Kleiner Perkins

The Unseen Fuel of the Intelligence Age Artificial intelligence has triggered a modern gold rush, but the miners are not looking for gold. They are looking for power. Silicon Valley has poured billions into chips, algorithms, and data centers. Yet, the entire system relies on a physical asset: the electrical grid. Bloom Energy founder and CEO KR Sridhar argues that the digital revolution has hit a physical wall. The mechanical-age infrastructure built over a century ago cannot handle the explosive needs of high-performance computing. This gap between software speed and hardware reality is creating massive friction. Traditional utilities run on slow regulatory cycles and rigid distribution networks. Meanwhile, AI startups and tech giants need gigawatts of power yesterday. The challenge is not just about producing more electricity. It is about restructuring how power is generated, delivered, and consumed. Sridhar believes the answer lies in distributed power at the edge, a concept his company has championed for twenty-five years. The Failure of One-Size-Fits-All Infrastructure To understand the energy crisis, we have to understand how our current system works. Traditional grids rely on centralized power plants built far away from cities. High-voltage lines carry electricity over hundreds of miles to sub-stations, which then distribute it to consumers. This mechanical setup made sense when energy demands were predictable and uniform. Today, it is a liability. Centralized generation loses vast amounts of energy in transmission. It is also vulnerable to weather disruptions, cyberattacks, and physical sabotage. For a modern AI factory, a power outage is not a minor inconvenience. It is catastrophic. When millions of dollars are riding on continuous training runs, even a millisecond drop in voltage ruins valuable progress. Furthermore, the legacy grid is structurally incapable of handling the rapid spikes of AI workloads. Graphics processing units (GPUs) consume power in violent bursts. They pulse with intense activity, then drop to idle. Steam turbines and gas combustion engines cannot ramp up or down fast enough to match these digital fluctuations. This leaves operators dependent on massive batteries to smooth out the loads, adding cost and complexity to already expensive builds. Walking the Shop Floor with Andy Grove Building a company to challenge the centralized energy model required more than just technical brilliance. Sridhar, a former NASA rocket scientist, had to learn how to scale manufacturing from scratch. In the early days of Bloom, the company struggled to transition from handmade laboratory prototypes to reliable factory-built units. The turning point came during a high-stakes board meeting featuring legendary Intel leader Andy Grove. After reviewing binders of engineering data, Grove dismissed the entire executive team from the room, leaving only Sridhar. He did not ask about technical specifications or material science. Instead, he asked a direct question: "What is wrong with you?" Grove pointed out that Sridhar had lost touch with the workers actually assembling the product. He urged Sridhar to walk the factory floor and listen to the technicians. That advice shifted Sridhar's leadership style toward radical empathy. He realized that the people building the machines knew more about the manufacturing flaws than any executive looking at a spreadsheet. By bringing that direct feedback loop back to engineering, Bloom solved its early reliability issues. This hands-on, detail-oriented approach laid the foundation for the company's current ability to scale production to multiple gigawatts. Why Energy Sovereignty Rules the New Geopolitical Order For decades, tech executives focused on software sovereignty and model development. They believed that owning the best algorithms and datasets would guarantee global dominance. Sridhar argues this is a profound misunderstanding of the physical world. Energy sovereignty is the actual bottleneck of the coming decade. Nations cannot build robust digital economies if they rely on hostile adversaries for fuel. The war in Ukraine exposed Europe's dangerous dependence on Russian gas. It showed that foreign energy dependence can cripple an economy overnight. Moving forward, true power will belong to nations that can generate and secure their own energy supply. This reality is forcing a major shift in how governments and corporations think about resource allocation. Sridhar advocates for a "free fuel from a free world" strategy. He believes democratic nations must collaborate to build secure energy supply chains. By using abundant domestic resources like natural gas and hydrogen, societies can insulate themselves from geopolitical blackmail. Decoupling Power from the Grid To solve the energy bottleneck, companies must bypass the grid entirely. The solution is distributed generation, placing power plants directly at the edge, next to the data centers they support. This strategy eliminates transmission losses and removes the risk of grid failure. Oracle recently embraced this model. Under the leadership of Larry Ellison, the tech giant turned to Bloom Energy's solid-oxide fuel cells to power its critical data centers. When traditional infrastructure delays threatened Oracle's timeline, Bloom deployed a fifty-megawatt system in just fifty-five days. This rapid deployment is possible because fuel cells operate like computer hardware rather than heavy machinery. They are solid-state devices that generate electricity through an electrochemical process. They have no large moving parts, meaning they do not require the extensive maintenance shutdowns that plague mechanical turbines. More importantly, solid-state power systems are modular. Operators can add capacity in small blocks, much like adding server blades to a data center rack. If one module needs service, technicians can hot-swap it without taking the entire facility offline. This modular architecture matches the natural scaling trajectory of growing tech companies, allowing them to scale their power generation in lockstep with their compute capacity. Bridging the Gap to a Hydrogen Future While the immediate demand for power is being met by domestic natural gas, the long-term vision points toward hydrogen. Sridhar sees natural gas as a necessary transition fuel. It has the lowest carbon footprint of any available fossil fuel molecule, making it a pragmatic choice for the current infrastructure scale-up. However, the ultimate goal is localized self-reliance. In the future, microgrids will capture wind, solar, and geothermal energy when it is abundant. They will convert the excess electricity into hydrogen on-site. When renewable sources drop offline, fuel cells will consume that stored hydrogen to provide continuous power. This closed-loop system will allow isolated communities and critical industrial sites to operate independently of weather patterns and unstable national grids. This shift to localized generation also addresses the massive waste inherent in centralized systems. Centralized power plants discard the thermal energy generated during electricity production. By placing power generation at the edge, operators can capture this waste heat and use it to cool the data centers themselves. This double-efficiency loop dramatically lowers the total environmental impact of high-compute facilities. The Path to Energy Abundance Rather than fearing a prolonged energy crisis, Sridhar views the AI infrastructure squeeze as a massive catalyst. The urgent need for computational power is forcing the world to build the next generation of energy technology. The capital flowing into tech-adjacent energy projects will accelerate clean energy innovation, driving down costs for everyone. Just as safety features like airbags started in high-end luxury vehicles before becoming standard in budget cars, digital power technologies will trickle down. The high-margin business of manufacturing intelligence is funding the development of robust, decentralized energy systems. Eventually, this technology will provide clean, reliable, and affordable power to developing regions that have historically suffered from energy poverty. By decentralizing power, society can finally democratize opportunity. When every community can secure its own energy and digital access, economic growth will no longer be concentrated in a few crowded metropolitan hubs. The intelligence revolution is fundamentally an energy revolution. Winning that revolution requires building a smarter, cleaner, and faster grid at the edge.