Tesla recently launched FSD in China, which has led many people to wonder exactly how they did it so quickly. Tesla isn't allowed to send training data out of China, meaning that it can’t leverage the capacity of the new Cortex Supercomputer Cluster at Giga Texas.
Instead, Tesla is using their generalized model, in combination with Synthetic Training Data, to train FSD for China. Of course, Tesla also uses this same synthetic data to supplement training for North America and for training for Europe. With European FSD on the horizon, we’ll likely see more and more use of synthetic training data for a sure-fire means to handle edge cases.
Simulated Content
Tesla officially refers to the synthetic training data as “Simulated Content” throughout their patent, which is titled “Vision-Based System Training with Synthetic Content.” Let’s break it down into easier-to-understand chunks.
Vision-Only Training
As you may well know, Tesla’s approach to autonomy focuses on using Tesla Vision. That means cameras providing visual data are the primary - and really only - means of acquiring data from outside of the vehicle. They no longer use radar and only use LiDAR to ensure vision sensor accuracy during training.
Capturing all the information from around the car builds a 3D environment that the vehicle uses to plan its path and conduct its decision-making. All that data is processed to build a fairly comprehensive view of what is around the vehicle and what is predicted to be around the vehicle in the future. All of that is also tagged and characterized to help the system prioritize various decisions.
Supervised Learning Model
Tesla’s FSD training is done through a supervised learning model. That means that the training model is fed data that is already labeled, either by humans or by Tesla’s unique AI model. The objects in the images that are being fed are identified and also tagged with position, velocity, and acceleration. This information acts as a ground truth for the AI model to learn from, allowing it to recognize and interpret similar objects and situations when encountered in real-world driving.
Ground Truth Label Data
The ground truth label data is a critical portion of this supervised learning process. The labeled data provides the model with accurate information about objects and their characteristics in the images. This enables Tesla to develop FSD’s robust understanding of the environment around it while it's driving. This data is typically collected from real-world driving scenarios and is either manually or automatically annotated with data.
Generating Simulated Content
Supplementing the real-world ground truth label data, Tesla employs a simulated content system to generate synthetic training data - which is really the key portion of this patent. This system generates synthetic training data that closely resembles the labeled ground truth data from above.
Content Model Attributes and Contextual Labeling
The generation of that simulated content is guided by what Tesla calls “content model attributes,” which are essentially the key characteristics or features that are extracted from the ground truth label data. These could include things like road edges, lane lines, stationary objects, or even dynamic objects like vehicles or pedestrians.
By varying these attributes, the system can create a wide array of simulated scenarios - which means that FSD’s training program is exposed to as many unique and normal situations as possible.
In addition to the attributes, the system also incorporates contextual labeling - which involves adding labels to the simulated content to help refine it with even more detail. These labels can include things like weather conditions, time of day, or even the type of road or environment the vehicle is driving in. All this information is useful context to help develop FSD’s understanding of driving environments.
Training Data Generation
Tesla’s simulated content system generates vast amounts of training data by creating variations of the content models. These variations generally involve making tweaks to the attributes of the objects in the scene - thereby changing environmental conditions, or introducing new types of driving scenarios, like heavy traffic or construction.
Training FSD
Wrapping it all up - the combined dataset of both real-world data and simulated data is then used to train FSD. By continuously providing new sets of both types of input, Tesla can continue to refine and improve FSD further.
Why Use Simulated Content?
It might seem counterintuitive that Tesla utilizes simulated content for training their autonomous driving system when their vehicles already collect vast amounts of real-world driving data. Their vehicles drive hundreds of millions of miles a month, all across the globe - providing them access to an unfathomable amount of unique data. Well, there are a few reasons to do so.
Not a Tesla App
Cost Reduction
One of the primary advantages of using simulated content is cost reduction. By not having to collect, transmit, sort, label, and process the incoming data from the real world, Tesla can instead just create data locally.
That cuts costs for data transmission, data storage, and all the processing and labeling - whether by human or machine. That can be a fairly significant amount when you think about just how much data goes through Tesla’s servers every single day from vehicles all around the world.
