Tesla needs to come clean about HW3 before the word ‘fraud’ comes out
The walls are closing on Tesla’s claim that millions of its vehicles with Hardware 3 (HW3) computers will be capable of unsupervised self-driving.
Tesla needs to come clean before the word “fraud” comes out.
Making a mistake is not a fraud. If Tesla really thought that it could deliver unsupervised self-driving to vehicles equipped with HW3 and, at one point, it figured out that it couldn’t, it’s not fraud even though it used that as a selling point for millions of vehicles for years.
However, the moment Tesla figures out that it can’t, it needs to stop selling its Full Self-Driving package to HW3 vehicle owners and come clean to owners about what their vehicle will and will not be able to do, like a robotaxi service.
Has the moment come?
Delivering self-driving on Tesla HW3/self-driving computer
In 2016, Elon Musk announced that all future Tesla vehicles would come equipped with the necessary hardware for self-driving capabilities, even specifying “level 5 self-driving,” which implies the ability to operate autonomously under any conditions. However, shortly after, Musk acknowledged that Tesla might require more onboard computing power than initially thought, leading to the introduction of Hardware 3 (HW3), which Tesla also called its “self-driving computer”.
Musk assured that HW3 would enable full self-driving (FSD) capabilities, promising retrofits for earlier models that had purchased the FSD package. When I bought my own Tesla Model 3 in 2018, it was equipped with the original computer, but since I had purchased the FSD package, Tesla upgraded my car with the new “self-driving computer” in 2019.
Following this, Tesla introduced Hardware 4 (HW4), a more advanced onboard computer system, but did not offer retrofits for older models with HW3, maintaining that HW3 was sufficient for achieving self-driving through software updates.
Musk said that it wouldn’t be “economically feasible” to retrofit HW3 vehicles with HW4, which not only includes a more powerful computer but also better cameras.
Initially, Musk claimed that FSD improvements would first be optimized for HW3, suggesting that HW4 might lag behind by at least “six months”. However, Tesla reversed this approach with the release of FSD version 12.5, which was first deployed to HW4 vehicles. Musk explained that optimizing the software for the less powerful HW3 would take additional time.
This has raised a significant red flag hinting at the limitations of HW3 in handling the latest software advancements towards unsupervised self-driving, a capability Tesla promised to HW3 owners since 2016.
The concern is especially significant within the context that Tesla still has a lot of work to do to deliver its unsupervised self-driving capabilities.
Tesla has always gone out of its way not to release any data regarding its FSD program. Therefore, we have to rely on crowdsourced data, which shows Tesla is currently at about 122 miles between critical disengagement:
According to most experts, Tesla needs a ~1,000x increase in miles between disengagement to deliver on its unsupervised self-driving promises. As you can see, this data shows that Tesla achieved a ~2x improvement over the last 3 years.
On top of this situation, CEO Elon Musk got people even more worried during the launch of the Robotaxi last week.
While discussing his claim that “all Tesla vehicles will be capable of self-driving,” someone in the crowd asked him about the Cybertruck, which Musk quickly answered with a “yes.”
However, when someone asked him about HW3 vehicles, instead of simply responding “yes”, Musk said “Let’s not get nuanced here” and then quickly asked for the next slide:
Now, still at the Robotaxi event last week, some have been pointing to this interaction with Tesla executives Franz von Holzhausen and Lars Moravy saying again that robotaxi-level self-driving is coming to “all cars” after being asked more specifically about HW3 as evidence that Tesla believes it’s still possible to deliver FSD unsupervised on HW3:
With all due respect to von Holzhausen and Moravy, they wouldn’t be the best people to ask. The former is in charge of design and the latter of vehicle engineering, which you would think the FSD program would fall under, but no.
Ashok Elluswamy leads the program at Tesla and reports directly to CEO Elon Musk.
That’s evidenced by some mistakes made even in this short interaction like Moravy saying that Tesla announced its self-driving effort in 2014 when it was in 2016 and him asking if a 2018 Model 3 has HW4, which has never been available on early Model 3 vehicles.
Speaking of the Robotaxi event, Musk said that the new Robotaxi is equipped with a new hardware suite, especially a new on board computer called AI5. He didn’t elaborate on the capacity of the new computer. The vehicle also has a bumper camera, which only the Cybertruck has and no other Tesla vehicle on the road today. The onboard compute power is one thing, but it’s also not the only potential bottleneck for Tesla with older hardware.
