News
Tesla owners share first impressions of FSD V12.5 on Hardware 3 vehicles
It appears that Tesla is now ramping the release of FSD V12.5.1.5 to more Hardware 3 vehicles. Based on comments from HW3 Tesla owners on social media, it appears that the recently released update represents a good step forward for the advanced driver-assist system.
As noted by Tesla Director of Autopilot Software Ashok Elluswamy in a recent comment on social media platform X, the company was able to achieve similar performance as HW4’s FSD V12.5 release with a smaller model for Hardware 3 vehicles. He also noted that deploying the larger model required Tesla to “implement a few kernels in the compiler to emulate the same operations that are natively supported on the AI4 hardware.”
First Drive of 12.5.1.5 on HW3! Fixed the One Main Issue on this route! pic.twitter.com/6RDBIwwD2l— Rocco Speranza (@R0cco_Speranza) August 23, 2024
While initial reports of FSD V12.5’s HW3 release still suggested that the software was being rolled out to employees, later posts on social media suggested that regular consumers were also receiving the update. Reviews posted on social media from longtime Tesla owners so far suggest that FSD V12.5 may very well be the best version of the advanced driver assist system that has been rolled out to HW3 vehicles to date.
? FSD 12.5.1.5 Initial impression on HW3 (compared to 12.4.3)
• Feels smoother
• Seems to be able to anticipate other cars' movement much better
• Overall it feels "quicker" and more confident in making decisions
• Highway stack definitely feels more decisive
• Too low… https://t.co/TpUpGHNiQk pic.twitter.com/pgd9jTU2Mv— Effetto Tesla (@EffettoTesla) August 23, 2024
Among the most prominent comments from HW3 vehicle owners was that FSD V12.5 definitely feels smoother than previous iterations of the advanced driver assist system. The build reportedly also feels quicker and more confident in making decisions. This is quite evident in the freeway, as the highway stack has reportedly become more decisive with the V12.5 update as well. That being said, some owners also mentioned that V12.5’s auto speed is a bit too slow in Hardware 3 vehicles.
FSD 12.5.1.5 first impressions on HW3 “legacy” Model S (~45 min drive)
– the 12.5 magic is definitely still there. it takes natural paths through all the turns, has very good lane positioning, and much better acceleration / braking than any version so far
– it’s not hands free… pic.twitter.com/FYgEQcckOz— ΛI DRIVR (@AIDRIVR) August 23, 2024
Longtime FSD tester @AIDRIVR, who uses a Hardware 3 Tesla Model S with no interior camera, observed that the new update tends to take natural paths through turns. He also noted that FSD V12.5 has very good lane positioning. The update reportedly also featured the best acceleration and braking compared to any FSD version that has been released to date.
I just took FSD 12.5.1.5 (HW3) out for a quick after dark drive. It was flawless and extremely comfortable. However….!!! I think I could have said the same about many previous releases under the same situation. More testing required to form a full opinion. @BarryGoodStocks https://t.co/FKAAxvuNNw— FSD (Beta) Test Pilot (@j32pmxr) August 23, 2024
Overall, it does appear that FSD V12.5 is a solid release for Hardware 3 vehicles. However, it does present some concerns about how fast older Teslas would be receiving the latest updates from the electric vehicle maker. Tesla first started selling FSD as an add-on for Enhanced Autopilot in 2016, after all, and Hardware 3 was only made available for vehicles in 2019. Hardware 4 was released in 2023.
Watch a video of FSD V12.5 on a Hardware 3 vehicle below.
Don’t hesitate to contact us with news tips. Just send a message to simon@teslarati.com to give us a heads up.
News
Rolls-Royce makes shocking move on its EV future
When Rolls-Royce unveiled its first all-electric model, the Spectre, in 2022, former CEO Torsten Müller-Ötvös declared the brand would cease production of internal combustion engine vehicles by the end of the decade.