Simulating Challenging Conditions
Simulated content allows Tesla to train FSD in a wide range of environmental conditions that might be rare, difficult, or even dangerous to encounter consistently in real-world driving. This can include challenging conditions like heavy rain, fog, or snow - or even nighttime driving in those conditions.
By training the system on this type of content without trying to pull it from real vehicles, Tesla can ensure that FSD remains operable and fairly robust even in more difficult scenarios in the real world.
Edge Cases & Safety
Another crucial benefit of simulated content is the ability to train FSD on edge cases. While we sometimes jokingly refer to edge cases as things like stopping for a school bus, there are real edge cases that may not be frequently encountered in real-world driving scenarios but can pose real safety risks for drivers, occupants, pedestrians, or other road users. Think of things that you could see happening but have never actually seen, like a car falling off a transport trailer or a highway sign falling down.
As such, Tesla simulates many unique edge cases, including sudden pedestrian crossing, unexpected obstacles in the road, or even erratic behavior from other drivers. All these unique simulations are fairly hard to capture regularly in the real world, which means simulating and training on them is essential to ensure safety.
Efficient and Continuous Optimization
Finally, the vast amount of diverse training data that can be generated by Tesla on demand means that they can quickly and efficiently iterate on FSD without needing to wait for real-world data. This means they can keep a continuous learning process going, ensuring that FSD is always improving bit by bit.
If you’re interested in reading more about the guts that make FSD tick, check out our entire series on FSD-related patents from Tesla here.
We’d also recommend our deep dive into Nvidia’s Cosmos - which is a training system for autonomous vehicles that primarily uses synthetic data to train machine models. It's a different take on Tesla’s FSD training cycle that primarily relies on real data, but it does have some similarities to this particular means of using simulated content.
Subscribe
Subscribe to our newsletter to stay up to date on the latest Tesla news, upcoming features and software updates.
For Tesla owners with vehicles equipped with Hardware 3 (HW3)—once hailed as the key to unlocking autonomy, Robotaxi functionality, and unsupervised Full Self-Driving—the landscape is quickly changing. FSD updates were previously available for HW3 and HW4, but now, HW3 is often excluded from newer FSD updates due to compute and memory constraints. While HW3 vehicles still run a capable version of FSD, they are considerably less smooth than HW4 vehicles.
This has many HW3 owners curious about Tesla’s plans to upgrade these older vehicles, which were once promised to be capable of true autonomy. Let’s take a look at everything Tesla has said and what we should expect.
The HW3 Predicament
Introduced around April 2019, HW3 was a big leap over HW 2.5 and HW 2, with Tesla billing it as the computer that would eventually deliver true self-driving. For a long time, it powered the FSD Beta program. However, as FSD Supervised becomes more complex and data-intensive, particularly with neural networks moving towards an end-to-end AI approach, questions about HW3’s long-term lifespan have grown.
While HW3 vehicles are still receiving FSD updates, with the latest version, V12.6, launching in January 2025, the latest improvements in FSD v13 appear to be stretching even the more modern capabilities of AI4 hardware. This has understandably led to concerns that HW3 will not support Robotaxis and true autonomy.
Tesla’s HW3 Upgrade Promise
To address these concerns, Elon Musk has made increasingly definitive statements. After initially suggesting an upgrade would happen "if needed," he confirmed at the Q4 2025 Earnings Call that Tesla will upgrade HW3 computers for customers who purchased the FSD package.
Musk stated, “That's going to be painful and difficult, but we'll get it done. Now I'm kinda glad that not that many people bought the FSD package, haha.”
While Musk initially stated that Tesla would offer a hardware upgrade if needed, he gave more details this time, stating that the complimentary upgrade would be available for those who purchased the FSD package. Subscribers and non-subscribers will likely need to pay a fee similar to the HW 2 / HW 2.5 upgrade. Interestingly, Tesla was later sued for charging a fee to upgrade to HW3 and had to waive the cost.
