Over the weekend, Tesla (TSLA) just delivered a rare double whammy to Nvidia (NVDA).
CEO Elon Musk reveals Tesla’s much-anticipated AI5 autonomous driving chip It is almost complete and the next project, A16, is already in progress.
Speaking of the AI reasoning part, Musk said on X on Sunday: Dojo 3 is restarting, pushing Tesla to reinvest in large-scale artificial intelligence training after previously withdrawing.
However, when NVIDIA launched “alpamayo” at CES 2026 (Open Source Autonomous Vehicle AI Toolkit), aiming to become Default autonomous platform Powering numerous brands.
Musk reacted quickly, downplaying the risks.
Obviously, this is a very interesting time for the self-driving industry, with the tug of war between the two giants Nvidia and Tesla.
For Tesla, it’s all about building a closed loop that covers the entire AV stack.
-
In-vehicle computing designed by Tesla (This includes AI5 which is “nearly completed” and AI6 which is already in progress)
-
Tesla’s camera-first software stack
-
Tesla’s data flywheel is powered by its own fleet of vehicles
So for Tesla, it’s all about keeping the self-driving part within its robust ecosystem, while Nvidia wants to power everyone else.
For investors, these promises are nothing new, which makes follow-up all the more important.
Tesla is looking to tighten control over its self-driving hardware.
More tech stocks:
Musk announced in an X post on Saturday that the electric car giant is close to completing its AI5 self-driving computer chip, while AI6 is already in development.
Musk said that AI5 chips produced by TSMC will enter the market Achieve mass production in 2027replace the AI4 hardware. In addition, Tesla also cooperates with Samsung Electronics to produce chips in the United States.
It’s easy to get lost in AI terminology, so it’s important to be clear about what’s happening every step of the way.
RELATED: BlackRock CEO issues blunt warning on U.S. Treasuries
The A15 and A16 initiatives are essentially about “edge reasoning.” This is basically how it works Tesla’s fully autonomous driving neural network in the car instead of relying on third-party computing stacks.
So if Tesla ran the software on its own chips, it would gain a major competitive advantage:
-
Tesla’s cars don’t need Nvidia’s in-vehicle SoC (or its full “DRIVE” platform).
-
Tesla gained control over unit costs, supply chain leverage, and chip design.
It’s worth noting, though, that Tesla has abandoned Nvidia because of its In-vehicle computing in 2019so the latest move is more of a doubling down than a transformation.
NVIDIA Providing automakers with a full-stack solution is essentially a shortcut to full autonomy. below NVIDIA drives It’s basically selling an integrated “brain, operating system and toolkit.”
RELATED: Goldman Sachs adjusts Microsoft stock price target ahead of earnings release
Therefore, instead of building custom chips, software, security frameworks, etc., automakers can simply plug into NVIDIA’s powerful ecosystem and get started. A big part of its appeal is that it’s essentially a hack for companies that don’t have Tesla’s decade-long autonomous efforts or billions of dollars in research and development.
What Nvidia bundles:
-
Drive AGX On-board computers, e.g. Olin and Thor.
-
A complete software stack including driver operating system and drive factory.
-
Piloting Hyperiona reference vehicle platform equipped with proven sensors and architecture.
-
Popular security and verification tools NVIDIA Halo Umbrella, as well as powerful artificial intelligence models such as alpamayo,accelerate training and simulation.
Tesla is improving its in-car AI chips, but it’s clear that Nvidia still has a key advantage in computing power.
AI5 and AI6 are tailor-made for edge inference, but Training cutting-edge scale models It’s a completely different challenge.
Training modern artificial intelligence systems is extremely demanding on computing resources.
Longer term, Meta said it trained its AI model Llama 3.1 (405B) using the following methods: Over 16,000 Nvidia H100 GPUs. So if we take into account 700 watts per chip, this is almost 11.2 MW power Only available on GPU. This level of scale is where Nvidia’s economics, availability, and ecosystem continue to dominate.
However, Tesla decided to reboot Dojo 3 because it wanted to get into the training game again.
But at this moment, I feel Dojo 3 returns Most likely points to a hybrid future.
Tesla will continue to use AI5 and AI6 architecture to enhance its training capabilities, while still relying on Nvidia where scale and economics are important.
The race for training really escalates when we see strong evidence of large-scale training clusters running on Tesla chips, supported by throughput and cost data.
RELATED: Morgan Stanley adjusts Rocket Lab stock price target after rally
This article was originally published by TheStreet on January 19, 2026, and first appeared in the Technology section. Click here to add TheStreet as your preferred source.