Another important piece of evidence pointing to Tesla not being able to deliver unsupervised self-driving on HW3 vehicles is the fact that it doesn’t have any compute redundancy anymore.
Electrek spoke with a well-known Tesla hacker called ‘green‘ who often reveals information about Tesla through his deep dives into the automaker’s software. He actually released the first HW3 images back in 2019.
Green reports that starting in late 2023, Tesla started to use both nodes for its FSD program on HW3 – running some new neural nets on the extra node. Originally, the idea was to have one for redundancy, which is necessary for higher levels of autonomy like levels 4 and 5, but arguably also level 3.
Now, green says that if one of the nodes fails, FSD doesn’t drive anymore. It can still produce FSD visualizations, but that’s about it. That alone basically kisses goodbye to robotaxi-level self-driving on HW3.
It’s also worth noting that shortly after green noticed this change happened, Tesla started to shift its priority from releasing new software on HW4 first rather than HW3.
Tesla is reducing its liability
Tesla has been trying to actively reduce its legal liability regarding HW3 by encouraging people who bought FSD to upgrade to newer vehicles.
For years Tesla owners have been asking Tesla to allow them to freely transfer their FSD package to a new vehicle. It makes sense. Tesla hasn’t delivered the product they have paid for. It’s the bare minimum to allow them to transfer it to a new car.
After years of refusing, Musk eventually agreed to FSD transfer last year, but he called it a “one-time amnesty” and said to take advantage of it.
That turned out not to be true. Tesla brought back the FSD transfer twice more since – with last quarter Musk saying “one more time”. And then, sure enough, Tesla brought it back for a fourth time this quarter.
This fake incentive to upgrade your older car with FSD to a newer one now because it’s the “last time” has a positive effect on Tesla’s liability regarding HW3.
When Tesla resales those used HW3 vehicles with FSD, they use their new language called “(Supervised) Full Self-Driving”, which opens the door for Tesla to say that they are only selling you self-driving that needs to be “supervised” by a driver.
But interestingly, for HW2 vehicle owners who never purchased FSD, Tesla is still selling them a $1,000 HW3 computer upgrade and $2,000 FSD software package ($2,000 if you have Enhanced Autopilot) with still the old language in the upgrade page:
That’s where Tesla would be adding liability as it would be “upgrading” a car to a 5-year-old computer that is already lagging behind on updates to its newer 2-year-old computer (HW4).
Electrek’s Take
Let’s be honest. Tech is rarely supported with software updates after 5-7 years. Tesla Hardware 3 is entering that zone. It is becoming obsolete and normally, it wouldn’t be a problem, but Tesla sold a Full Self-Driving capability package for up to $15,000 based on this hardware that it never delivered.
At the minimum, it will have to reimburse that, but owners can even argue that they bought the car because Elon Musk told them it would become self-driving over time and become an “appreciating asset.”
This could quickly become a very large liability for Tesla, and the way it handles it is also important.
Musk said that retrofits are not economically feasible from HW3 to HW4. It’s true that it would be quite expensive and also likely create an insurmountable amount of work for Tesla’s already overworked service teams. The HW4 computer doesn’t have the same power harness or camera harnesses as the HW3, and it doesn’t share a form factor that fits in the exact same spot.
Also, the cameras have been upgraded with HW4, which raises the question, “Is the computing power the only problem, or does the camera also need to improve?”
If it’s just the computing power, Tesla could potentially design a new computer that could be more easily retrofitted in HW3 cars, but even then, that’s something that needs to be disclosed.
As I said, if Tesla knows that it can’t deliver unsupervised self-driving on HW3, it needs to let owners know right now and stop selling the software package to HW3 owners without a clear plan to make things right. Otherwise, this quickly becomes fraudulent.
The fact that Elon and Tesla have been wrong so many times about self-driving is already not a great confidence builder for them delivering on HW4 vehicles or even on the new AI5 (Robotaxi), but if they are also actively misleading owners, then Tesla becomes untrustworthy.
I am seriously concerned that Tesla is going to rely on the “corporate puffery” defense to frame Elon’s promises as “mere puff”.
After I first brought up the potential of Tesla reaching the limits of HW3 earlier this year, many Elon superfans started to make claims that Tesla and Elon never promised robotaxi-level self-driving capabilities on HW3 cars, which is plain ridiculous.
