British "Financial Times" columnist June Yoon wrote on Wednesday that Tesla's market value of US$1.3 trillion shows that the market expects the company's future to rely more on AI rather than electric vehicles.To meet this expectation, Tesla must have greater control over the hardware needed to support autonomous driving, robotics and AI training.

Musk plans to build his own chip factory TeraFab

Tesla CEO Elon Musk said that the company therefore needs to build and operate a so-called "TeraFab" large-scale chip factory to produce its own chips. But can Tesla really make its own chips?

This idea should be viewed separately from Musk's more ambitious visions. Compared with fanciful projects such as colonizing Mars or brain-computer interfaces, building chip manufacturing factories can rely on existing mature technologies.

It doesn’t have to be cutting-edge

The industry's doubts about new entrants largely stem from a misunderstanding: when people talk about chip manufacturing, they often think of the 3-nanometer and 5-nanometer cutting-edge process technologies used by TSMC to produce Nvidia's flagship AI chips. Such high-end chips require state-of-the-art equipment and decades of accumulated trial and error experience. By this standard, new entrants indeed seem to have no chance of winning.

However, this comparison has its own problems. Chip manufacturing companies do not have to reach TSMC's most advanced process level to gain survival space. There is actually a long and narrow middle ground in the chip industry: new entrants can meet the technical requirements, and at the same time, the process performance is sufficient to support AI computing loads.If Tesla's wafer fab is built, it will inevitably need to target the older generation technology around 7 nanometers. This process node is widely considered to be the last advanced chip generation before complexity and capital requirements soar.

TSMC first began mass production of 7-nanometer chips in 2018, and the technology is still widely used in AI and data centers. Even if it lags behind by several years, Tesla can in principle produce AI computing chips suitable for autonomous driving and bionic robots.

Many challenges

This means that 7nm will be a suitable technology benchmark for Tesla, but it is by no means an easy target. It still requires ASML's extreme ultraviolet lithography machines, ultra-clean facilities with large-scale power, cooling and water treatment systems, hundreds of precision equipment and advanced chip packaging capabilities.

Most critically, it requires hundreds of engineers experienced in reducing chip defect rates, and this scarce talent is currently mainly concentrated in TSMC. Initial production typically takes three years or more, with high material loss and a lengthy trial-and-error process to achieve usable output.

TSMC

Even if Tesla can reach the technical threshold, feasibility alone cannot guarantee value creation. In the field of chip manufacturing, the current benchmark is TSMC, which spent more than $40 billion in capital expenditures last year. The expense is justified by spreading the risk across a large customer list and thousands of designs, helping absorb production losses and accelerate technology iterations.

Tesla cannot replicate this model.Musk said Tesla does not plan to sell chips externally.Without global orders, the chip business becomes a fixed-cost operating system, and it is difficult to avoid structural losses. TSMC's past investments in building factories in the United States indicate that each fab costs at least $20 billion to build (even including subsidies). Unlike most industrial assets, chip manufacturing plants require continuous reinvestment to remain competitive. Even under the most optimistic assumptions, the payback period could be decades.

Economic factors aside, execution risk remains the greater challenge.Despite Intel's decades of industry experience, the company's transition to 10-nanometer chips since the mid-2010s has been repeatedly frustrated by aggressive timelines and internal pressure, resulting in years of delays and a permanent loss of industry leadership.

There have been reports that the quality of Tesla's electric vehicles is uneven, such as uneven gaps in body panels and repairs after delivery, but this does not necessarily directly equate to insufficient technical capabilities. However, it does reflect the company's tendency to rush to deliver products before the production process is fully stable. For cars, most of these defects are cosmetic and can be repaired. But in the field of chip manufacturing, this operating model simply does not work.

Failure case

GlobalFoundries in the United States is another case worth learning from. After acquiring IBM's money-losing chip business in 2015, it took just three years for the company to conclude that advanced chip manufacturing was not economically viable.

Tesla will face both types of risks: the cultural pressure that Intel experienced, and the adverse economic realities that forced GlobalFoundries to exit. Historical experience shows that this double risk superposition is particularly likely to lead to value destruction, and this consequence often becomes apparent only after capital investment is completed.