Author: Hualin Wuwang

In the 1980s, Japan’s semiconductor industry was at its zenith, commanding over 50% of the global chip market.

Americans grew restless—not because they couldn’t buy chips, but because they realized that a nation unable to control its own “silicon” effectively surrendered its strategic lifeline to others. Thus came the Semiconductor Agreement, Intel’s subsequent resurgence, and today’s CHIPS Act.

Forty years later, the same anxiety resurfaced—but this time in a far more personal form, gripping Elon Musk.

Only this time, it’s not a nation seeking control over “silicon,” but an individual—and the constellation of companies under his command.

On May 6, Bloomberg disclosed a document revealing that SpaceX has formally proposed a $55 billion investment in Grimes County, Texas, to launch construction of its “Terafab” chip manufacturing facility. If all subsequent phases proceed as planned, total capital expenditure could reach $119 billion.

Let’s first convert those figures to grasp the scale.

$55 billion is more than double TSMC’s full-year 2023 revenue. $119 billion approaches NVIDIA’s entire peak revenue for fiscal year 2024. This isn’t merely an “investment”—it’s a high-stakes gamble—or rather, a strategic declaration.

The project is a joint initiative between SpaceX and Tesla. Musk’s logic is clear-cut: his cluster of companies—SpaceX, Tesla, and xAI—consume staggering volumes of GPU compute annually. Training Grok requires chips; Starlink’s ground station network requires chips; Tesla’s autonomous driving systems require chips; and future humanoid robots like Optimus will require chips too.

Rather than paying NVIDIA billions every year, why not redirect that money inward?

Strategically, this logic is unassailable.

01 Musk’s Vertical Integration Ambition

To understand Terafab, you must first understand what Musk has been doing over the past two years.

In 2025, xAI acquired the social media platform X. Earlier this year, SpaceX acquired xAI entirely via stock. Meanwhile, SpaceX’s IPO process is advancing, with its roadshow window reportedly set to open around June 8 and its S-1 registration filing expected by late this month.

This is a rapidly tightening web: rockets, satellite broadband, AI models, social platforms, electric vehicles, humanoid robots—and now, chip manufacturing.

DataCenter Knowledge cut straight to the point: Terafab should not be viewed merely as a “factory.” Instead, it represents a “full-stack AI infrastructure strategy”—an effort to unify compute production, energy procurement, and compute deployment under one roof.

It’s akin to Amazon not only selling books, but also building its own freight network, data centers, and logistics satellites—except Musk intends to add one more layer: building his own “factory for producing computation.”

Apple’s decision to internally design its A-series chips is widely regarded as one of the most successful vertical integration moves in tech history. Yet Apple only designs chips—their manufacturing remains outsourced to TSMC. Musk aims to take even manufacturing in-house.

That ambition—even Apple dared not attempt.

02 The “15-Year Strategy” Behind It

Still, there remains a chasm between strategic vision and engineering reality.

Ben Bajarin, semiconductor analyst at Creative Strategies, offered an incisive observation: he described what Musk is undertaking as a “15-year strategy.” While this phrasing may sound complimentary, its underlying implication is stark: don’t expect near-term returns.

Morgan Stanley’s forecast is even more direct. Even under the most optimistic construction assumptions, Terafab’s initial chip output won’t begin until mid-2028—at least two-and-a-half years from now. By then, no one knows where AI chip technology generations will stand.

Semiconductor manufacturing is brutally unforgiving—perhaps the world’s least tolerant industry of “PPT promises.”

Building an advanced-node wafer fab typically takes three to five years. It demands extremely precise lithography tools (only ASML manufactures cutting-edge EUV machines), thousands of highly specialized engineers, stable supplies of ultra-pure water and electricity, and a comprehensive cleanroom management system. Intel spent tens of billions of dollars and multiple years struggling to close the process-technology gap with TSMC.

Finance Monthly directly highlighted this risk: chip projects are often underestimated in terms of execution difficulty—prone to delays, costly overruns, heavy dependence on specialized machinery, skilled labor, and an already severely strained supply chain.

Interestingly, when asked about Terafab, Intel CEO Dr. Lip-Bu Tan responded with carefully worded diplomacy: “We’re excited to explore innovative ways to rearchitect silicon process technologies.” This statement can be read either as a signal of potential collaboration—or as tacit acknowledgment of tightening market supply-demand dynamics—or perhaps both.

03 More Than Just Business

Yet viewing Terafab solely through the lens of ROI calculations would misframe the analysis from the outset.

What makes this truly fascinating is how it reflects the AI industry’s deepening obsession with “compute autonomy.”

Over the past three years, the AI arms race has evolved—from “whose model is smarter?” to “who can secure more compute?” NVIDIA’s H100, H200, and GB200 chips are perpetually in short supply; TSMC’s advanced-node capacity has been booked years in advance. Microsoft, Google, Amazon, and Meta have each invested tens of billions of dollars into developing their own AI chips.

Musk’s logic aligns precisely with these tech giants: in the AI era, compute *is* the means of production—and whoever controls chips controls AI.

Moreover, Terafab’s announcement—timed alongside SpaceX’s imminent IPO—adds yet another layer of complexity. A tech media editor commented that part of the announcement’s design appears intended to “tie Tesla—under mounting pressure—to the upcoming IPO of SpaceX and the AI supercomputing narrative.”

This assessment may not be wholly objective—but neither is it baseless. Just months ago, Musk himself admitted “xAI was built incorrectly,” necessitating SpaceX’s intervention to integrate it. Against that backdrop, Terafab’s high-profile unveiling serves both as strategic positioning and as capital-market storytelling—two objectives that need not conflict.

Musk has never done just one thing.

Now, Starship Flight 12’s launch window falls between May 12 and 18; a Dragon cargo spacecraft is also scheduled to launch to the International Space Station on the same day. SpaceX’s rocket business continues operating at full throttle.

Meanwhile, in Grimes County, Texas, a plot of land that could reshape the AI-era chip supply chain exists—for now—only as coordinates on a document.

No one can guarantee whether $119 billion will materialize into a fully operational wafer fab. But one thing is certain: when a company renowned for building rockets decides to build chips, the boundaries of this industry have once again been redrawn.