After all of the hype, Tesla has still not moved beyond small demonstrations and is not accumulating meaningful autonomous miles.
TL,DR:
All existing vehicle hardware (4 and lower) will not run newer, larger FSD models (10x parameter) because of insufficient memory (at least). Hence Tesla’s AV efforts have likely plateaued on existing vehicles and hardware in 2026 - so we can expect no big, scaling autonomy advances in 2026 for all existing Tesla corporate and retail fleets.
New AI5 hardware to run new, larger models may be available next year, mid-late 2027. So major improvements and scaling in autonomous driving capability and safety are unlikely until 2027, at the earliest.
Recent news (easy to google):
Tesla is discussing a new compact SUV, that might make use of their Cybercab manufacturing platform
Tesla released FSD software version V14.3 on hardware 4 (latest available at scale), with only modest, incremental improvements expected (“polishing” via point releases)
Tesla says the proposed large, “10x parameter” increase in the FSD model won’t arrive until software version 15
Tesla has not yet applied for autonomous testing permits in California
What does the above tell us?
Because new revenue from new AI initiatives like autonomous vehicles (including Cybercab) is nowhere in sight, and Tesla needs revenue and needs to use spare manufacturing capacity, hence the “AI company” is still seriously considering manufacturing of new vehicles, a materially non-AI endeavor.
The new, larger 10x parameter FSD model is still not yet implemented on hardware 4 and not expected until the new major software version 15. This suggests a newer, larger, more capable model cannot run on all existing hardware (4 and lower) without a hardware upgrade.
The above also suggest that substantial advances in autonomous driving won’t happen until new hardware AI5 is available in 2027 to run new, larger models.
All of the above is consistent with Tesla eliminating the option to purchase FSD outright and eliminating the promise (and liability) that current hardware can support autonomy.
Tesla autonomy is faltering.
We can see their struggle in their lack of scaling of Robotaxi and it’s consistent with the above product news. Of course, Tesla has not explicitly stated all of the above setbacks, but that is the writing on the wall.
[ceo] has confirmed that the highly anticipated “10x parameter” model won’t arrive until version 15.
The update came via X, where [ceo] explained the reasoning behind the delay. “Our rate of advancement with the small model has been so fast that the large model has not yet caught up. V15 will be the large model,”
How will all Tesla owners feel if their HW4 vehicles will never support autonomy?
That statement is very consistent with my observations for several years now. As I read through the quoted passages, it seems the situation is even worse than what I have believed previously. Why would anyone buy into this mess right now? This makes me compelled to sell which I have been resisting as I have been badly beaten up with capital gains tax (yes I know — one could have worse problems)
How many releases of FSD thus far have supposed to have been the one that is finally what everyone has been waiting for. I think there have been numerous ones since back with the myriad versions of Rel 12. My guess there are few that expect rel 15 to be any better
Is there a chance one could release the 10X parameter version on a car with hardware to support it, evolve it until it is working really well, and then intelligently prune the parameters so that one ended up with a version that would work on older hardware?
Wouldn’t that depend on what the issue is with the older hardware? If it’s just the chip capabilities, it might be theoretically possible. But if it’s a missing front bumper camera or having the “wrong” cameras (lower res or not self-cleaning) or something like that, then it might not be possible - or worth the effort.
Well, maybe, but understand that I am not thinking of a perfect solution, but rather just one that is a leap forward on what they have now on those cars.
Nobody seems to address the elephant in the room, so I will: robotaxi cannot happen with the hardware Tesla currently ships in its cars. This isn’t a matter of opinion it’s a basic engineering reality. The sensor suite isn’t there. Software, no matter how good, cannot make up for hardware that was never designed for full autonomy. Until Tesla addresses what’s actually in the vehicle, the robotaxi narrative is just that a narrative. If this was a software issue, it would have solved years ago, NO other manufacturer has as much driving data as tesla and yet….
That quote comes from a post in the “comments” section of an SA article.
The author of the comment IMO does a good job of summarizing Tesla shortcomings with both the capabilities of FSD when it comes to full autonomy. He was responding to a silly hype article on SA advocating buyimg at these prices.
Anybody who wants FSD for a Level 2 ADS, keep paying all you want for it – whether you paid in full - or hoping that monthly payment is worth it — just don’t keep dreaming that you may be able to sleep in the back seat someday unless issues eventually get addressed in a much better way than they have in these many past years. We stockholders deserve better..
