Bernstein "throws cold water": Musk's super chip factory would require $5-13 trillion in capital expenditures, "harder than landing on Mars"

Elon Musk's ambitious "Terafab" plan is being scrutinized rationally by Wall Street.

According to Chase Wind Trading Desk, Bernstein Research’s latest report provides detailed quantitative analysis, pointing out that to achieve Musk’s envisioned annual production target of 1 terawatt of computing power, it would require capital expenditures of $5 to $13 trillion, and wafer production equivalent to the total current global semiconductor capacity. This challenge is "harder than landing on Mars."

Last weekend, Musk announced the launch of the "Terafab" project, planning to expand human computing power production to 1 terawatt per year, about 50 times the current global supply (approximately 20 gigawatts). The first phase involves building an advanced wafer fab in Austin, Texas, covering the full production chain of computing chips, logic chips, memory, packaging, and masks. Bernstein analyst Stacy A. Rasgon and his team immediately published a report, offering a systematic quantitative evaluation of the plan’s feasibility.

Bernstein points out that if investors truly believe Musk can achieve this goal, semiconductor equipment stocks would be the most direct beneficiaries; but for now, the actual impact of this plan on the industry "may not be significant and remains mostly hype."

Astronomical Capital Expenditure Estimates

Bernstein used the existing computing power architecture (Nvidia racks) as the benchmark to roughly estimate wafer demand for three core chips: GPU, HBM memory, and CPU. The report shows that to produce 1 terawatt of computing power annually, it would require a monthly startup volume of 7 to 18 million 300mm wafers, with HBM memory demand dominating.

Using a factory with a monthly capacity of 50,000 wafers as a unit, this would be equivalent to building 140 to 360 such factories. With a capital expenditure of $3.5 billion per factory, the total investment would reach $5 to $13 trillion. Bernstein clearly notes in the report that these are "very rough estimates" and do not include other semiconductor categories like networking, optical, analog, and power chips.

Production Requirements Rival Rebuilding the Global Semiconductor Industry

Bernstein’s calculations reveal the shocking scale of production capacity needed for this plan. The report points out that the wafer production required for 1 terawatt of computing power is roughly equivalent to the current total global semiconductor wafer capacity (about 16 million 300mm equivalent wafers per month).

If only "related" semiconductors are considered—that is, memory plus advanced logic wafers at 4nm or below—current global installed capacity is about 5 million 300mm equivalent wafers per month, while the capacity required for the 1 terawatt goal is several times that amount. In other words, Musk’s plan essentially requires expanding advanced capacity by multiple times over what currently exists, an extremely difficult feat.

Industry Impact: Equipment Stocks Benefit, No Immediate Threat to Foundry Model

Bernstein believes that Terafab’s short-term real impact on the semiconductor industry is limited but highlights several investment aspects to watch.

The report points out that if the market believes Musk can carry out this plan, semiconductor equipment (semicap) would be the most direct beneficiary.

Regarding whether Musk’s in-house chip fabrication would impact existing chip companies, Bernstein takes a relatively optimistic view: in an environment with such strong computing demand, all participants will face scalable opportunities far beyond their capacity, including memory suppliers.

On the business model front, Bernstein notes that Terafab integrates logic, memory, mask manufacturing, chip design, and packaging into one—essentially forming a "super IDM" model. However, this model has proved far less efficient than the "foundry + fabless + specialized memory IDM" division of labor system.

Therefore, the report believes that Terafab poses limited threats to foundries like TSMC for now.

Bernstein: Doesn’t Rule Out Musk, but Challenges Are "Extremely Difficult"

Despite the jaw-dropping numbers, Bernstein does not entirely rule out the possibility of Musk succeeding.

The report states that Musk has repeatedly accomplished what outsiders believed impossible in the past, and "we won’t easily dismiss him."

However, the report also clearly states, "A true Terafab, to us, feels more like an extension, especially if based on the current computing paradigm."

Bernstein raises two possible alternative paths:

First, if Musk cannot push forward independently, he may seek partnerships with existing chip manufacturers;

Second, Musk may have some "even more unexpected" technological path to break the current paradigm, but the report admits, "We don’t know what that would be."

~~~~~~~~~~~~~~~~~~~~~~~~

The above content is from Chase Wind Trading Desk.

For more details, including real-time commentary, frontline research, and other content, please join [Chase Wind Trading Desk ▪ Annual Membership]

Risk Warning and DisclaimerThe market carries risk; invest with caution. This article does not constitute personal investment advice and does not take into account individual users’ specific investment targets, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article suit their individual situation. Investments made based on this are at your own risk.