Lambda's Stephen Balaban on data center misinformation, US energy sovereignty, and why offshoring compute would be a national tragedy

Lambda's Stephen Balaban on data centers, misinformation, and energy sovereignty

The case against AI data centers rests largely on bad facts, and Stephen Balaban argues the fastest way to win that debate is to correct them directly rather than fight the political framing around them.

Water is his first example. The dominant narrative draws on older evaporative cooling systems that consume significant water. The vast majority of new AI facilities now use closed-loop dry cooling, where water never leaves the building. Balaban says that single clarification, offered to polling respondents, lifted net approval scores by roughly 30 percentage points.

The harder problem is distribution. Anti-AI content on YouTube attracts multiples of the views that factual corrections do. Balaban notes, citing Taylor Lorenz, that AI safety organizations quietly pay creators to produce undisclosed negative content, compounding the asymmetry.

---

“The vast majority of new AI factories that are being built are completely recycling the water and using a dry cooler. It's a closed loop water system where the water never even leaves the facility at all... We've got a data center out in Kansas City and 300 folks worked on it during the construction period and 50 full time jobs, many making over $100,000 a year... AI is a national strategic asset in my opinion.”

Jobs and local economics

Lambda's Kansas City data center involved 300 construction workers and created 50 permanent jobs, many paying over $100,000 a year. Over five years, Lambda expects to pay $47 million in property taxes to the local community from a 10,000-GPU facility.

Balaban's position on governance is that this is already a local decision, not a federal one. Communities that welcome data centers attract investment; those that don't, won't. He points to Loudoun County, Virginia as an example of a community that pioneered the model and benefited. He expects data centers to build reputations by paying taxes, keeping utility rates down for residents, and supporting local supply chains rather than lobbying for preferential treatment.

The political framing is more complicated than a clean partisan split. In Kansas, steamfitters and construction unions are actively backing a pro-data center gubernatorial candidate, which cuts against the narrative of a bipartisan anti-data center coalition.

---

Offshoring compute would be a national tragedy

Balaban is direct that regulatory hostility pushing data center investment to Mexico, South America, or Asia Pacific would be a strategic mistake. His argument is not purely economic. AI compute, in his view, is a national strategic asset on par with the electrical grid, and offshoring it at the moment the US is rebuilding its grid would undermine a much larger reindustrialization thesis.

The grid argument is specific: 80% of new megawatts coming online in 2026 are solar or storage. Demand from data centers is accelerating that buildout, and Balaban argues the resulting grid capacity creates conditions for broader industrial onshoring, including arc electric furnace mills, specialty metal smelting, and turbine manufacturing.

He also makes a harder national security point. When one frontier AI model experienced an outage, it felt, in his words, like having electricity cut off. He says compute access is as strategically significant as grid access, and that the risk of a foreign government severing that access is real, not theoretical.

---

Energy and the next misinformation cycle

Balaban expects the water debate to give way to an emissions debate. As compute scales and natural gas remains part of the power mix, data-center-to-emissions equivalencies are likely to become the next viral attack vector.

His policy prescription is grid interconnection reform. Behind-the-meter power generation, often used when interconnection queues are too slow, relies on simple-cycle gas rather than more efficient combined-cycle plants. Getting data centers onto the main grid faster, he argues, is both cleaner and more efficient. Deregulating the interconnection process is the single policy lever he prioritizes.

On the emissions question itself, Balaban resists framing carbon as the central issue. The more relevant comparison, he says, is the trajectory of American energy production versus China's, both of which are now adding predominantly solar and storage. The grid will decarbonize as storage and renewables scale; the question is whether the US builds the industrial base to benefit from that transition or exports the opportunity.