Power, not incentives, now decides where AI data centres get built, says FiberLight CEO

He points to Oklahoma and Western Pennsylvania as the clearest evidence of the shift.

The site selection conversation has changed

Major says developers are no longer opening negotiations by asking what incentives a state can offer. “The conversation starts to change from, ‘What incentives can you give me?’ to, ‘Can you actually give me the power I need, and when can I get it?’” he says. Economic sweeteners, in his view, don’t solve the underlying problem: if a site cannot deliver the power a large AI facility needs within the developer’s timeframe, “the economics don’t matter.”

Oklahoma and Western Pennsylvania are, in his account, where that logic is playing out most visibly. Oklahoma has attracted major investment from Meta and Google, while Amazon has committed $20 billion to AI infrastructure in Pennsylvania, with developers drawn to the state’s existing energy resources as a way to get power online faster. Major frames the pattern as broader than any single deal: “developers are following the infrastructure, particularly power, that can support AI at scale.”

Land and incentives still matter, he says, but they’re no longer sufficient on their own. The advantage now goes to markets that combine power, land, water where required, workable permitting, and fibre connectivity to link that compute back into the wider network. “The next generation of data centre hubs won’t necessarily be the markets that won the last cycle,” Major says. “They’ll be the ones that can deliver power, connectivity and capacity quickly enough to keep up with where the demand is going.”

Behind-the-meter power is a workaround, not a solution

Asked about the growing use of behind-the-meter generation and natural gas as alternatives to grid connection, Major is cautious rather than bullish. “There’s no magic bullet here,” he says. Off-grid generation can shorten timelines when the grid can’t deliver power fast enough, but developers take on capital costs, fuel, emissions and permitting complexity in exchange. His broader point is about patience, or the lack of it: “developers aren’t going to sit around indefinitely waiting for one solution. If the traditional grid connection takes too long, they’ll look at other options.”

He also warns against treating power and connectivity as separate problems. A facility can secure its own generation and still be stranded without the fibre to move data in and out. Major expects site decisions to increasingly treat power, compute and connectivity “as integrated” rather than sequential considerations.

FiberLight isn’t slowing down in Texas

Major, whose company is headquartered in Texas, was asked directly how the state’s new data centre audit has affected deal timelines. He calls Abbott’s directive “a measured approach to new data centre development,” adding that scrutiny of grid impact, ratepayer costs, water resources and infrastructure funding is warranted “given the state’s experience when its power grid is challenged.”

For FiberLight specifically, he says the audit hasn’t disrupted anything: the company’s projects are already approved and under construction on a five-year planning horizon. Earlier this year, FiberLight announced a $500 million capital investment in West Texas to expand its network for data centre, hyperscaler and AI workload demand, a commitment Major says reflects an expectation that demand keeps growing regardless of near-term regulatory scrutiny. “Taking a more deliberate approach to new development doesn’t mean the underlying demand and construction is going away,” he says.

A familiar pattern, at a different speed

Major draws a direct comparison to early public resistance to cell towers, which he says followed a similar arc before the technology became indispensable. What’s different with AI, in his view, is the pace and physical scale of the buildout, driven by a technology that went from emerging to mainstream in a remarkably short period.

He casts the current wave of local opposition as a prompt for the industry to solve harder problems faster, rather than a reason to pull back. “The U.S. has a tremendous opportunity to lead in AI,” he says, “but maintaining that position means continuing to invest and finding practical solutions that allow us to build at the pace the technology and the market demand.”

Data centres could use 2.7% of Texas’s water by 2030: Where they land matters