When Will the HW3 Upgrade Be Available?
Despite Musk’s confirmation of an upgrade, Tesla hasn’t provided any timelines or estimates for HW3 retrofits. The prevailing logic, and one that aligns with Tesla’s approach to engineering challenges, is that the company is unlikely to initiate a mass upgrade program until FSD is significantly closer to being “solved,” meaning it has achieved true, unsupervised autonomy where a driver is not needed.
Until Tesla knows the final, stable computing power and architectural requirements needed for that level of autonomy, rolling out an interim upgrade wouldn’t make sense. It would risk needing yet another upgrade down the line. Therefore, HW3 owners are in a waiting game - will they wait it out, or will they take one of Tesla’s FSD transfer deals?
What to Expect With the HW3 Upgrade
One thing is clear. The upgrade will not be a simple swap to the current generation of HW4 hardware. AI4, as found in newer Tesla vehicles, has different physical dimensions, power and cooling requirements, and connector configurations that make it incompatible as a direct retrofit into HW3-designed vehicles. It’d require a lot of effort and cost to adapt HW4 for HW3 vehicles.
This means Tesla will have to develop another custom-designed retrofit FSD computer specifically for HW3 replacements. This computer must fit within an existing and defined physical space and operate within the power and cooling budget of older vehicles.
Speculation naturally turns to Tesla’s next-generation FSD hardware, HW5 or AI5. Elon previously indicated that AI5 would appear in new vehicles near the end of 2025, initially citing a timeframe of 12-18 months back in mid-2023. However, it now looks like it’ll ship sometime in the first half of 2026.
Potentially, a variant of this new AI5 computer, perhaps a more power-efficient or underclocked version, could be engineered to form the basis of the HW3 retrofit solution. This is plausible, as newer chip architectures often bring considerably greater efficiency, potentially allowing a more powerful new design to operate within HW3’s constraints.
What About HW4 and HW5?
The current-generation FSD computer, HW4, is already facing some constraints with the latest FSD v13 updates. This means buyers and owners of AI4 vehicles are also starting to have this question creep into the back of their heads… “What about my vehicle?”
Based on Tesla’s official statements on AI5, it is poised to be a powerhouse of an upgrade. That means up to 10 times the processing capability of AI4. This is an immense increase in processing power, and over time, Tesla will likely use every bit of it to make FSD handle as many edge cases as possible. While AI4’s computing power was a modest increase from HW3, the leap from AI4 to AI5 is expected to be significantly larger.
Tesla’s executive team has stated that the existing cameras on HW3-equipped vehicles are “capable” and that the upgrade will be focused on the FSD computer. While the AI4 cameras offer a much higher resolution than HW3, Tesla says they’re not needed. This appears to contradict what Tesla is doing as of FSD v13.2. In that update, Tesla introduced processing FSD camera feeds at full resolution, suggesting that there is some advantage to the higher-resolution cameras.
Musk also stated that cameras would not be upgraded in HW3 vehicles.
As we’ve previously covered, the newer HW4 cameras offer several advantages over the HW3 camera generation, which include:
Higher Resolution: The AI4 cameras feature 5 megapixels, compared to the 1.2 megapixels on HW3 cameras, which allows the vehicle to see things further away and in sharper detail.
Improved Dynamic Range and Low-Light Performance: The improved dynamic range allows the system to see more clearly in low-light conditions, such as during sunrise or sunset, or at night.
Wider Field of View: The rear camera on AI4 features a significantly larger field of view, providing greater awareness of the vehicle's surroundings.
It's known that AI4 processes camera data at these higher resolutions, which undoubtedly contributes to its increased performance in decision-making, object recognition (especially at a distance or for small details, such as text on signs), and overall FSD smoothness.
Therefore, while a new, more powerful retrofit computer for HW3 vehicles will bring substantial improvements, it will still be processing input from the older-generation cameras. Another technical challenge that Tesla will need to address is how to maximize FSD performance using the existing HW3 cameras.
Infotainment (MCU) Upgrade?
