September looms as a risk month for stocks, Yardeni says
Investing.com -- Tesla is reportedly scaling back its ambitious in-house custom silicon compute project, Dojo, shifting focus towards partnerships with suppliers like NVIDIA for inference capabilities.
Bloomberg reported that Tesla (NASDAQ:TSLA) is disbanding its Dojo supercomputer team, with its leader, Peter Bannon, set to depart.
Sources reportedly said CEO Elon Musk ordered the shutdown as the company leans more on external partners, including Nvidia (NASDAQ:NVDA), AMD (NASDAQ:AMD), and Samsung (KS:005930), for chip manufacturing.
Morgan Stanley (NYSE:MS) noted that the sources also suggested some Dojo team members have joined the AI startup DensityAI, while others will be reassigned internally.
In its note reacting to the report, the bank shared five key thoughts:
-
Tesla’s “DOGE-like efficiency drive” will help reduce costs, said Morgan Stanley. The bank explained that Tesla’s Q2 results showed higher operating expenses due to AI projects and expanded AI training compute, with FY25 capex guidance over $9 billion, mainly on “AI-related capital expenditures.”
-
DensityAI, led by former Tesla AI head Ganesh Venkataramanan, is continuing Dojo’s mission in stealth mode, stated Morgan Stanley.
-
The bank believes Elon Musk’s xAI is expanding AI “brain” development, using data from X (formerly Twitter) and Tesla’s real-world vision data. The analysts add that scaling back Dojo could boost collaboration between Tesla and xAI.
-
Tesla is said to be shifting emphasis to robotics and inference capability at the edge, leveraging its vehicle fleet as a “massive distributed inference cloud,” said Morgan Stanley. Incremental capex and R&D will focus on improving robot BOM cost and manufacturability ahead of Optimus commercialization.
-
Finally, the bank notes that the GPU compute shortage is less acute than two years ago. They believe Tesla remains a strong Nvidia customer, but supply-demand balances have improved.