquois,” but in proposals to expand their existing pipeline systems. It took six years to traverse the intense opposition, but Iroquois was ultimately built and started operations in 1991—and has been expanded several times since.
The third historical example is completely unrelated to the prior two. Like most developed countries, the availability of water for human consumption and agriculture is critical to the economy and human well-being. Periods of drought, no matter when or where they occur, are problematic. The United States is no exception, and drought has long been a major concern, especially in the Western United States. Primary water sources, like the Colorado River, have been strained to an alarming degree, with no easy solution. In these circumstances, turning to seawater desalination was one possible approach, and a solution adopted in many parts of the world (especially the Middle East). In 2001, a newly formed company, Poseidon Water, embarked upon several large-scale seawater desalination projects. Poseidon succeeded in launching successful plants in Florida and California but then faced aggressive opposition with several roots. The environmental opposition—grounded in the notion that the intake and outfall systems of desalination plants have material adverse impacts on marine life—took hold in the proceedings for permits to construct and operate the plants. The governmental opposition took hold, not in public proceedings, but in a variety of efforts to discourage private ownership of major sources of potable water, leading to further cancelled projects.
So how do these three historical anecdotes bear on the tension between AI-driven data center development and growing public concerns about AI? There are three parallels which cause me to look at the historical perspective. First, there will always be some opposition to major advances in infrastructure development. That opposition will almost always be based in part on environmental concerns. In the case of data center development, environmental opponents cite persistent noise; excessive energy consumption causing secondary effect on retail energy prices; and excessive water requirements straining indigenous water resources. While not necessarily environmental, another area of public concern is with the aesthetics of data centers (big nondescript buildings in clear public view), among other things. While environmental-based opposition is not always successful (e.g., the Iroquois pipeline), it often is and in all events is a major cause of project delay. It must be taken seriously.
Second, opposition to data center projects need not have a direct connection to the project itself. North Star is an example, where a human rights issue, and not an issue with the project per se, carried the day against what was otherwise a widely supported project. One could argue that the growing opposition to AI relates not to data center projects as infrastructure developments, but to the intended uses of AI.
Third, and perhaps most relevant, is the fact that opposition to specific projects can be unpredictable and therefore hard to manage. The opposition to North Star was not predicted (and some would say was unpredictable, given its origins in a highly public international trade agreement). The opposition to Poseidon’s California plant was predictable, but the shape it took—denial of a crucial governmental permit in the face of strong support from the public and the California governor—was not. In the case of data center development and the AI-based opposition, it is hard to predict where the opposition to AI will go. Will it go toward legislative restrictions on AI at some governmental level(s)? Will environmental objections to large data center development be effective and perhaps cause AI to take a different or slower course? Will a cacophony of opposition chill the financial institutions who provide the capital for the AI boom? Will the pace of development be disrupted by major advances in technology?
The AI data center revolution involves an amazing number of stakeholders, which is why AI data center development has been likened to a symphony orchestra—multiple diverse components which, alone, can do nothing but together under the guidance of a conductor can make beautiful music. At this point in the evolution, there is no conductor, nor is there any sense of who could come forward as a conductor—industry or government? Much remains to be seen, and historical perspectives can be useful.