Most HW3-era vehicles are equipped with the older Intel Atom-based infotainment computer, known as MCU 2. Newer Teslas, as well as newer HW3 vehicles, use the considerably faster AMD Ryzen-based MCU 3. Given that Tesla sometimes packages the FSD computer and infotainment computer together, it wouldn’t be too surprising to see an MCU upgrade as part of an FSD computer retrofit.
While this would be a welcome improvement, providing a snappier user interface and better media capabilities, Tesla has not confirmed any such plans. The FSD computer and the MCU are technically separate systems, but Tesla usually bundles them together to save on costs. While Tesla has offered paid MCU upgrades in the past (e.g., from the older MCU 1 to MCU 2), there is currently no official upgrade path from MCU 2 to MCU 3.
It’s best to assume that the promised free FSD computer upgrade will not automatically include an infotainment system upgrade as well, but it’s certainly possible, given that Tesla usually bundles these together.
Playing the Waiting Game
For Tesla owners who purchased FSD with their HW3 vehicles, the commitment for a free hardware upgrade is on the record. However, the "when" and "what" remain tied to the challenge of achieving true, unsupervised autonomy. Once Tesla understands the compute power required to solve FSD, we’ll likely hear more about this hardware upgrade. Until then, we’ll have to hold on tight with FSD v12.6.
In just 8 months, Tesla has gone from breaking ground to delivering electrons at its most ambitious Supercharger project to date, just in time to be ready for the busy Fourth of July holiday weekend. Project Oasis, the world’s largest Supercharger site, is now partially open to customers for its first phase in Lost Hills, California.
What makes this remarkable is the speed of execution. In just eight months, Tesla has constructed a site that will eventually feature 168 stalls (84 stalls are now open), supported by 11 MW of solar power and 10 Megapacks of battery storage. That construction speed is pretty impressive, but what is even more impressive is how this new station operates and what it means for future Supercharging infrastructure.
Self-Sufficient Energy Oasis
Not a Tesla App
The first 84 stalls at Lost Hills are now open, and according to the Tesla Charging team, they are currently powered solely by the sun and operate off-grid.
This makes it more than just a new Supercharger site. It serves as a proof of concept for a new type of Supercharger. Unlike nearly every other charging site in the world, which draws power from local utilities, this station generates its own clean electricity from its massive solar array and stores it in its array of on-site Megapacks.
Self-sufficient charging stations are something completely different than what we see today. They are highly resilient since they’re not reliant on the grid. That means that even if there is a local power outage, brownout, or blackout, one can always come to Lost Hills to Supercharge.
If you’ve got a Cybertruck, you could take advantage of the Cybertruck’s Powershare feature and charge up at Lost Hills to help keep your home powered during a blackout, utilizing the Cybertruck as a portable battery charger. Now that’s true independence and self-reliance.
The Future of Charging
Solar-powered Superchargers help avoid massive new loads on already stressed electrical grids, especially during peak afternoon and evening hours, when demand is the highest.
This is Tesla’s vision for the future of charging: a clean, fully closed-loop ecosystem that sustains itself. The sun’s energy is captured, stored, and delivered directly to vehicles on site at any time of day without relying on the electrical grid or fossil fuels.
Largest Supercharger in the World
This opening of 84 stalls is just the first phase of the project. Tesla says that the remaining stalls, as well as a new on-site lounge, are coming later this year. Once complete, the 168-stall site will be the largest Supercharger site in the world.
While the speed of building such a massive project in just eight months is a testament to Tesla’s execution, the true innovation is actually that self-sustainability. Let’s hope we see even more large, self-sufficient Supercharger sites across the world in the near future.
![Tesla Updates Robotaxi App: Adds Adjustable Pick Up Locations, Shows Wait Time and More [VIDEO]](https://www.notateslaapp.com/img/containers/article_images/tesla-app/robotaxi-app/25-7-0/robotaxi-app-25.7.0.webp/4ac9ed40be870cfcf6e851fce21c43b9/robotaxi-app-25.7.0.jpg)