Rolls-Royce made a shocking move on its EV future after planning to go all-electric by the end of the decade. Now, the company is tempering its expectations for electric vehicles, and its CEO is aiming to lean on its legacy of high-powered combustion engines to lead it into the future.
In a significant reversal, Rolls-Royce Motor Cars has scrapped its ambitious plan to become an all-electric manufacturer by 2030. The luxury British marque announced the decision amid sustained customer demand for traditional combustion engines and shifting regulatory landscapes.
When Rolls-Royce unveiled its first all-electric model, the Spectre, in 2022, former CEO Torsten Müller-Ötvös declared the brand would cease production of internal combustion engine vehicles by the end of the decade.
The move aligned with the industry’s broader push toward electrification, promising silent, effortless power befitting the “Rolls-Royce of cars.”
However, new CEO Chris Brownridge, who assumed the role in late 2023, has reversed course. “We can respond to our client demand … we build what is ordered,” Brownridge stated.
The company will continue offering its iconic V12 engines, which remain a cornerstone of its heritage and appeal to discerning buyers who appreciate the distinctive sound and character. He noted the original pledge was “right at the time,” but “the legislation has changed.”
While not abandoning electric vehicles entirely, the Spectre remains in production, with an electric Cullinan option forthcoming; the decision marks the end of a strict all-EV timeline. Relaxed emissions regulations and slowing EV demand, evidenced by a 47 percent drop in Spectre sales to 1,002 units in 2025, forced the reconsideration.
It was a sign that perhaps Rolls-Royce owners were not inclined to believe that the company’s all-EV future was the right move.
Rolls-Royce joins a growing roster of automakers reevaluating aggressive electrification targets.
Fellow luxury brand Bentley has pushed its full electrification from 2030 to 2035, while continuing to offer hybrids and ICE models. Mercedes-Benz walked back its 2030 all-EV goal, now aiming for about 50% electrified sales while keeping combustion engines into the 2030s. Porsche has abandoned its 80% EV sales target by 2030, delaying models and extending hybrids.
Mainstream giants are following suit. Honda canceled its U.S. EV plans, including the 0-Series and Acura RSX, facing a $15.7 billion hit as it doubles down on hybrids. Ford and General Motors have incurred tens of billions in writedowns, canceling models and pivoting to hybrids amid an industry total exceeding $70 billion in charges.
This trend reflects a pragmatic shift driven by infrastructure gaps, consumer preferences, and policy changes. In the ultra-luxury segment, where emotional connection reigns, automakers are prioritizing flexibility over rigid deadlines, ensuring brands like Rolls-Royce evolve without alienating their core clientele.
News
Elon Musk teases expectations for Tesla’s AI6 self-driving chip
This optimistic timeline for tape-out—the stage where chip design is finalized before manufacturing—signals Tesla’s push to rapidly advance its silicon capabilities.
Tesla CEO Elon Musk is outlining expectations for the AI6 self-driving chip, which is still two generations away. Despite this, it is already in the plans of the company and its serial entrepreneur CEO, who has high expectations for it.
Musk provided fresh details on the company’s aggressive AI hardware roadmap, spotlighting the upcoming AI6 chip designed to supercharge Tesla’s self-driving tech, humanoid robots, and data center operations.
In a post on X dated March 19, Musk stated, “With some luck and acceleration using AI, we might be able to tape out AI6 in December.”
With some luck and acceleration using AI, we might be able to tape out AI6 in December
— Elon Musk (@elonmusk) March 19, 2026
This optimistic timeline for tape-out—the stage where chip design is finalized before manufacturing—signals Tesla’s push to rapidly advance its silicon capabilities.
The announcement builds on progress with the predecessor AI5. Earlier in January, Musk announced that the AI5 design was “in good shape” and “almost done,” describing it as an “existential” project for the company that demanded his personal attention on weekends.
He characterized AI5 as roughly equivalent to Nvidia’s Hopper class performance in a single system-on-chip (SoC) and Blackwell-level as a dual configuration, but at significantly lower cost and power usage.