Tesla could also blame regulators as this is the new language that you have to agree with when buying what is now called “Full Self-Driving (Supervised)”:
The currently enabled Autopilot and Full Self-Driving (Supervised) features require active driver supervision and do not make the vehicle autonomous. Full autonomy will be dependent on achieving reliability far in excess of human drivers as demonstrated by billions of miles of experience, as well as regulatory approval, which may take longer in some jurisdictions. As Tesla’s Autopilot and Full Self-Driving (Supervised) features evolve, your vehicle will be continuously upgraded through over-the-air software updates.
On the very same day that Tesla presented its new Robotaxi, Former President Donald Trump, who Tesla CEO Elon Musk is financially backing to become the next president and who he says he is “all-in” on, said that he would “ban autonomous vehicles on American roads.”
This situation is quite a mess to say the least.
FTC: We use income earning auto affiliate links. More.
Tesla is now pulling on all the demand levers in the US with new discounts and incentives as sales are crashing due to brand damage.
Over the last few days, Tesla has introduced a series of new discounts and incentives in the US.
Previously, Tesla had a program to offer a $1,000 discount for US military personnel, but the automaker has now extended it to “students, teachers, first-responders, military veterans, retirees, active-duty members, their spouses, and surviving spouses.”
The update incentive applies to Tesla’s entire lineup of new vehicles.
Tesla also introduced a new incentive for Lyft drivers. They are eligible to $1,000 in Tesla credits when taking delivery and $1,000 from Lyft if they complete 100 deliveries by July 13.
The automaker wrote on its website:
Eligible Lyft drivers who purchase a new Tesla vehicle can receive $1,0001 in Tesla Credits upon taking delivery and a $1,000 incentive from Lyft after completing 100 trips on or before July 13, 2025. Tesla Credits can be used toward Supercharging, a new Tesla vehicle, service appointments or select Tesla Shop or upgrade purchases. Offer available to active Lyft drivers in good standing.
Tesla also started reaching out to Cybertruck reservation holders to let them know that they only have a month before they can’t take advantage of lower FSD prices.
The automaker wrote in the email:
As an early reservation holder, you have access to a reserved Full Self-Driving (Supervised) price of $7,000. To keep this price, you’ll need to take delivery by June 15, 2025. After June 15, 2025, FSD (Supervised) will be available at the latest price, which is currently $8,000.
When Tesla started taking Cybertruck reservations in 2019, Tesla said that by reserving the truck, reservation holders were locking in the then $7,000 price for its ‘Full Self-Driving’ package.
It looks like Tesla is now putting a deadline to take advantage of this deal to boost orders of the Cybertruck, which has proven to be a commercial flop.
On top of all these incentives, Tesla is also subsidizing interest rates to offer 0% financing on Model 3, and 1.99% financing on Model Y.
All those incentives in place point to Tesla having significant demand issues in the US.
Tesla’s global sales came about 50,000 units below expectations, which the company blamed on the production changeover of Model Y, its most popular model by far.
However, production is now back up to normal in Q2, and Tesla is clearly having issues selling the updated Model Y.
The automaker has no backlog of orders for the new Model Y and vehicles are already piling up in inventory:
We reported last week that Tesla employees wrote an open letter calling for Elon Musk’s removal as CEO due to the damage he has caused to the brand.
In the letter, the employees confirmed Tesla’s demand issues, saying that thousands of new Model Ys are now sitting unsold on lots in the US.
Electrek’s Take
This is not a great sign for Tesla. These are end-of-quarter level incentives when we are just about halfway through the quarter.
And that’s just in the US, where Tesla’s sale performance is more opaque.
In Europe and China, where we know for a fact that Tesla is struggling with sales, the automaker is virtually offering 0% financing on its entire lineup.
FTC: We use income earning auto affiliate links. More.
The electric box van experts at Harbinger announced a new, EREV version of their medium-duty van that pairs a big battery with a small, gas-powered ICE engine to offer fleets that are hesitant to electrify a massive 500 miles of autonomy on a single charge + tank.
The American truck brand is putting its latest $100 million raise to good use, developing a cost-competitive EREV chassis that marries a low-emissions 1.4L inline four-cylinder gas engine with a close coupled 800V generator sending power to a 140 or 175 kW battery for up to 500 miles of fully loaded range. More than enough, in other words, to meet the needs of just about any fleet you can think of.
