Tesla has lost the technical program manager behind its Cortex supercomputer in Texas to OpenAI.

The latest example of a massive talent exodus at Tesla over the last year.

Tesla has heavily invested in large data centers to house supercomputing power for training its neural networks to achieve self-driving capabilities.

Its most recent facility is a supercomputer called ‘Cortex‘ inside Gigafactory Texas in Austin. Tesla plans for the system to support 500 MW of power in the long term, but the initial phase is approximately 100 MW.

The project suffered a few delays, and Elon Musk even redirected NVIDIA computers that were supposed to be used for the super cluster to xAI, an AI startup under his control.

A source familiar with the project told Electrek that Tesla brought in Adam Wilson, a data center design engineer from Meta, to lead the technical program last year.

Now, we learn that Wilson left Tesla to lead Data Center Design for OpenAI.

OpenAI is sort of a nemesis to Tesla CEO Elon Musk.

Musk originally co-founded OpenAI, but he left in 2018, citing a conflict of interest due to Tesla’s focus on AI and the potential of competing for talent.

As evidenced by employee movements between Tesla and OpenAI, including now Wilson, this appears to have been the right call.

However, Musk ended up also founding xAI to more directly compete with OpenAI after threatening Tesla shareholders that he would not build AI products at Tesla unless he gained 25% control over the company’s shares, a level of control he ultimately lost after selling shares to acquire Twitter.

We have constructed a concrete timeline of Musk’s shifting attempts to gain control over AI, first at OpenAI, then at Tesla, and now at xAI.

Now, Wilson is going to help OpenAI deploy data centers through its “Stargate” project, which plans to deploy hundreds of billions of dollars worth of computing power across the US to power OpenAI’s AI products.

Tesla also needs to build many more large data centers to train its planned self-driving vehicles.

Progress on its consumer “Supervised self-driving” has stalled for months as Tesla focuses on training for a localized version of the software for the ride-hailing fleet it plans to launch in Austin in June.

Some experts believe that Tesla would need to 10x its training compute before it can achieve level 4 or 5 autonomy based on its current approach.