SpaceX plans to invest $55 billion in Texas to build a chip manufacturing facility called Terafab, with total future investment potentially reaching $119 billion—nearly matching NVIDIA’s revenue in fiscal year 2024. This project is a joint initiative between SpaceX and Tesla, as Musk’s companies consume massive amounts of GPU computing power for applications including xAI’s Grok model, Starlink ground stations, Tesla’s autonomous driving systems, and the Optimus humanoid robot. The goal is to achieve vertical integration in chip manufacturing by unifying computation production, energy procurement, and computing deployment under one management structure. Analysts view this as Musk’s “15-year strategy,” with initial chip output not expected until mid-2028. The plan reflects the AI industry’s deep obsession with computational autonomy, and SpaceX’s IPO window is set to open around June 8.

Author and source: GeekPark

In the 1980s, Japan's semiconductor industry was at its peak, once accounting for more than 50% of the global chip market.

Americans could no longer sit idle—not because they couldn’t get chips, but because they realized that a nation unable to control its own “silicon” is handing over its strategic lifeline to others. This led to the Semiconductor Agreement, followed by Intel’s resurgence, and today’s CHIPS Act.

Forty years later, the same anxiety appeared in Elon Musk in a more personal way.

This time, however, it is not a nation but one person—and the cluster of companies under his control—that seeks to dominate "silicon".

On May 6, Bloomberg disclosed a document stating that SpaceX has formally proposed investing $55 billion in Grimes County, Texas, to initiate the construction of a chip manufacturing facility called "Terafab." If all subsequent phases proceed, total capital expenditures could reach $119 billion.

First, convert the numbers to get a sense of the scale.

$55 billion is more than twice TSMC's full-year revenue in 2023. $119 billion is close to NVIDIA's peak revenue for the entire 2024 fiscal year. This is not an "investment"—it's a high-stakes gamble—or rather, a strategic declaration.

This project is a joint initiative between SpaceX and Tesla. Musk’s logic is straightforward: his ecosystem of companies—SpaceX, Tesla, xAI—consumes massive amounts of GPU computing power annually. Training Grok requires chips, Starlink’s ground station network requires chips, Tesla’s autonomous driving system requires chips, and future humanoid robots like Optimus will also require chips.

Instead of paying NVIDIA every year, pay yourself.

From a strategic standpoint, this is flawless.

01 Elon Musk’s vertical integration ambitions

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

In 2025, xAI acquired the social media platform X. Earlier this year, SpaceX swallowed xAI in an all-stock deal. Meanwhile, SpaceX’s IPO plan is progressing, with the roadshow window reportedly opening around June 8, and the S-1 registration filing expected by the end of this month.

It’s an increasingly dense web: rockets, satellite broadband, AI models, social platforms, electric vehicles, humanoid robots… and now, chip manufacturing is also being brought into the fold.

DataCenter Knowledge's analysis cuts to the chase: Terafab should not be understood as a "factory," but rather as a comprehensive AI infrastructure strategy—aiming to unify compute production, energy procurement, and compute deployment under one roof.

It's like Amazon not only wanting to sell books but also building its own shipping network, data centers, and logistics satellites—except Musk is adding one more layer: building a "factory that produces computing."

Apple’s decision to internally develop its A-series chips was considered one of the most successful vertical integration moves in tech history. But Apple only “designed” the chips, leaving manufacturing to TSMC. Musk aims to take over manufacturing as well.

This ambition, even Apple hasn't dared to touch.

02 Behind the "15-Year Strategy"

However, there is always a deep chasm between strategic vision and engineering reality.

Creative Strategies’ chip analyst Ben Bajarin used an interesting phrase, saying that Musk is pursuing a “15-year strategy”—this sounds like a compliment, but the underlying meaning is: don’t expect short-term returns.

Morgan Stanley’s forecast is more direct. They estimate that, even under the most optimistic construction assumptions, Terafab’s initial chip output won’t arrive until mid-2028. That’s more than two years from now, and by then, no one knows where AI chip technology will stand.

The cruelty of semiconductor manufacturing lies in the fact that it is the least forgiving industry in the world when it comes to "PPT promises."

Building a state-of-the-art semiconductor fab typically takes 3–5 years, requires extremely precise lithography machines (only ASML can produce high-end EUV machines globally), thousands of highly specialized engineers, a stable supply of ultra-pure water and electricity, and a comprehensive precision cleanroom management system. Intel has spent tens of billions of dollars and years of effort, yet still struggles to catch up to TSMC in process technology.

Finance Monthly's analysis directly highlights this risk: chip projects are easily underestimated in terms of execution difficulty, are slow and expensive, prone to delays, and rely on specialized machinery, skilled labor, and supply chains that are already under significant strain.

Interestingly, when asked about Terafab, Intel CEO Patrick Gelsinger gave a nuanced response, saying he was “excited to explore innovative ways to reimagine silicon manufacturing technology.” This statement can be interpreted as a signal of potential collaboration, a subtle acknowledgment of market supply constraints, or both.

03 More than just business

But viewing Terafab solely through the lens of a business return on investment calculation may be the wrong analytical framework from the start.

What’s truly interesting about this is that it reflects the AI industry’s growing obsession with “compute autonomy.”

Over the past three years, the essence of the AI arms race has shifted from “whose model is smarter” to “who can secure more computing power.” NVIDIA’s H100, H200, and GB200 are in extremely short supply, and TSMC’s advanced manufacturing capacity has been locked in years in advance. Microsoft, Google, Amazon, and Meta have each invested tens of billions of dollars in developing their own AI chips.

Elon Musk's logic is the same as that of these big companies: In the AI era, computing power is a means of production; whoever controls the chips controls AI.

The emergence of Terafab,叠加在 SpaceX 上市之前，又增添了一层更复杂的含义。一位科技媒体编辑在评论中写道，这一公告的部分设计意图是「将面临压力的特斯拉，与即将上市的 SpaceX 和 AI 超级计算叙事绑定在一起」。

This judgment may not be entirely fair, but it’s not without basis. Just a few months ago, Musk openly admitted that “xAI was built incorrectly” and had to be integrated by SpaceX. Against this backdrop, Terafab’s high-profile debut is both a strategic move and a capital narrative—and these two things are not mutually exclusive.

Elon Musk never does just one thing.

The launch window for Starship Flight 12 is between May 12 and 18, and the Dragon cargo spacecraft will also be heading to the International Space Station on the same day. SpaceX’s rocket operations continue to operate at full steam.

In Grimes County, Texas, a plot of land that could reshape the chip supply chain in the AI era is still just a set of coordinates on a document.

$119 billion—no one can guarantee it will become a fully operational semiconductor fab. But one thing is certain: when a company known for building rockets decides to make chips, the boundaries of this industry are once again being redefined.