Tesla's "Chip-Making" is Coming, Will It Challenge TSMC?

This marks another major expansion for Tesla beyond its core electric vehicle business, launching a large-scale, costly project.

Tesla’s self-built chip factory project, “Terafab,” is set to commence soon. On March 14th, U.S. time, Tesla CEO Elon Musk disclosed this information on social media, stating that in “seven days,” they will launch the massive chip factory plan Terafab, aiming to produce AI and autonomous driving chips in-house.

This will signify another significant expansion beyond Tesla’s core electric vehicle business, initiating a large-scale, expensive project.

Tesla “Chip Shortage”

Currently, advanced chip manufacturing is highly concentrated among a few foundries like TSMC, Samsung, and Intel. Tesla is sourcing AI chips from companies like NVIDIA while also pushing forward with internal R&D, having signed manufacturing agreements with TSMC and Samsung.

Tesla is currently designing its fifth-generation AI chip (AI5), planned for application in 2027 to support its autonomous driving initiatives. Musk previously previewed this chip and expressed concerns about supplier capacity.

“Even if we assume the best-case scenario from our chip suppliers, production still won’t be enough,” Musk said at Tesla’s 2025 annual shareholder meeting. “So I think we may have to build a ‘Terafab.’ It’s like a ‘Gigafactory,’ but much larger. This is something we must do.”

He pointed out that the factory would not only cover AI system chips but also storage chips, which are likely to become a supply bottleneck. Therefore, an integrated production line covering logic, storage, and packaging is necessary to ensure chip supply for the next three to four years.

In January this year, Musk further stated during Tesla’s earnings call that existing partners Samsung, Micron, and TSMC “cannot meet Tesla’s demand three to four years from now” in terms of capacity and supply levels. He added that to eliminate this “likely bottleneck,” Tesla must build its own Terafab-level wafer factory to gain control over key AI and automotive chip capacity.

According to Musk, the goal of Terafab is to produce 100 billion to 200 billion chips annually. If achieved, it could become one of the world’s largest chip factories and potentially one of the largest-scale chip manufacturing projects globally.

Musk also questioned the long-standing cleanroom standards in semiconductor manufacturing. In an interview, he said he wants to build a 2-nanometer chip factory where workers can “smoke cigars and eat burgers.”

Additionally, Musk did not disclose details about Terafab’s location, investment amount, or partners. Little is known about the project externally.

Speculation suggests one possible model for Terafab is that Tesla might sign technology licensing agreements with chip manufacturers like TSMC and Intel, with Tesla funding the establishment of production lines. Since TSMC is open to pre-booked capacity collaborations, this could be one pathway for Terafab’s realization.

Furthermore, as a U.S.-based company, Tesla’s collaboration with Intel is also possible, and Musk has publicly mentioned this direction.

However, entering chip manufacturing entails significant capital expenditure and operational costs. Building an advanced process chip factory can cost billions or even hundreds of billions of dollars, facing challenges such as rapid technological iteration, long yield ramp-up cycles, and tight supply of high-end talent. Especially in a landscape where TSMC, Samsung, and others have highly mature foundry models, whether Tesla can achieve cost and efficiency advantages through self-building remains to be seen.

AI-driven Growth in Foundry Industry

Currently, the chip manufacturing industry Musk aims to enter is being reshaped by the AI wave.

A report from TrendForce on March 12th shows that in Q4 2025, the top ten global foundries will collectively generate $46.3 billion in revenue, a 2.6% quarter-over-quarter increase, with annual revenue reaching approximately $169.5 billion, a 26.3% year-over-year growth, setting a new record.

Driven by AI demand, the storage chip market is experiencing a super cycle, affecting upstream wafer foundries, midstream chip manufacturers, and downstream terminal vendors.

TrendForce states that in advanced processes, demand for AI server GPUs and Google TPUs remains strong, coupled with flagship chips for new smartphones, leading to robust shipment performance. In mature processes, server and edge AI power management orders support high utilization of 8-inch lines, with some even considering price hikes. Meanwhile, 12-inch capacity utilization remains roughly stable.

Among leading foundries, TSMC’s quarterly revenue grew 2% quarter-over-quarter to $33.7 billion, holding a 70.4% market share, remaining the leader. Despite a slight decrease in overall wafer shipments, flagship AP chips for iPhone 17 drove demand for 3nm wafers, boosting overall average selling prices (ASP).

In January, TSMC announced that its 2025 capital expenditure would reach a record high of $52–56 billion, about 30% higher than the $40.9 billion spent in 2024, further increasing investment.

TSMC Chairman and CEO Wei Zhejia stated during the earnings call that in 2025, revenue from AI chips will account for a high-teens percentage of total revenue, and the foundry industry is expected to grow 14% year-over-year in 2026, with TSMC surpassing that, expecting over 30% growth in revenue.

Samsung benefited from shipments of its 2nm new products and the start of HBM4 logic chip production. Its foundry revenue this quarter increased 6.7% quarter-over-quarter to nearly $3.4 billion, turning profitable, with market share rising from 6.8% to 7.1%, ranking second.

SMIC’s Q4 revenue grew 4.5% quarter-over-quarter to nearly $2.49 billion, driven by increased wafer shipments, slight ASP increases, and higher mask shipments at year-end, ranking third.

SMIC Co-CEO Zhao Haijun noted that AI’s strong demand for storage products has squeezed supply for mid- and low-end mobile devices, leading manufacturers to reduce production of these products. Meanwhile, demand for AI, storage, and mid-to-high-end applications continues to grow.

Rising AI demand is further pushing up wafer foundry prices. Zhao said the company adjusts pricing based on supply and demand: when upstream shortages occur and mainstream suppliers raise prices, SMIC follows suit. He specifically pointed out that rising storage prices have driven higher foundry quotes, with power and power-specific processes like BCD in high demand from AI, data centers, and automotive scenarios. Additionally, continuous improvements in product yield and quality support price increases for certain categories.