At the 2025 Consumer Electronics Show, Nvidia showed off its new consumer graphics cards, home-scale compute machines, and commercial AI offerings. One of these offerings included the new Nvidia Cosmos training system.
Nvidia is a close partner of Tesla - in fact, they produce and supply the GPUs that Tesla uses to train FSD - the H100s and soon-to-be H200s, located at the new Cortex Supercomputing Cluster at Giga Texas. Nvidia will also challenge Tesla’s lead in developing and deploying synthetic training data for an autonomous driving system - something Tesla is already doing.
However, this is far more important for other manufacturers. We’re going to take a look at what Nvidia is offering and how it compares to what Tesla is already doing. We’ve done a few deep dives into how Tesla’s FSD works, how Tesla streamlines FSD, and, more recently, how they optimize FSD. If you want to get familiar with a bit of the lingo and the background knowledge, we recommend reading those articles before continuing, but we’ll do our best to explain how all this synthetic data works.
Nvidia Cosmos
Nvidia’s Cosmos is a generative AI model created to accelerate the development of physical AI systems, including robots and autonomous vehicles. Remember - Tesla’s FSD is also the same software that powers their humanoid robot, Optimus. Nvidia is aiming to tackle physical, real-world deployments of AI anywhere from your home, your street, or your workplace, just like Tesla.
Cosmos is a physics-aware engine that learns from real-world video and builds simulated video inputs. It tokenizes data to help AI systems learn quicker, all based on the video that is input into the system. Sound familiar? That’s exactly how FSD learns as well.
Cosmos also has the capability to do sensor-fused simulations. That means it can take multiple input sources - video, LiDAR, audio, or whatever else the user intends, and fuse them together into a single-world simulation for your AI model to learn from. This helps train, test, and validate autonomous vehicle behavior in a safe, synthetic format while also providing a massive breadth of data.
Data Scaling
Of course, Cosmos itself still requires video input - the more video you feed it, the more simulations it can generate and run. Data scaling is a necessity for AI applications, as you’ll need to feed it an infinite amount of data to build an infinite amount of scenarios for it to train itself on.
Synthetic data also has a problem - is it real? Can it predict real-world situations? In early 2024, Elon Musk commented on this problem, noting that data scales infinitely both in the real world and in simulated data. A better way to gather testing data is through real-world data. After all, no AI can predict the real world just yet - in fact, that’s an excellent quantum computing problem that the brightest minds are working on.
Yun-Ta Tsai, an engineer at Tesla’s AI team, also mentioned that writing code or generating scenarios doesn’t cover what even the wildest AI hallucinations might come up with. There are lots of optical phenomena and real-world situations that don’t necessarily make sense in the rigid training sets that AI would develop, so real-world data is absolutely essential to build a system that can actually train a useful real-world AI.
Tesla has billions of miles of real-world video that can be used for training, according to Tesla’s Social Media Team Lead Viv. This much data is essential because even today, FSD encounters “edge cases” that can confuse it, slow it down, or render it incapable of continuing, throwing up the dreaded red hands telling the user to take over.
Cosmos was trained on approximately 20 million hours of footage, including human activities like walking and manipulating objects. On the other hand, Tesla’s fleet gathers approximately 2,380 recorded minutes of real-world video per minute. Every 140 hours - just shy of 6 days - Tesla’s fleet gathers 20 million hours of footage. That was a little bit of back-of-the-napkin math, calculated at 60 mph as the average speed.
Generative Worlds
Both Tesla’s FSD and Nvidia’s Cosmos can generate highly realistic, physics-based worlds. These worlds are life-like environments and simulate the movement of people and traffic and the real-life position of obstacles and objects, including curbs, fences, buildings, and other objects.
Tesla uses a combination of real-world data and synthetic data, but the combination of data is heavily weighted to real-world data. Meanwhile, companies who use Cosmos will be weighting their data heavily towards synthetically created situations, drastically limiting what kind of cases they may see in their training datasets.
As such, while generative worlds may be useful to validate an AI quickly, we would argue that these worlds aren’t as useful as real-world data to do the training of an AI.
Overall, Cosmos is an exciting step - others are clearly following in Tesla’s footsteps, but they’re extremely far behind in real-world data. Tesla has built a massive first-mover advantage in AI and autonomy, and others are now playing catch-up.
We’re excited to see how Tesla’s future deployment of its Dojo Supercomputer for Data Labelling adds to its pre-existing lead, and how Cortex will be able to expand, as well as what competitors are going to be bringing to the table. After all, competition breeds innovation - and that’s how Tesla innovated in the EV space to begin with.
Subscribe
Subscribe to our newsletter to stay up to date on the latest Tesla news, upcoming features and software updates.
There has been extensive reporting on what the long-anticipated affordable EV option would become, and we’ve seen numerous news stories mentioning that it was delayed or even arriving on time. The executive team revealed the near-term and long-term plans for Tesla’s upcoming vehicles, and there’s definitely good news to share here.
