Samsung Electronics has begun turning its $16.5 billion Tesla foundry contract into silicon, but investors still want clear evidence that the deal can repair its chipmaking economics.
Samsung started trial production of Tesla’s AI5 processor on 2-nanometre lines at Taylor, Texas this week, with yield verification under way before full-scale supply next year.
The contract dates to July 2025 and the trial production is the fresh catalyst.
Samsung shares were only modestly higher on Wednesday after falling 4.05% on Monday and another 0.2% on Tuesday. That muted response reflects execution risk rather than indifference to Tesla.
Tesla now has to validate Samsung’s 2nm execution
Samsung’s Taylor plant has begun producing 2nm wafers for Tesla’s AI5, a processor expected to power vehicles, Cybercab, Optimus and AI data-centre applications.
Samsung aims to complete mass-production verification by year-end before ramping supply next year.
Samsung’s foundry problem has rarely been ambition. It has been lower yields, underused capacity and weaker profitability than TSMC.
TrendForce put Samsung’s global foundry share at 5.9% in the second quarter, only slightly ahead of SMIC at 5.4%, while TSMC controlled 72.5%.
KB Securities analyst Kim Dong-won told ChosunBiz that Samsung’s 2nm GAA yield has improved from below 50% at the start of the year to above 70%.
“Starting in the third quarter this year, the foundry business is likely to swing to profit,” Kim said.
Tesla now needs to prove that those yield gains translate into reliable, profitable mass production.
Bigger rerating needs customers beyond Tesla
The $16.5 billion Tesla agreement is one of Samsung’s most important foundry wins, but a lasting rerating requires repeatability.
One marquee customer can lift Taylor utilization and validate Samsung’s leading-edge manufacturing.
Several major customers would show that its 2nm platform is competitive rather than dependent on one flagship programme.
Mirae Asset analyst Young-gun Kim argues that Tesla and other advanced-node wins validate Samsung’s ability to manufacture data-centre-grade accelerators. Mirae values the foundry operation at about 138 trillion won and says its previous valuation discount is fading.
There is also a potential industry opening.
TrendForce says TSMC’s advanced 3nm, 4nm and 5nm capacity remains heavily utilized by AI demand, creating room for an alternative supplier if Samsung can deliver consistently.
That makes Tesla less the end of Samsung’s turnaround story than a test case for winning the next large customer.
Memory still matters far more to Samsung profits
The complication for shareholders is that Samsung is not a pure foundry stock.
Most of its current earnings momentum still comes from DRAM, NAND and HBM.
Investors can therefore become more optimistic about Tesla while simultaneously worrying that the much larger memory cycle is approaching a peak.
BNK downgraded Samsung to Hold and cut its target price to 270,000 won from 300,000 won, arguing that aggressive industry capacity expansion could eventually soften memory pricing.
That helps explain why Tesla trial production has not triggered the share-price reaction seen when the contract was announced in 2025.
KB Securities may be right that foundry profitability is improving, while BNK may also be right that memory momentum is becoming less favourable.
Tesla has already solved one Samsung problem: finding a marquee customer willing to bet on its advanced process.
The post Samsung stock is ignoring a $16.5 billion Tesla catalyst: should investors worry appeared first on Invezz
