— Forward Cap (@forwardcap) August 7, 2025 Dojo’s pitch was straightforward: purpose-built silicon would process vast video data from Tesla’s global fleet faster and cheaper than off-the-shelf GPUs, accelerating FSD progress and enabling humanoid robot “Optimus” to learn in simulation. Losing that edge now adds a hurdle for Tesla to prove it can still achieve said goals. With Dojo shelved, Tesla appears set to lean harder on external suppliers. Just last week, Samsung Electronics inked a US$16.5 billion agreement to provide custom AI chips to the automaker. Tesla already relies on Nvidia and AMD for training clusters, and abandoning homegrown hardware may blunt capex but exposes the firm to tight supply and pricing risk in the broader AI accelerator market. Tesla ended Q2 with roughly $22 billion in cash, but free cash flow has been lumpy as the company juggles factory expansions, the Cybertruck rollout, and now multibillion-dollar AI chip commitments. This development notably comes after Tesla board awarded Musk a new $29 billion compensation package, even as his previous record-breaking pay deal remains tied up in Delaware courts. Entire DOJO team quits and Musk says it will license all $TSLA Tesla Ai from xAI after $30 billion pay package. It will cost Tesla $5 billion for 1% ownership, meaning all Ai profits of FSD and Optimus become property of xAI through licensing. https://t.co/Nrs0M8Tyrm pic.twitter.com/OVhBjAqmrc