After years of confident predictions about imminent robotaxis, the Tesla boss has admitted that many cars sold as “Full Self‑Driving ready” will actually need costly hardware upgrades before they can run the company’s next‑generation autonomy software.
A promise that has aged badly
Back in the mid‑2010s, Musk repeatedly claimed that every new Tesla rolling off the line had all the hardware needed for full autonomy. Owners just had to wait for the software to catch up. Many paid thousands of dollars extra for the “Full Self‑Driving” (FSD) package on that basis.
At the end of January 2025, during a call with major shareholders, Musk finally walked that narrative back. Speaking about cars equipped with Tesla’s Hardware 3 computer, he acknowledged that they will not be able to run the upcoming self‑driving system as promised.
He said Tesla will have to upgrade Hardware 3 computers for customers who bought FSD, calling the process “painful and difficult” but necessary.
That single admission cuts directly against years of marketing and conference‑stage optimism. It also raises an awkward question for regulators and investors: if the hardware was never fully ready, what exactly were FSD buyers paying for?
How Tesla’s hardware story keeps shifting
Tesla’s autonomy pitch has always leaned on a simple idea: buy the car now, and it will magically get smarter later with software updates. Under the bonnet, though, the computer platform has changed multiple times.
From hardware 2 to hardware 4
Tesla’s driver‑assist evolution roughly looks like this:
- Hardware 2.0 / 2.5 – Introduced around 2016–2017, sold as capable of “full self‑driving” in the future.
- Hardware 3 – Rolled out from 2019, described as the dedicated FSD computer, with Tesla offering upgrades to some earlier owners.
- Hardware 4 – Now in current production, more powerful, with revised camera placements and extra computing headroom.
In 2019, when Hardware 3 arrived, some owners with Hardware 2.0 and 2.5 were told they needed a new computer to unlock the full FSD potential. That already triggered lawsuits accusing Tesla of misleading advertising.
Now the same pattern appears again, this time with Hardware 3 at risk of being sidelined by more demanding AI models and sensor data flows.
Repeated hardware revamps are starting to look less like future‑proofing and more like a moving target for owners who bought into long‑term promises.
What Musk actually admitted
On the investor call, Musk did not mince words about the strain of upgrading earlier cars. He framed the plan as a kind of necessary damage control: Tesla will retrofit the onboard computers of customers who paid for FSD but only have Hardware 3.
The phrase “painful and difficult” hints at several headaches at once:
- Cost: Swapping computers on a large fleet can run into hundreds of millions of dollars in parts and labour.
- Logistics: Cars are spread across states and countries, requiring service capacity that Tesla already struggles with in busy markets.
- Legal risk: If upgrades are slow or incomplete, frustrated owners could revive or expand claims of deceptive marketing.
Musk also implied the company has little choice. Without upgrades, a sizeable group of FSD buyers may never see the level of autonomy they were sold on, just as competitors ramp up their own driver‑assist systems.
Customers caught in a long wait
For early adopters, the frustration is not only technical; it is emotional and financial. Many bought FSD as a kind of ticket to the future, encouraged by Musk’s timelines for robotaxis “next year” that slipped again and again.
In the US, the FSD option has cost well over $10,000 at different points, or a monthly subscription running into hundreds of dollars annually. Some owners budgeted for that on the assumption that the car would eventually drive itself most of the time, possibly even earning money as a robo‑Uber when they were not using it.
Instead, what they have today is an advanced driver‑assist suite that still requires constant supervision and, in many regions, remains officially a beta product.
Current Teslas roll off the line with Hardware 4, yet the fully autonomous mode Musk has long teased still has no clear launch date. For those who paid upfront years ago, patience is starting to thin.
Legal storm clouds and financial stakes
Regulators have already taken notice. In several jurisdictions, authorities have questioned whether the names “Autopilot” and “Full Self‑Driving” overstate the technology’s real‑world capability.
Previous legal actions around Hardware 2 upgrades showed that courts are willing to hear complaints when reality fails to match marketing slides. Musk’s latest comments could fuel a new wave of scrutiny, since they implicitly admit that the promised hardware sufficiency was not accurate.
On the financial side, Tesla faces a difficult calculation: invest heavily to upgrade legacy cars and try to preserve trust, or quietly narrow eligibility and risk further angering a vocal and digitally savvy customer base.
| Issue | Risk for Tesla | Risk for customers |
|---|---|---|
| Hardware upgrades | High capital outlay, service bottlenecks | Long waits, uncertain eligibility |
| Legal actions | Fines, settlements, reputation damage | Complex claims process, unclear remedies |
| Technology delays | Loss of perceived leadership | Paid features not matching expectations |
What “full self‑driving” really means today
Part of the confusion comes from language. The phrase “Full Self‑Driving” suggests a car that can handle every situation on its own. In industry terms, that would be Level 4 or Level 5 automation.
Right now, Tesla’s system is closer to Level 2: the car can steer, brake and accelerate, but the human driver has to stay alert and ready to take over at any moment. Cameras watch the road; in many cases there is no lidar or radar backup, by design.
This gap between name and function has led safety advocates to argue that FSD and Autopilot encourage overconfidence. Musk’s fresh admission about hardware limitations underlines how far the company still has to go to reach true autonomy.
How this could play out for owners
For a Tesla driver with Hardware 3 and a paid‑up FSD package, the realistic near‑term scenarios look something like this:
- They receive a free or heavily discounted hardware upgrade, but only after long waiting lists and limited service slots.
- Tesla restricts upgrades to certain markets or vehicle ages, leaving some owners arguing they have been left behind.
- Class‑action settlements in some countries offer partial refunds or credits instead of physical upgrades.
There is also the technical question of how much day‑to‑day driving will actually change after the upgrade. Many safety rules require continuous human supervision, which caps the perceived benefit of a powerful new computer in the dash.
Why AI demands keep breaking “future‑proof” hardware
From a computing perspective, Tesla’s problem is not unique. Modern AI models, especially those for vision and planning in traffic, are becoming more complex and data‑hungry every year.
A chip that looked more than adequate in 2019 may struggle with 2025’s neural network workloads. Frames per second, camera resolution, and the number of parameters in the model all climb rapidly. That leads to harder choices: either simplify the algorithms to fit older hardware, or push performance and drop support for legacy platforms.
Car buyers, though, do not think in three‑year chip lifecycles. They think in ten‑year ownership cycles. When marketing promises collide with that reality, upgrade programmes like the one Musk hinted at become almost inevitable.
What other carmakers are watching
Traditional manufacturers and tech rivals will study this episode closely. Many are rolling out their own advanced driver‑assist systems, and some are tempted to promise software‑based upgrades down the line.
Tesla’s struggle shows how sensitive those promises can be. Once a company claims that today’s hardware is enough for tomorrow’s autonomy, it risks inheriting an open‑ended obligation to keep that hardware relevant. That has real costs, both financial and reputational, when AI expectations keep climbing.
For drivers looking at a new EV or hybrid, this saga offers a practical lesson: view “future” software features as a bonus, not a guarantee. Check what the car can do on the day you buy it, and treat everything else as a bet on how honest and realistic the manufacturer will be when the hard engineering trade‑offs arrive.
