Elon Musk’s $119B “Terafab” chip mega-plant in Texas could reshape the global AI arms race

Elon Musk is floating one of the biggest manufacturing bets in modern tech: a proposed semiconductor “gigafab” in Texas that local documents peg at $55 billion in an initial buildout, and as much as $119 billion if it expands to its full vision.

The project, dubbed “Terafab,” surfaced not through a Musk keynote but in a public notice from Grimes County, north of Houston. A public hearing is set for June 3 to consider a property-tax abatement deal, an early, telling sign that the plan is moving into the gritty world of permits, incentives, and infrastructure.

A Texas county document puts real numbers on Terafab: $55B now, up to $119B later

The clearest snapshot of Terafab so far comes from local government paperwork. Grimes County’s notice describes an estimated $55 billion first phase, with a total investment that could climb to $119 billion if additional phases are built.

That wording matters. In megaprojects, “up to” often means a ceiling, not a guarantee, more a roadmap of possible expansions than a firm commitment to spend every dollar.

The same notice tees up the June 3 public hearing focused on a property-tax abatement agreement. In Texas, those deals are a standard tool to lure big industrial projects, with local officials betting that jobs and long-term economic activity will outweigh near-term tax breaks.

Near Houston, the proposed site is all about power and cooling

The location named in the filing sits near the Gibbons Creek Reservoir, about 81 miles northwest of Houston. The man-made lake was built in the 1980s to help cool a power plant, an on-the-nose detail for a facility meant to churn out advanced chips and potentially support massive AI computing loads.

Chip fabs and data centers are energy-hungry and water-intensive, especially when you’re talking about high-density computing that generates enormous heat. Putting Terafab near an area already shaped by the energy industry suggests a strategy: plug into existing infrastructure rather than start from scratch.

But the scale being discussed would still test local capacity, transmission lines, substations, water systems, roads, housing, and emergency services. Those pressures tend to surface quickly once a project moves from paper to earthmoving.

Musk’s eye-popping claim: a terawatt of computing per year

Musk has framed Terafab as a solution to a bottleneck he says is getting worse: access to chips and compute for AI. He has described the goal in staggering terms, producing a “terawatt” of computing power per year.

Even allowing for the fuzzy way tech leaders sometimes talk about “compute,” the message is clear: this isn’t pitched as just another factory. It’s being sold as strategic infrastructure for AI model training, robotics, and data processing tied to space operations.

Industry veterans tend to split into two camps when they hear claims like this. One side sees a familiar playbook, vertical integration, like when major tech companies built their own supply chains to control destiny. The other side hears alarm bells: semiconductor manufacturing is brutally complex, and success is measured not by ambition but by yields, process stability, and execution over years.

Tesla was mentioned publicly, but the county paperwork names SpaceX

When Musk discussed Terafab, he suggested it would be a joint effort involving Tesla and SpaceX. But in the Grimes County notice, only SpaceX appears by name.

That doesn’t prove Tesla is out. Big projects often shift between corporate entities for financing, land ownership, and tax negotiations. Still, the discrepancy underscores how early-stage the structure may be, and how much remains unsettled behind the scenes.

The proposed mission is broad: semiconductors and advanced computing for AI, robotics, and space-linked data centers. Musk has argued that demand inside his companies could outstrip what the global chip market can reliably supply, an argument that resonates as AI accelerators remain scarce and expensive.

The real choke point: energy, not hype

If Terafab advances, the biggest constraint may not be concrete or clean rooms, it may be electricity. AI data centers are already straining regional grids across the U.S., and chip manufacturing adds another layer of heavy, continuous demand.

That’s why the June 3 hearing matters beyond taxes. Local residents and officials often use these meetings to press for commitments on infrastructure upgrades, water use, traffic, and housing impacts, especially in areas that could be transformed by a massive construction surge.

Musk’s terawatt talk, you interpret it, spotlights a reality the entire AI industry is running into: the race for compute eventually collides with power generation, grid buildout, permitting timelines, and politics.

Intel and top chip-equipment makers are reportedly in the mix

Reports around Terafab have also pointed to outreach involving Intel and major semiconductor equipment suppliers such as Applied Materials, Tokyo Electron, and Lam Research, companies that make the specialized tools modern fabs can’t function without.

Those names add credibility to the idea that Musk is trying to plug into the existing semiconductor ecosystem rather than reinvent it. But they don’t erase the central challenge: building a fab is one thing; achieving high yields at scale is another, and it typically takes deep expertise and time.

Terafab is reportedly aimed at supplying chips for SpaceX, Tesla, and Musk’s AI venture xAI. That kind of in-house pipeline could reduce dependence on overstretched global suppliers, but it also raises the risk of building for fast-moving internal needs in a field where architectures can change before a factory is fully ramped.

If the project becomes real, it won’t just be a Texas economic-development story. It could become a stress test for how far one region, and one billionaire-led ecosystem, can push the next phase of the AI boom before the grid, the supply chain, or the timeline pushes back.