Tesla has shut down its Dojo supercomputer and custom AI-chip program, according to internal notices dated 11–12 Aug. The move dissolves a multiyear effort to develop proprietary hardware for training the company’s Full Self-Driving software and reallocates dozens of semiconductor and firmware engineers to other teams. Chief Executive Officer Elon Musk decided to end Dojo after determining it would be inefficient to fund two distinct chip architectures, people familiar with the matter said. Tesla will instead expand its reliance on external suppliers, including Nvidia and AMD for compute chips and Samsung for fabrication. The strategic reversal coincides with continued turnover in Tesla’s leadership ranks. Piero Landolfi, director of service for North America, exited the electric-vehicle maker on 11 Aug. after almost nine years, becoming at least the tenth senior executive to depart this year. Earlier departures include heads of sales, battery architecture and software engineering. The dismantling of Dojo and the executive exodus come as Tesla works through a slump in deliveries and profit margins. Musk has cautioned investors that the company faces "a few rough quarters" before it begins to generate material revenue from self-driving software and related services.