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.
Tesla recently launched software update 2025.8.3, which included a bunch of “minor updates.” Nestled away in the release notes are a couple of interesting features - including some major changes to the Energy App.
We’re not quite sure we’d call these minor updates, so let’s take a bit of an exploration of the revitalized Energy App.
The Energy App has received some new categorization features in the Consumption tab. You can once again limit the graph by distance. As part of the 2024 Holiday Update, Tesla redesigned the consumption tab and brought it to the new Model S and Model X for the first time. However, with the redesign, Tesla removed the ability to choose the distance shown in the graph, instead providing a static display of the last 200 miles (300km).
The first part of the improved Energy App is bringing back this distance filter. You can now choose between showing the last 10, 100, or 200 miles (15, 150, or 300 km). This differs from the previous available distances of 5, 15 and 30 miles, but brings back the ability for the driver to choose a distance that may be more applicable to how they’re driving.
We’d love for Tesla to introduce custom distances by pinching and zooming the graph or simply add the ability to choose “This Drive Only” to the list of available distances.
Reset Energy App History
@EV3_Model3 on X
Sadly, the “Instant Range” button that was available in the Energy App before is still not available, but Tesla introduced a new feature that may be even better. You can now clear your driving history so that the graph only reflects your current driving style.
This is a little hidden, but if you tap the little info icon at the top near your average Wh/mi, you’ll now be presented with a dialog that lets you clear your history.
This could be useful if you’ve just come from some hard driving on a track or off-road, where energy consumption could be exaggerated. It could also be useful if you just started towing or a different type of driving that is drastically different from your current driving.
Other Changes
In addition, it looks like future predictions of the Energy app now take into account your driving history and apply that in addition to expected vehicle consumption, weather, altitude, and the multitude of other factors that your Tesla constantly takes into account while calculating your expected range.
While these are smaller improvements to the Energy app, they added some important functionality, as it helps users better understand their vehicle consumption and display data that is more applicable to the current driving style.
This is a great example of Tesla listening to its customers and bringing back features they had previously been removed. Check out the video below by akide on the updated Energy app in update 2025.8.3.
Sandy Munro once again had the opportunity to chat (video below) with Lars Moravy, Tesla’s Vice President of Vehicle Engineering, as well as Franz Von Holzhausen, Tesla’s Chief of Vehicle Design. This time, rather than focusing on the improvements to the refreshed Model Y and all the new engineering behind it, the focus was on Tesla’s autonomous ambitions.
In this case, the Cybercab and Robovan were the key highlights, with a distinct focus on the Cybercab. With that all said, let’s dig into all the fun new stuff. You can check out the entire video at the end of the article.
Cybercab Unboxed Process
The Cybercab will be the first of Tesla’s vehicles to use their new unboxed assembly process, which builds vehicles in parallel sections, and then brings them together all at once for final assembly. Traditionally, vehicles are assembled from the ground up, and sometimes even disassembled for parts of production to take place.
This innovative method involves using a big rear and front casting, brought together by a structural battery pack on the floor. The door rings are hot-stamped and laser-welded to form the side shell of the Cybercab, resulting in a very rigid structure that can also absorb crash impacts due to its unique manufacturing.
Paint-Free Panels
We’ve already talked about the Cybercab’s unique paint-free panels, but we’ve never heard until now exactly how Tesla intends to manufacture them. Tesla will inject colored PU plastic onto the backside of another plastic part (formed from various processes) and then ultrasonically weld that entire piece to an inner part.
That essentially combines multiple parts into one smooth piece that is easy to repair - because all you have to do is remove the entire panel - it just comes off as the interior fasteners are removed. There is no subframe holding it together - instead, the subframe is part of the panel.
The only area of the vehicle that will have any paint will be the hot-stamped door rings - which will be painted to improve corrosion resistance to protect the steel.
Aerodynamics
Aerodynamically, the Cybercab boasts the largest aero-cover-to-wheel ratio of any Tesla vehicle. This updated aero cover improves efficiency, even with the Cybercab’s relatively large tires—chosen to reduce rolling resistance. Interestingly, the impact of unsprung weight on efficiency is minimal compared to the benefits of improved rolling resistance and aerodynamics.
Another key to the aerodynamics is the teardrop shape of the Cybercab itself. With its low profile and smooth shape, it is extremely aerodynamically efficient. While neither Sandy nor Lars delved into drag coefficients, we expect it will likely be more efficient than the Model 3’s already fantastic Cd of 0.219.
Range and Battery Pack
Tesla currently has prototypes undergoing real-world testing at Giga Texas to evaluate range and efficiency. The goal is to ensure the vehicle can operate throughout an entire day in the city before returning to charge.
Tesla is targeting a battery pack under 50kWh, delivering close to 300 miles of real-world range—an impressive efficiency of around 166Wh/mi, even outperforming the Model 3’s lowest at 181Wh/mi.
When Does It Arrive?
So, with all that new knowledge - when does the Cybercab arrive? Well - the prototypes for the June Robotaxi network deployment in Austin are getting prepared now - but Tesla expects to begin production and sales sometime in the first half of 2026.