Samsung lags behind TSMC in chips. So why is Musk making a $16.5B bet on it?

[NEWS ANALYSIS]

Elon Musk’s recent decision to tap Samsung Electronics for AI5 and AI6 chips for Tesla has come as a surprise, given the lingering technology gap in custom fabrication between the Korean chip giant and Taiwan’s TSMC.

The Tesla CEO cited a surge in demand for high-performing AI chips essential in self-driving cars, humanoid robots and data centers, but industry insiders say he is ultimately pursuing an end-to-end chip supply chain spanning multiple vendors to secure cost advantages and tighter control.

“The only option is to go build some, like, very big chip fab,” Musk said at a recent shareholder meeting on November 7 when asked about plans for Tesla’s own chip factory. “And then you go to solve memory and packaging too.”

That vision is some way off, however, as he has only floated Tesla’s own chip factory without any indication of concrete plans or a timeline, requiring him to expand orders with Samsung.

It is a shift from the AI5 chip production trajectory Musk discussed earlier when he indicated that Taiwan’s TSMC, the world’s largest contract chipmaker, would exclusively handle production. Tesla will instead adopt a dual-track supply chain, splitting orders between TSMC and Samsung to stabilize production and accelerate output.

“This is a beautiful chip,” Musk said in an earnings call after announcing third quarter results on October 22. “I’ve poured so much life energy into this chip personally. I’m confident this is going to be a winner, next level. It makes sense to have both Samsung and TSMC focus on AI5.”

According to Musk, AI5 chip samples will roll out next year, with mass production targeted for 2027, followed by AI6 in 2028. The contract between Tesla and Samsung is valued at $16.5 billion for now, but Musk’s ambition to build Tesla’s AI chip ecosystem could see that figure rise sharply.


Musk’s endgame: ‘Gigantic chip fab’ vision
Despite dual sourcing from Samsung and TSMC, Musk insists the capacity will not be enough.

“It’s basically a money-printing machine for TSMC and Samsung,” he said. “The faster they make the chips, the faster we send them money — but it’s still not fast enough.”

Musk’s eventual goal is to build a Tesla-owned mega foundry, consolidating chip design, manufacturing, memory and packaging under one roof — a facility capable of producing up to one million wafers per month.

His appetite for high-end chip supply comes at a critical juncture of his expansion of autonomous driving across the United States, as well as Tesla’s Optimus robots, with third-generation mass production slated to begin late 2026.

Analysts anticipate Tesla’s value to be driven by advanced autonomous driving technologies and its humanoid robots, and in order for that to become a reality, they must be powered by the latest AI chips suited for Tesla software.

Tesla noted that full self-driving (FSD) version 15 will represent the company’s biggest upgrade since version 12, integrating much of the robotaxi-grade FSD model into the consumer version of the Optimus V3, which is expected to show vast improvements from prior generations, able to be dispatched to industry manufacturing and general purpose commodities spanning household chores, logistics and health care.

“To ensure the stable implementation of such performance improvements, large-scale AI infrastructure investment is essential,” said Lee Ji-soo, an analyst at Korea Investment & Securities. “Tesla said it has expanded its AI training infrastructure to the equivalent of 81,000 H100 processors, which proves its focused commitment to improving computing capacity.”

Shareholders recently approved a $1 trillion stock option compensation plan for Musk, tied to milestones such as cumulative sales of 20 million EVs, 1 million robotaxis and 1 million humanoids — a vision that entails a demand for tens of millions of AI chips in the years ahead.

Samsung: Tesla’s new chip guru?
Musk’s past comments about Samsung’s Taylor plant suggest he intends to be closely involved in overseeing the chip manufacturing process — a level of access not possible with TSMC, which still holds an edge over the Korean firm in process maturity and yield rates.

For TSMC, its top-tier clients are Nvidia and Apple. Tech trade presses estimate that combined orders from those two firms represent about 40 percent of TSMC’s total annual revenue based on projections for 2025, while Tesla claims a far lower portion in the low-to-mid single digits, limiting Tesla’s influence over the company if Musk wants to focus on hands-on production.

“If AI6 reaches mass production smoothly, chip design and manufacturing could become a core competitive advantage across Elon Musk’s businesses — enabling greater flexibility and lower costs,” said Ming-chi Kuo, a semiconductor analyst at Taiwan’s TF International, on X. “While Samsung may not fully catch up with TSMC in advanced nodes, it has at least discovered a new business model that actively involves customers in the manufacturing process.”

BY LEE JAE-LIM [lee.jaelim@joongang.co.kr]