Where AI Builds: How Algorithms and AI Are Rewriting the Map

It’s no surprise that algorithms have shaped our world. As Kevin Slavin argues, physical creations—like stock trading floors, buildings, and entire city infrastructures—are now designed around algorithmic needs rather than human use. Think fiber optic cables buried under miles of land to shave off microseconds in trading. More than their physical influences, they essentially shape our reality: From some of the most impactful things we have to contend with, like credit applications to more mundane, like Netflix scripts, algorithms are no longer just helping us; they’re actively remaking our environments, culture, and daily routines. Further sill, algorithms optimize with objectives no one fully controls or understands. It’s not just AI, most of our current systems are complex black boxes. Once in place, they mainly operate outside human comprehension, often with impacts we can’t predict or reverse. For example, in finance, the fastest algorithm wins – even if it’s wrong or opaque. I read an interesting paper here on how LLM’s aren’t transparent even when they are “transparent” through chain of thought (COT) which frames it in a way I hadn’t really given too much thought to up until now.

Essentially, algorithms have become agentic in a way. In any case, they have become more than calculators. They reshape markets, cities, entertainment, and even surveillance. The place where we are seeing (as in where it’s most visible, but potentially not where the most action is taking place) is with electricity prices and demand. Even though overall inflation has cooled (depending on what you read) electricity prices continue to rise rapidly. In the U.S, they rose 4.5% over the last 12 months (according to the consumer price index), which is nearly double the inflation rate for all goods and services. Retail electricity prices are set to outplace inflation through 2026. There are lots of isolated and intersecting reasons for this (some beyond my comprehension), but essentially and at a 30,000ft level, the growth in electricity demand coupled with the deactivation of power-generating facilities are far outpacing the rate at which new electricity generation is coming online/being added to the grid).

Prices are regional

Unlike, say, oil prices, the price of electricity is regional and costs can vary widely based on where the consumer lives (and their consumption of course). The average U.S. household paid about 17 cents per kilowatt-hour of electricity in March 2025. That being said, using averages cannot give much real insight. The range in terms of cost is huge, with a low of about 11 cents per kWh in North Dakota to a high of about 41 cents per kWh in Hawaii, according to EIA data. However, certain electricity trends are happening nationwide, not just regionally.

Enter The Data Center

You may recall there was a House energy hearing in March (2025). According to Jennifer Curran, senior vice president of planning and operations at Midcontinent Independent System Operator, who testified, electricity demand growth was “minimal” in recent decades due to increases in energy efficiency. Meanwhile, U.S. “electrification” swelled via use of electronic devices, smart-home products and electric vehicles. Now, demand is poised to surge in coming years, and data centers are a, if not thee, major contributor. In 2023 alone, Data center electricity use tripled to 176 Terawatt-hours in the decade, and use is projected to double or triple by 2028. This means, data centers are expected to consume about 14% of total U.S. electricity by 2028. Nothing else compares. The U.S. economy is set to consume more electricity in 2030 for processing data than for manufacturing all energy-intensive goods combined, including aluminum, steel, cement and chemicals, and if you’ve followed me for any amount of time (or read Vaclav Smil’s How the World Really Works, you will know what an incredible and gargantuan amount that is (aluminum, steel, cement are what the entire world is built on; and chemicals are what make feeding us possible).

All about infrastructure

Data centers are increasingly showing up in areas with abundant and affordable power sources, driven by the need for reliable energy to support their growing operations, particularly with the rise of AI and other high-demand technologies. This shift is evident as traditional hubs like Northern Virginia face capacity and power limitations, prompting operators to explore new regions with better energy availability.

• Northern Virginia: Once a major hub for data centers, it has pretty much reached capacity, with 245 centers covering 25,000,000 square feet and consuming 3.6 GW of power.
• California: Similarly, the state has seen a saturation of data centers, with 208 centers owned by 63 operators, and 53 in Santa Clara alone.
• Columbus, Ohio: Emerging as a secondary market, it offers available land and power, attracting data center operators looking for growth opportunities.
• Austin, Texas: A growing location for data centers, with operators seeking access to power and talent.
• Arizona: Data centers are expanding in Arizona, particularly in areas like Phoenix, where there is ample land and hydroelectric power available.
• Pennsylvania: The state is seeing significant investment in power infrastructure, with projects like the repowering of the Bruce Mansfield coal plant into a natural gas power station to support data centers.
• Texas: Known for its abundant energy resources, Texas is a key location for data centers, with projects like the one in Portland, Oregon, using fuel cells and geothermal energy.
• If you’ve followed us for a while, you’ll know that subsea data centers reduce the power consumption by 40% (meaning all power used is IT power) and we can sit in water anywhere.

Just as Kevin Slavin warned that algorithms have escaped the screen and begun reshaping the physical world—from stock market fluctuations to architectural design—AI is now pushing that influence even further into our landscapes. The rise of large-scale AI models has created unprecedented demand for computational resources, forcing data centers to migrate from traditional urban tech hubs to (sometimes) remote areas with abundant energy. Just as it was with algorithms, this shift isn’t purely digital—it is geographic, architectural, and environmental.

Data centers, once hidden in plain sight near corporate campuses or suburban zones, are now appearing in places chosen not for human access, but for energy access, including underwater in ports, rivers, dams and further offshore (see subseacloud.com for more). AI’s insatiable power hunger—often requiring tens to hundreds of megawatts per site—means that infrastructure is being reoriented around the physics of electricity rather than the economics of real estate (which is odd to think about because, reductively speaking, data centers have always been a real-estate play). Rivers, dams, geothermal vents, and even stranded wind projects are becoming the new centers of AI development, dictating where digital economies bloom, which in turn helps dictate who benefits most.

To sum up, this physical readjustment has profound implications. It redefines what constitutes a “tech hub”, challenges national and regional grids, and introduces new geopolitical and environmental stakes. In short, AI is not only changing how we think and decide, but where we build and live. The virtual continues to be a huge driver of the physical.

Future-Proofing

In the future, I think it's likely that the physical footprint of AI will be as consequential—and visible—as highways, ports, or power plants were in previous industrial eras. We will likely see entire regions reshaped around the logistical needs of AI: abundant renewable energy, cooling water sources, low-latency fiber routes, and political stability. These new AI zones will, in someways hark back or resemble modern factory towns, except instead of human labor, they’ll be filled with racks of machines performing trillions of operations per second, around the clock.. hopefully for us.

Countries with surplus green energy—like Iceland, parts of Canada, or whoever has gone nuclear—should become digital resource exporters, not by shipping goods, but by hosting models and processing global workloads. Conversely, countries without stable or clean power may fall behind in their ability to harness or regulate AI. In this way, energy becomes destiny again. It drives home what Vaclva Smil says, which is that energy is the only universal currency. Or, as Napoleon said, geography is destiny.

On a more granular level, data centers may get smaller and more distributed—modular pods deployed offshore, underground, or even on decommissioned oil rigs—chasing not just power but climate resilience and regulatory arbitrage. Some will be fully autonomous, self-maintaining with robotics and AI-powered optimization. Others may be embedded into national infrastructure, like railway systems or desalination plants, creating synergistic energy-sharing ecosystems.

The boundary between the digital and physical will blur further. Where power flows, intelligence will follow. And in turn, AI may begin shaping not just where we build—but how. Future cities, ports, and even rural communities may be co-designed with their digital twin first, optimized by AI, and constructed to support both human and machine needs simultaneously. The data center won’t be a back-end warehouse anymore—it will be the beating heart of our built environment.