That’s a good thing, too, because medium-duty trucks are put to work in just about any circumstance you can think of, as well – a fact that’s not lost on Harbinger.
“Medium-duty vehicles serve an incredibly diverse range of applications, just like the fleets and operators that rely on them, ” explains John Harris, Co-founder and CEO, Harbinger. “There are some fleets whose needs simply can’t be met with a purely electric vehicle—and we recognize that. Our hybrid is designed for use cases and routes that go beyond what an all-electric system typically supports. The series hybrid delivers the benefits of an electric drivetrain, along with the added confidence of a range extender when needed.”
In addition an up-front cost that should make it an attractive prospect for fleet buyers, the new Harbinger EREV pack performance that should made it attractive for its drivers, too. The new chassis’ electric powertrain delivers 440 hp and 1,140 lb-ft of tq for quick acceleration into traffic and smooth running, even under load. Charging performance is also quick, with the ability to get the big battery from 10-80% charge in just under an hour on a 150 kW port.
You’ve heard all this before
If that sounds familiar, that’s because it is. This medium-duty chassis was first shown last year, making its debut under a Thor Class A motorhome concept that we covered in September. That vehicle promised the same great EREV range and capability to a market that values independence and spontaneity more than most, and bringing those values to a medium-duty commercial market that’s lapping up “messy middle” propaganda from Shell NACFE is just smart business.
The new Harbinger chassis’ batteries are manufactured by Panasonic. No word on who is making the 1.4L ICE generator, but my money’s on the GM SGE four-cylinder last seen in the gas-powered Chevy Spark. You guys are smart, though – if you have a better guess who the supplier might be, let us know in the comments.
SOURCE | IMAGES: Harbinger.
If you’re considering going solar, it’s always a good idea to get quotes from a few installers. To make sure you find a trusted, reliable solar installer near you that offers competitive pricing, check out EnergySage, a free service that makes it easy for you to go solar. It has hundreds of pre-vetted solar installers competing for your business, ensuring you get high-quality solutions and save 20-30% compared to going it alone. Plus, it’s free to use, and you won’t get sales calls until you select an installer and share your phone number with them.
Your personalized solar quotes are easy to compare online and you’ll get access to unbiased Energy Advisors to help you every step of the way. Get started here.
FTC: We use income earning auto affiliate links. More.
Environment
Trump wants coal to power AI data centers. The tech industry may need to make peace with that for now
President Donald Trump wants to revive the struggling coal industry in the U.S. by deploying plants to power the data centers that the Big Tech companies are building to train artificial intelligence.
Trump issued an executive order in April that directed his Cabinet to find areas of the U.S. where coal-powered infrastructure is available to support AI data centers and determine whether the infrastructure can be expanded to meet the growing electricity demand from the nation’s tech sector.
Trump has repeatedly promoted coal as power source for data centers. The president told the World Economic Forum in January that he would approve power plants for AI through emergency declaration, calling on the tech companies to use coal as a backup power source.
“They can fuel it with anything they want, and they may have coal as a backup — good, clean coal,” the president said.
Trump’s push to deploy coal runs afoul of the tech companies’ environmental goals. In the short-term, the industry’s power needs may inadvertently be extending the life of existing coal plants.
Coal produces more carbon dioxide emissions per kilowatt hour of power than any other energy source in the U.S. with the exception of oil, according to the Energy Information Administration. The tech industry has invested billions of dollars to expand renewable energy and is increasingly turning to nuclear power as a way to meet its growing electricity demand while trying to reduce carbon dioxide emissions that fuel climate change.
For coal miners, Trump’s push is a potential lifeline. The industry has been in decline as coal plants are being retired in the U.S. About 16% of U.S. electricity generation came from burning coal in 2023, down from 51% in 2001, according to EIA data.
Peabody Energy CEO James Grech, who attended Trump’s executive order ceremony at the White House, said “coal plants can shoulder a heavier load of meeting U.S. generation demands, including multiple years of data center growth.” Peabody is one of the largest coal producers in the U.S.
Grech said coal plants should ramp up how much power they dispatch. The nation’s coal fleet is dispatching about 42% of its maximum capacity right now, compared to a historical average of 72%, the CEO told analysts on the company’s May 6 earnings call.
