- Tesla’s AI focus on robotaxis and autonomous driving, what Musk refers to as “real-world AI,” means it doesn’t risk being disrupted by DeepSeek chatbot R1, argue analysts at investment bank Piper Sandler. If anything, potentially less demand for Nvidia’s AI training chips could be a benefit.
When DeepSeek’s impact obliterated a trillion dollars of value collectively from U.S. equity markets, artificial intelligence stocks found themselves at ground zero.
But analysts at Piper Sandler argue Tesla unjustly fell victim to Monday’s bloodbath, which ensued after investors—seeking safety from the surprisingly adept Chinese AI chatbot—dumped overboard shares in anything vaguely related to Nvidia.
“We don’t think DeepSeek represents a threat,” the investment bank told clients in a research note on Monday.
That’s because Elon Musk built a competitive moat around Tesla that makes it impervious to what venture capitalist Marc Andreessen labeled AI’s Sputnik moment. That bulwark is called robotaxis, self-driving cars that Musk likes to call “real-world AI.”
If anything, the effect on Tesla’s business would be neutral at worst, and quite likely even positive, should DeepSeek’s more efficient AI model cool demand and therefore the cost of Nvidia’s most advanced graphics processing units Musk purchases to train his AI.
“As a consumer of GPUs,” the bank wrote, “Tesla may actually benefit from this trend.”
Musk has been gradually repositioning Tesla as an AI play
It’s not hard to see why the world’s largest EV manufacturer made for a logical target of Monday’s selloff.
Musk is one of the most important customers of Nvidia. The serial entrepreneur’s latest startup, xAI, is the lab behind Grok, the chatbot used on Musk’s social network, X. This artificially intelligent chatbot competes directly with DeepSeek’s reasoning model, R1, which has taken the world by storm.
Musk has spent billions of dollars on Nvidia CEO Jensen Huang’s signature H100 chips for AI training, and not just for xAI. The compute also powers the new “Cortex” data center that Tesla has situated directly next to its vehicle production site in Austin.
Musk has also spent all of last year gradually repositioning investor perception of Tesla as an AI-first company, with a captive car business to boot.
Almost all AI stocks hit—but Tesla isn’t in the chatbot business
Tesla shares sank 2.3% in Monday’s trading, unusual given investors tend not to position themselves all too heavily going into earnings week. Tesla reports fourth-quarter results after the close of trading on Wednesday.
Granted, this drop was substantially better than the historic $600 billion wipeout suffered by Nvidia, or the plunge in data-center cooling equipment supplier Vertiv, down 30% in a single session.
The 4.2% drop in Google AI parent Alphabet was also brutal for a $2.3 trillion tech leader that still earns the bulk of its money from ads on its dominant internet search engine. A rare exception to the selloff was Meta, likely because it, too, relies on open-source rather than closed-source AI models like Google’s Gemini.
But while R1 may hurt Musk’s ability to raise funding for xAI, his separately held carmaker is itself not at risk. That’s because Tesla’s not interested, first and foremost, in intelligent chatbots like DeepSeek’s R1.
Treasure trove of camera-based data so FSD can learn how to drive
While it’s safe to say that one day a large language model like R1 or Musk’s own Grok will eventually be integrated into a Tesla as a form of virtual personal assistant, the real problem Tesla aims to solve is unsupervised autonomous driving. This requires feeding its Cortex data center petabytes of video captured by its cars’ cameras in order to train its AI model responsible for Full Self-Driving.
FSD is now fully based on AI inference, meaning the vehicle doesn’t rely on hard-coded commands when executing driving maneuvers. Everything is calculated directly on Tesla’s proprietary silicon in its FSD computer, now called AI4 instead of the previous nomenclature Hardware 4 (HW4).
“Instead of using the internet as a trove of text-based data to train chatbots, Tesla is using vehicles as a trove of camera-based data to train robots,” Piper Sandler wrote. “And in this regard, we think Chinese auto companies are laggards.”
There’s a catch, however. Investors expect tangible progress: The top ranked question that Tesla will have to answer on Wednesday’s Q4 investor call is whether it will live up to its promise to launch unsupervised driving in Texas and California this year.
Shares in the company, which didn’t reply to Fortune’s request for comment, are trading down 0.8% on Tuesday, underperforming the moderate rebound in the overall tech market.