Updated Timeline
The most important piece of news is that despite all the talk of delays from supposed inside sources, Tesla has confirmed that the plan for its more affordable model remains on schedule for production to begin in the first half of 2025. Tesla’s executive team narrowed that timeline down further - and said that they expect production to kick off as soon as June and that the new model will be in the market shortly thereafter.
While the production timeline itself is on track, Tesla did note that the subsequent ramping process will likely be slower than initially hoped, citing global tariff and financial impacts as challenges to overcome to prepare its production lines.
Hybrid Production Approach
Tesla has once again confirmed that this will not be their next-generation vehicle, built using new production methods. Instead, they outlined a relatively more pragmatic approach for this new model.
Tesla will utilize aspects of both the next-generation platform as well as some parts of its current platforms (namely the Model 3 and Model Y). This means that Tesla will produce this new vehicle on the same manufacturing lines as the Model 3 and Model Y.
This strategy allows Tesla to bring the vehicle to the market more quickly, while also managing capital expenditures more efficiently by using existing infrastructure. However, Tesla’s executive team also noted that this approach, while faster, will result in fewer cost reductions than what might have been achieved with an entirely new platform and dedicated manufacturing process.
Vehicle Design
Using these existing production lines means that Tesla’s new vehicle will likely share some considerable similarities with either the existing Model 3 or Model Y. Rather than being a radically different and smaller vehicle, this new model will resemble the overall form factor and shape of Tesla’s current core offerings, while being optimized for a lower cost.
This doesn’t mean that Tesla is forgetting the overall goal here. Their ultimate goal is to reduce the initial cost of ownership and lower monthly payments for customers while maintaining a standard of excellence and safety.
Not Unboxed
Both at this Earnings Call and previous ones, Tesla has indicated that this new vehicle will not be using the innovative unboxed assembly method, at least for the time being. That relatively unique method will be developed and implemented specifically for the purpose-built Cybercab and for future vehicles on the next-generation platform.
We’re just a few days away from May, so it won’t be long before we see more about this upcoming vehicle. Stay tuned.
In a follow-up move to the current US administration’s goals to introduce a federal framework for autonomous vehicles, the US Department of Transportation (USDOT) is loosening autonomy restrictions following an announcement from Secretary Sean Duffy on X. This new initiative helps streamline complex regulatory processes and foster home-grown innovation.
From the Wright Brothers to the first astronauts on the moon, our nation has always been at the forefront of transportation technology.
That’s why today we're unveiling a new Automated Vehicle Framework from @USDOT’s Innovation Agenda ⬇️ ⬇️ pic.twitter.com/W3kbMUwQSn
As part of the broader upcoming USDOT Innovation Agenda, the newly unveiled AV Framework is designed to promote American innovation and strengthen domestic engineering while maintaining existing safety standards. The framework centers around three key principles:
Prioritize Safety
Unleash Innovation
Enable Commercial Deployment
To kickstart this AV framework, USDOT announced two initial steps focused on streamlining processes and expanding opportunities.
Crash Reporting Requirements
Under the first principle to Prioritize Safety, the National Highway Traffic and Safety Administration (NHTSA) will maintain its Standing General Order requiring crash reporting on Advanced and Automated Driver Assistance Systems (ADAS and ADS).
However, the reporting process will be streamlined following feedback from AV innovators, likely including Tesla. The goal here is to focus on collecting critical safety information while removing unnecessary or duplicative items from the reporting process, thereby reducing the burden without compromising safety.
Cutting Red Tape
Directly tied to the second principle of Unleash Innovation, the framework also seeks to slash red tape. The first step here is the expansion of the Automated Vehicle Exemption Program, or AVEP. This program allows manufacturers to petition for temporary exemptions from certain federal motor vehicle safety standards (FMVSS) for testing or deployment purposes.
Previously, the standard excluded domestically produced vehicles. Now, domestically produced AVs will not need to meet FMVSS, which will broaden the scope for manufacturers to test more innovative and unique designs and technologies.
Single National Standard for AVs
Finally, tied to the third principle of Enable Commerical Development, USDOT intends to move the United States closer to a single national standard for autonomous vehicles. This aims to prevent a confusing and inefficient patchwork of state-level or city-level laws and regulations, which can create hurdles for companies attempting to innovate, deploy, and scale their technology.
A unified standard across the United States also means that Canada and Mexico will likely be able to follow, as they share homologization standards across North America, including for vehicle crash safety and some autonomy regulations.
What This Means for Tesla
These framework changes will likely have a substantial impact on Tesla. The move towards a national standard is potentially the most impactful change, as Tesla identified regulatory hurdles as one of the most significant challenges it will face with the deployment of both Unsupervised FSD and its Robotaxi network.
The reduction of FMVSS requirements and streamlined reporting will likely play a role in the future as well. The FMVSS requirements are probably already being worked on, if not already met, by the Cybercab and other vehicles in Tesla’s lineup.
Meanwhile, the streamlined reporting will be helpful once Tesla officially launches its Robotaxi network in June.
![Tesla's 2025 Q1 Earnings Call: How to Listen [Listen to Replay]](https://www.notateslaapp.com/img/containers/article_images/multiple-models/group_81.jpg/b2695a53b51e4c7927802deba2534b09/group_81.jpg)
