Tesla Is Developing Its Own Custom AI Chips To Drive Level 5 Autonomous EVs

Tesla is indeed developing custom AI semiconductors that will help with autonomous driving chores, company CEO and founder Elon Musk has finally confirmed. It is not entirely clear if Tesla is working with AMD on specialized hardware for its electric vehicles, though Musk did offer up that Jim Keller, a microprocessor engineer who previously worked for AMD and Apple, is working on the hardware for Tesla.

Keller joined Tesla in 2016. He is the automaker's President of Autopilot Hardware and Software Engineering, and presumably is the one who is in charge of the company's effort to build to a custom AI chip. Previously he spearheaded both AMD's K8 and Apple's A-Series chip efforts. When Tesla brought him on board, the company said he would help develop the safest, most advanced autopilot systems in the world.
model 3 1
Musk reiterated that mission at the intelligence conference NIPS, telling attendees that "Tesla is serious about AI, both on the software and hardware fronts," and that "Jim is developing specialized AI hardware that we think will be the best in the world." In other words, Tesla is not only interested in designing its own chips rather than leaning on third parties, but it wants to set a high bar.

Not that any of this should be surprising. Tesla has been going in this direction for some time now. At the beginning of the year, Tesla managed to poach the former boss of Apple's Swift programming language, Chris Lattner. He joined Tesla as Vice President of Autopilot Software, taking over for Jinnah Hosein, who served in the role on a interim basis before going back to his full-time duties at SpaceX.

"Working with many phenomenal teams at Apple to launch Swift has been a unique life experience," Lattner said in a posting to Swift-Evolution at the time. "Apple is a truly amazing place to be able to assemble the skills, imagination, and discipline to pull something like this off. Swift is in great shape today, and Swift 4 will be a really strong release with Ted as the Project Lead."

Unfortunately Musk did not go into many details about its custom hardware. For example, in addition to not knowing if its is working with AMD on chip design, we also do not know if Tesla plans to replace NVIDIA's hardware in its autonomous vehicles, or continue using it alongside its own. We suspect the former, though there is certainly room for multiple slices of silicon in Tesla's increasingly sophisticated vehicles.

Tesla's ultimate goal is to achieve Level 5 autonomy, based on SAE International's standard. Level 5 represents full autonomation, with a driverless car being able to navigate any road in any driving condition that a human driver could.