Tesla taps Samsung for $16.5B next-gen AI chip deal

Tesla just dropped $16.5 billion to lock down its next generation of custom silicon.

According to an announcement by Elon Musk on X, Tesla signed a massive deal with Samsung to produce its upcoming AI6 processor (also known as Hardware 6). The chips will be made exclusively at Samsung’s new mega-fab currently under construction in Taylor, Texas.

Musk called the strategic value of the deal hard to overstate. He also noted that $16.5 billion is basically just a baseline, adding that actual spending and chip output will likely be several times higher.

One chip for cars, bots, and data centers

Tesla isn’t designing AI6 just for car dashboards. The chip is meant to be a unified silicon architecture capable of scaling across three completely different workloads: FSD driver assistance, Optimus humanoid robots, and high-performance AI training inside Tesla’s data centers.

Musk gave a quick roadmap update alongside the announcement. The preceding AI5 chip, which just completed tape-out, is handled by TSMC. TSMC will initially manufacture AI5 in Taiwan before shifting production to its Arizona plant. Prior chips, including AI4 and the original Samsung-built Hardware 3 launched in 2019, paved the road for this custom silicon push.

Musk also confirmed that Samsung agreed to let Tesla step inside the fab to help optimize efficiency. In typical Musk fashion, he noted he plans to walk the production line personally, helped by the fact that the Taylor fab sits conveniently close to his Texas home.

For Samsung, the contract is a massive lifeline. Its foundry division has been bleeding billions after losing major tech clients to TSMC, so landing Tesla’s next-gen silicon puts Samsung right back into the advanced manufacturing conversation.

Why it matters

Tesla’s mega-deal marks a major shift in how AI hardware empires are being built.

First, keeping production in Texas alongside Tesla’s primary operations cuts out serious geopolitical and logistics headaches. Having the foundry practically in Musk’s backyard gives Tesla immediate feedback loops during pilot runs.

Second, the move is a clever hedge against TSMC. Relying solely on TSMC puts any company in direct competition with Apple and Nvidia for fab allocation. By pairing AI5 with TSMC and committing AI6 to Samsung, Tesla secures dual-source leverage and protects its hardware roadmap.

Finally, unifying cars, humanoid robots, and data centers on a single internal architecture is a massive competitive threat to merchant silicon vendors. If Tesla proves an all-in-one chip design works at scale, it will pressure every automotive and robotics rival still paying Nvidia markups. The obvious risk? Samsung’s advanced node yields have notoriously lagged behind TSMC. Elon Musk walking the factory floor might accelerate culture, but fixing semiconductor physics and wafer yields is an entirely different beast.