Elon Musk is setting high expectations for Tesla AI5 and AI6 chips
Musk highlighted that AI5 “will punch far above its weight” thanks to Tesla’s co-designed AI software and hardware stack, making maximal use of every circuit. While capable of data center training tasks, it is primarily optimized for edge computing in Optimus robots and Robotaxi vehicles.
For AI6, Musk envisions substantial gains. “In the same half reticle and same process node, we think a single AI6 chip has the potential to match a dual SoC AI5,” he explained.
The company is targeting ambitious nine-month development cycles for future chips, allowing rapid iteration to AI7, AI8, and beyond. AI5/AI6 engineering remains Musk’s top time allocation at Tesla, with the CEO calling AI5 “good” and AI6 “great.”
Samsung is expected to manufacture the AI6 chips, following deals worth billions, while AI5 will leverage TSMC and Samsung production. These chips will form the backbone of Tesla’s Full Self-Driving system, enabling safer and more capable autonomy, alongside powering dexterous movements in Optimus bots and efficient inference in expanding data centers.
Tesla to discuss expansion of Samsung AI6 production plans: report
Musk has also restarted work on the Dojo 3 supercomputer project now that AI5 is progressing. Long-term plans include in-house manufacturing via the Terafab facility.
By accelerating chip development with AI tools, Tesla aims to reduce dependence on third-party GPUs and deliver high-performance, energy-efficient solutions tailored to its ecosystem. Success with AI6 could mark a major milestone in Tesla’s journey toward full autonomy and robotics leadership, though timelines remain subject to manufacturing realities.
Elon Musk
SpaceX is quietly becoming the U.S. Military’s only reliable rocket
Space Force drops ULA for SpaceX on GPS launch after Vulcan rocket anomaly investigation halts flights.
The U.S. Space Force announced today it is switching an upcoming GPS III satellite launch from United Launch Alliance’s Vulcan rocket to a SpaceX Falcon 9, a move that is as much a reflection of Vulcan’s mounting problems as it is a validation of SpaceX’s growing dominance in national security space launch. The GPS III Space Vehicle 09, originally contracted to fly on Vulcan this month, will now target a late April liftoff on Falcon 9, marking the fourth consecutive GPS III satellite the Space Force has moved to SpaceX after contracts were originally awarded to ULA.
The immediate trigger is a solid rocket motor anomaly that occurred on February 12 during Vulcan’s USSF-87 mission. Although the payloads reached orbit and ULA declared the mission successful, the company characterized the malfunction as a “significant performance anomaly” and has since paused all military launches on Vulcan pending a root cause investigation.
“With this change, we are answering the call for rapid delivery of advanced GPS capability while the Vulcan anomaly investigation continues,” said Systems Delta 81 Commander Col. Ryan Hiserote. “We are once again demonstrating our team’s flexibility and are fully committed to leverage all options available for responsive and reliable launch for the Nation.”
The broader reality is that SpaceX’s reliability record and launch cadence have made it the path of least resistance for the Pentagon, and bodes well with Elon Musk’s plans to IPO SpaceX sometime this year. Its Falcon 9 is the most flight-proven rocket in history, and the Space Force’s Rapid Response Trailblazer program was specifically designed to enable exactly this kind of provider swap for GPS missions, and effectively building SpaceX’s flexibility into the national security launch architecture by design.
For ULA, the stakes are existential. The company entered 2026 with aspirations of finally turning a corner after years of Vulcan delays, with interim CEO John Elbon pointing to a backlog of over 80 missions as reason for optimism. Meanwhile, SpaceX’s contracts with the Space Force have given it a formal pathway to take on even more national security launches going forward.
The significance of today’s announcement extends beyond one satellite swap. It reinforces that America’s most critical space infrastructure, including GPS, missile warning, and beyond, is increasingly dependent on a single commercial provider.