The Company Kept Winning. I Started Questioning The Moat.
I sat down to write this thinking I needed to work out whether LiDAR still had a future.
That’s not actually where I landed.
If anything, I feel more sure about LiDAR now than when I started digging. What I’m less sure about is whether Hesai is the one who gets to keep the money from it.
When I first bought the stock, it felt simple. Robotaxis were coming, most autonomous vehicles would need LiDAR, and Hesai (HSAI) was the clearest way to ride that. I wasn’t buying it for any one quarter’s numbers. I was buying it because I believed the roads would look different in five years’ time. Somewhere along the way I’d quietly assumed that if LiDAR became indispensable, Hesai would become indispensable right along with it. I never really stopped to check whether that second part was actually true.
On the surface, that thesis has not gone badly at all.
Hesai shipped 471,723 LiDAR units in Q1 2026, up 140.9% from a year earlier. Revenue rose 29.6% to RMB680.6 million, the company turned a GAAP profit, and it landed a strategic supplier agreement with Mercedes-Benz for L3 models across Europe and China. The robotaxi industry itself is moving forward too, with Pony AI and WeRide expanding commercial operations. This is no longer a story told through demo videos and management slides.
If I only looked at execution, I probably wouldn’t be writing this.
The problem is that execution was never the whole thesis. What’s shifted is my assumption about who would need LiDAR, and whether needing LiDAR necessarily meant needing Hesai.
LiDAR surviving may not be enough
When I first invested, I more or less assumed autonomous driving and LiDAR would develop together.
That’s less certain now.
Tesla has stayed firmly committed to a camera-only approach, and its robotaxi service is no longer just a demo. It now operates across several US cities, with the company reporting more than 380,000 miles of unsupervised driving across six of them as of this week’s Q2 earnings call. The rollout remains limited, and most of the fleet is still supervised, but vision-only can no longer be dismissed as something that only works on a presentation stage. XPeng has gone further still, publicly reaffirming a pure vision approach in May, with its CEO stating plainly that the auto industry no longer needs the sensor. XPeng was never a meaningful Hesai customer, so none of this hits current revenue directly.
But it still matters.
If Tesla and XPeng can prove vision-only works at scale, in real deployment rather than on a stage, the market may stop assuming every successful autonomous vehicle needs LiDAR by default. Hesai can keep growing and still find its ceiling priced lower than I once assumed.
At the same time, the industry isn’t converging the way that “vision wins” story would suggest.
Unitree’s humanoids use 3D LiDAR together with depth cameras, and Hesai’s JT128 equipped the Unitree robots used in this year’s Spring Festival Gala performance. Honor’s “Lightning” humanoid, which won this year’s Beijing robot half-marathon, also used Hesai’s JT128. The real picture isn’t LiDAR versus cameras. It’s different companies choosing different systems for different jobs.
That was actually the moment this piece changed direction on me. I’d been holding up Unitree as proof that Hesai had a real foothold in humanoids. But Unitree also builds its own LiDAR for some of its products. So the same company validating Hesai’s opportunity is also showing that it does not intend to depend on Hesai for everything. If Hesai’s technology gave it a truly durable hold over the customer, why would one of its best-known users still be developing alternatives of its own, even for simpler applications? That’s the question that stuck with me, and it’s a different question from whether LiDAR itself survives.
I had been treating Hesai’s robotics business as something that might matter in a few years. That was too dismissive. Robotics already made up roughly a quarter of Q1 shipments, and Unitree, Honor, and the lawn and delivery robot orders from Dreame and MOVA are named, current customers, not a pipeline promise. The opportunity is real. Whether Hesai retains enough of it, and earns attractive margins from it, is the part I’m no longer sure about.
Tesla’s Optimus is also a reminder not to get ahead of myself. Vision-only hasn’t been proven at humanoid scale yet, so the opportunity for Hesai remains real, just unsettled in either direction.
The number I keep coming back to
The part that concerns me most isn’t the cameras-versus-LiDAR debate at all.
It’s gross margin.
Hesai’s Q1 gross margin came in at 39.1%, down from 41.7% a year earlier, even as unit shipments more than doubled. One quarter alone wouldn’t be enough to call this commoditisation, but management says the decline is deliberate, with its in-house chip work allowing it to pass on cost savings and open a much larger market. I want to believe that story, but I’m not fully buying it. If the moat were strengthening, I’d expect Hesai to retain at least part of those cost gains. Instead, shipment growth is accelerating while revenue grows much more slowly and margin falls, and that makes me wonder whether the efficiencies are accruing to Hesai or being competed away for its customers’ benefit.
RoboSense, Hesai’s closest rival, makes me wonder even more. Its own margin climbed from single digits to over 20% through 2025, and it posted its first profitable quarter in the fourth quarter, over the same broad period Hesai’s margin has been sliding. Both companies also launched competing chip platforms within weeks of each other this year. To me, that looks less like one company generously sharing efficiency gains with the market and more like two companies racing each other on price, with Hesai the one currently giving up more margin while doing it.
Hesai could still end up the largest player in a very large market and still be a disappointing stock, if the margin race with RoboSense keeps going the way it’s going. That’s the bigger risk to me. Not whether LiDAR disappears.
And that’s really the same doubt from the Unitree moment, just showing up in the numbers now. My honest read is that Hesai’s moat may depend more on cost and manufacturing scale than I originally gave it credit for. The technology could still be genuinely good. It just may not be so irreplaceable that a customer like Unitree stops looking for its own version anyway.
Two questions, not one
When I bought Hesai, I thought I only needed one thing to be true. Robotaxis had to become real.
I didn’t realise I was quietly also betting on a second thing. That if LiDAR mattered, Hesai would be the one who got paid for it mattering. I used to think those two things were basically the same bet. Writing this made me see they’re not, and that at a valuation of roughly 5 times trailing sales, I don’t have much room to be wrong on the second one even if I’m completely right on the first.
I still think Hesai is a good company. Execution has been strong, its market position is real, and the robotics business is more substantial than I gave it credit for even a few weeks ago. The company hasn’t gotten worse. If anything, on the numbers, it’s gotten better.
I’m leaning toward selling anyway.
It is not financial advice. The author may hold positions in securities discussed. Readers should conduct their own due diligence and consult with a qualified financial advisor before making investment decisions.