Hall of Fame baseball player and manager Yogi Berra had a remarkable gift for expressing profound truths with simple words. One of his most famous observations was, “If you don’t know where you’re going, you might not get there.” Although intended humorously, his advice may be more relevant today than ever before as America races into the age of artificial intelligence.

The United States is investing hundreds of billions of dollars in AI. Technology companies are constructing enormous data centers, electric utilities are expanding generating capacity and communities across the country are competing to attract these projects. AI is widely promoted as the next Industrial Revolution, capable of transforming nearly every aspect of our economy.

But amid all the excitement, one fundamental question is receiving surprisingly little attention: Where are we going with AI — and what are we giving up to get there?

Before answering that question, we should recognize that artificial intelligence is not one technology, but two.

The first is generative AI. Systems such as ChatGPT, Claude, Gemini and Grammarly assist people by generating text, images, software code, music and designs. They increase productivity while leaving the final decisions to humans. Today, generative AI is proving valuable in engineering, medicine, education, business, and the creative arts.

The second is agent AI. Unlike generative AI, agent AI purports to be able to do work on behalf of people without human help. It hopes to manage schedules, execute financial transactions, monitor supply chains and eventually replace many routine jobs. Because Agent AI performs continuous searches and decision-making, it requires substantially more computing power — and therefore much more electricity — than generative AI.

That distinction matters because infrastructure, not technology, may become AI’s greatest limitation.

Generative AI has already demonstrated enormous value. In filmmaking, directors can preview scenes before spending millions of dollars on location shooting. Actors can be made younger or older, weather conditions altered, lighting adjusted and camera angles tested in minutes instead of weeks. Steven Spielberg summarized its value well when he observed that AI can be an extraordinary filmmaking tool, provided the final creative decision always remains with the filmmaker.

Agent AI presents a different challenge. While promising remarkable gains in efficiency, it also raises important questions about employment, energy consumption and whether the enormous resources required to support it might be invested more wisely elsewhere.

Jennifer Harris recently argued in The New York Times that America’s investment community has become consumed with AI while neglecting equally important priorities, including affordable housing, manufacturing, transportation infrastructure and modernization of the electric grid. Venture capital is flowing overwhelmingly toward AI while sectors critical to America’s long-term economic health struggle to attract investment.

History gives us reason for caution.

Railroads transformed America but also created speculative bubbles. The Internet revolution permanently changed commerce and communication, yet the dot-com collapse showed that revolutionary technology does not automatically justify unlimited investment. AI may prove even more transformative, but that alone does not guarantee that every proposed project is a wise investment.

The scale of today’s expansion is extraordinary.

Although there is no official federal count of AI data centers, estimates suggest the U.S. has between 4,000 and 5,000 data centers, including hundreds of hyperscale facilities and a rapidly growing number of AI campuses. More than 1,000 additional projects have been proposed. Whether they can all be built remains uncertain.

The greatest obstacle is not financing. It is infrastructure.

Electricity has become the largest bottleneck. Texas alone is evaluating requests totaling more than 438,000 megawatts of additional electrical demand, much of it associated with AI facilities. By 2030, U.S. data centers may need tens of gigawatts of additional power and could use hundreds of terawatt-hours of electricity per year. Some estimates put U.S. data-center electricity use at 6.7% to 12% of U.S. power consumption by 2030.

Every campus requires new generation, substations, transformers, transmission lines, fiber-optic networks, backup power and years of engineering before construction can begin.

Water presents another challenge. Depending on cooling technology, AI facilities may consume billions of gallons annually. Recent research estimates AI servers could increase U.S. water consumption by 200 billion to 300 billion gallons annually by 2030, placing additional stress on regions already struggling with drought.

Land requirements are equally significant. Modern AI campuses may occupy hundreds — or even thousands — of acres to accommodate buildings, substations, cooling systems, battery storage and future expansion.

The government is struggling to keep pace. While the federal government seeks to accelerate permitting, many state and local governments are increasing oversight because of concerns about electricity demand, water supplies, zoning, electric ratepayers and taxation, and community impacts.

These competing priorities illustrate a much larger national issue.

Every dollar invested in AI is a dollar unavailable for another national priority.

Every electrical substation built for a data center cannot simultaneously support affordable housing, manufacturing expansion, transportation electrification or modernization of an aging electric grid.

Every engineer designing AI infrastructure is not rebuilding flood-control systems, increasing our housing inventory, strengthening bridges, modernizing drinking water facilities or helping communities adapt to increasingly destructive extreme weather.

This is not an argument against artificial intelligence. It is an argument for balance.

America has always succeeded because it invested simultaneously in innovation and public infrastructure. We built interstate highways while expanding aviation. We developed nuclear power while modernizing our electrical grid. We landed astronauts on the moon while investing in universities, hospitals and research laboratories.

Artificial intelligence should become another chapter in that history — not the only chapter.

As an engineer, I believe technology should first solve society’s greatest problems. AI can certainly help us achieve that goal. But if our pursuit of AI diverts the financial, electrical, water, land and human resources needed to address housing shortages, aging infrastructure, medical care, manufacturing and climate resilience, we may find we have solved one problem while creating several others.

Yogi Berra’s simple wisdom still applies today.

Before America commits another trillion dollars to artificial intelligence, perhaps we should answer this question: Where exactly are we going — and is that where we truly want to be?

Dr. Michael Sills was the Chief Engineer of the New Hampshire Environmental Agency NHDES-WMD for nearly 30 years.