“We believe that all coal-powered generators need to defer U.S. coal plant retirements as the situation on the ground has clearly changed,” Grech said. “We believe generators should un-retire coal plants that have recently been mothballed.”
Tech sector reaction
There is a growing acknowledgment within the tech industry that fossil fuel generation will be needed to help meet the electricity demand from AI. But the focus is on natural gas, which emits less half the CO2 of coal per kilowatt hour of power, according the the EIA.
“To have the energy we need for the grid, it’s going to take an all of the above approach for a period of time,” Kevin Miller, Amazon’s vice president of global data centers, said during a panel discussion at conference of tech and oil and gas executives in Oklahoma City last month.
“We’re not surprised by the fact that we’re going to need to add some thermal generation to meet the needs in the short term,” Miller said.
Thermal generation is a code word for gas, said Nat Sahlstrom, chief energy officer at Tract, a Denver-based company that secures land, infrastructure and power resources for data centers. Sahlstrom previously led Amazon’s energy, water and sustainability teams.
Executives at Amazon, Nvidia and Anthropic would not commit to using coal, mostly dodging the question when asked during the panel at the Oklahoma City conference.
“It’s never a simple answer,” Amazon’s Miller said. “It is a combination of where’s the energy available, what are other alternatives.”
Nvidia is able to be agnostic about what type of power is used because of the position the chipmaker occupies on the AI value chain, said Josh Parker, the company’s senior director of corporate sustainability. “Thankfully, we leave most of those decisions up to our customers.”
Anthropic co-founder Jack Clark said there are a broader set of options available than just coal. “We would certainly consider it but I don’t know if I’d say it’s at the top of our list.”
Sahlstrom said Trump’s executive order seems like a “dog whistle” to coal mining constituents. There is a big difference between looking at existing infrastructure and “actually building new power plants that are cost competitive and are going to be existing 30 to 40 years from now,” the Tract executive said.
Coal is being displaced by renewables, natural gas and existing nuclear as coal plants face increasingly difficult economics, Sahlstrom said. “Coal has kind of found itself without a job,” he said.
“I do not see the hyperscale community going out and signing long term commitments for new coal plants,” the former Amazon executive said. (The tech companies ramping up AI are frequently referred to as “hyperscalers.”)
“I would be shocked if I saw something like that happen,” Sahlstrom said.
Coal retirements strain grid
But coal plant retirements are creating a real challenge for the grid as electricity demand is increasing due to data centers, re-industrialization and the broader electrification of the economy.
The largest grid in the nation, the PJM Interconnection, has forecast electricity demand could surge 40% by 2039. PJM warned in 2023 that 40 gigawatts of existing power generation, mostly coal, is at risk of retirement by 2030, which represents about 21% of PJM’s installed capacity.
Data centers will temporarily prolong coal demand as utilities scramble to maintain grid reliability, delaying their decarbonization goals, according to a Moody’s report from last October. Utilities have already postponed the retirement of coal plants totaling about 39 gigawatts of power, according to data from the National Mining Association.
“If we want to grow America’s electricity production meaningfully over the next five or ten years, we [have] got to stop closing coal plants,” Energy Secretary Chris Wright told CNBC’s “Money Movers” last month.
But natural gas and renewables are the future, Sahlstrom said. Some 60% of the power sector’s emissions reductions over the past 20 years are due to gas displacing coal, with the remainder coming from renewables, Sahlstrom said.
“That’s a pretty powerful combination, and it’s hard for me to see people going backwards by putting more coal into the mix, particularly if you’re a hyperscale customer who has net-zero carbon goals,” he said.
-
Sports3 years ago‘Storybook stuff’: Inside the night Bryce Harper sent the Phillies to the World Series
-
Sports1 year agoStory injured on diving stop, exits Red Sox game
-
Sports2 years agoGame 1 of WS least-watched in recorded history
-
Sports2 years agoMLB Rank 2023: Ranking baseball’s top 100 players
-
Sports4 years ago
Team Europe easily wins 4th straight Laver Cup
-
Environment2 years agoJapan and South Korea have a lot at stake in a free and open South China Sea
-
Environment2 years agoGame-changing Lectric XPedition launched as affordable electric cargo bike
-
Business3 years agoBank of England’s extraordinary response to government policy is almost unthinkable | Ed Conway