Getty Images

Today’s power grids aren’t designed for AI workloads—winners in the AI age will be those who secure affordable, reliable juice.

The big question facing the digital economy is no longer model architecture or GPU supply. It is electricity. Artificial intelligence, EVs, robots and drones are converging at the same bottleneck: an undersized, 20th century power system that was never designed for AI-era loads.

Consider the basic arithmetic. A single generative-AI query consumes roughly 10x the electricity of a standard web search. Replace millions of searches with AI copilots running all day in enterprises. Now add training clusters measured in hundreds of megawatt hours of electricity, and the curve bends fast.

Left alone, expect political disaster: electricity price wars that pit rich tech companies against households and small businesses. That is a recipe for civil disorder. If governments choose to ration supply, another disaster awaits. That country will lose its tech base and watch the AI revolution from the sidelines. The only durable answer is to expand supply and modernize grids at speed.

At the recent Forbes Global CEO Conference in Jakarta, AI infrastructure investor Kuok Meng Wei said global AI data center needs could reach 250 gigawatts by 2030—up from an estimated 50 gigawatts of global capacity today. “About a 1GW ‘takedown’ per week,” he explained. This is equal to one large nuclear power plant. Per week. The gap is not incremental; it is a chasm.

MORE FOR YOU

The panel, which included Alan Chan, managing partner of Full Vision Capital; Kuok, CEO and managing director of K2 Strategic; Otto Toto Sugiri, president director of DCI Indonesia; and Hendra Soetjipto Tan, CEO of Barito Renewables Energy, covered three major points:

First, electricity demand forecasts are too conservative. Most predictions assume electricity must grow at twice the rate of global growth. That’s not nearly enough. Today’s demand models focus on training loads but barely touch inference—the steady-state usage once billions of people run AI assistants continuously. Until storage and grids catch up, operators will need spare capacity simply to keep data centers within tight power-quality tolerance limits. The result: structurally higher demand than policymakers expect.

Second, the bottleneck is not just generation. It’s also transmission. Data centers are vulnerable to voltage surges and flickers; they need stable, high-quality power. Spain’s recent weather-related solar outages are cautionary tales. The solutions are no mystery: substations, high-voltage lines, interconnects and control systems. But upgrades will cost plenty. Kuok described spending tens of millions on a single substation to unlock just 70 megawatts of capacity. Here, China’s multiyear push into ultra-high-voltage transmission is a strategic benchmark, he added.

Third, energy strategy is going local. Nations want more than energy abundance. They also want energy security and data sovereignty. Indonesia will lean harder into geothermal and solar; the U.S. will tap gas and wind; others will mix hydro, nuclear or imports. There is no single technology savior in a world ravenously hungry for electricity. The energy winners will be the jurisdictions that can permit, connect and deliver faster than their peers.

Will this very pressure to get it right and fast lead to a data-center bubble? Yes—and no. Overbuild risk exists in every boom, as it did in telecoms and fiber three decades ago. But the structural drivers—AI across every function; electrified vehicles and logistics; factory automation; “AI at the edge” in robotics and devices—are not fads. They constitute the largest industrial revolution in history. Even if there is a capex shakeout, the load curve keeps rising. Think of temporary gluts as buying the future at a discount.

The headline risk in AI is often framed as safety, bias or employment. Those matter. But the binding constraint for the next decade is simpler: power. Intelligence is becoming the largest single consumer of electricity. Leaders who secure affordable, reliable juice—and the grid that delivers it—will capture the upside of AI. Those who bury their heads may discover that the scarcest input in the digital economy is not talent or capital. It’s the watt.

ForbesThe World PivotsBy Rich KarlgaardForbesTurning PointsBy Rich KarlgaardForbesDecline By DegreesBy Rich KarlgaardForbesVenture GapitalBy Rich KarlgaardForbesThe Age Of Digital DivinationBy Rich KarlgaardForbesAI Gold RushBy Rich KarlgaardForbesAI Wealth DisruptionBy Rich KarlgaardForbesProven Growth FormulaBy Rich KarlgaardForbesBits And AtomsBy Rich KarlgaardForbesNavigating America’s PopulismBy Rich KarlgaardForbesSmarts In The Age Of AIBy Rich KarlgaardForbesElectricity WarsBy Rich KarlgaardForbesVinod Khosla On AI’s Fast Rise, Robotics, Entrepreneurs And InnovationBy Rich KarlgaardForbesWho Learns Fastest, WinsBy Rich KarlgaardForbesMoonshots Vs. Marginal GainsBy Rich KarlgaardForbesGreat Plains Presidential Dark Horse — Doug BurgumBy Rich KarlgaardForbesThe Great AI Deflation BombBy Rich Karlgaard