Momenta Just IPO’d. It Doesn’t Belong in My Robotaxi Basket.
Momenta Global (HKEX: 6880) listed in Hong Kong on 8 July. The IPO was priced at HK$295.60 and raised about US$751 million, with retail demand oversubscribed more than 400 times and institutional orders well past HK$100 billion. The stock opened at HK$301, spiked intraday to HK$314.80, then drifted back to close near HK$299 — a gain of roughly 1%, essentially flat.
That gap is the interesting part. Demand for the deal was enormous, but once the stock started trading freely, the response was much more muted. Investors clearly like the technology story. The harder question is who ultimately owns the economics.
That’s the question I keep coming back to with autonomous driving as a theme, and it’s why Momenta isn’t in my basket alongside PONY, WRD, XPEV, GRAB, UBER and HSAI.
The Android comparison doesn’t hold up for me
Momenta’s pitch is straightforward: it doesn’t build cars, it licenses driving software to automakers who do. Toyota, Mercedes-Benz, GM, BYD and others are already customers, and nine of the world’s top ten automakers reportedly work with Momenta in some form. No factories, no capex on vehicles, just software sold across a widening base of partners.
On paper, that sounds like the Android of autonomous driving. I’m not convinced it plays out that way.
Tesla owns the car, the software, the customer relationship and the data loop end to end. If you buy a Tesla, part of what you’re buying is Full Self-Driving — it’s part of the brand’s identity, not just another feature. XPeng is building towards the same thing. When the customer sees the car as inseparable from its driving software, the company that owns both the car and the software has a much stronger claim on the value that autonomy creates.
Momenta doesn’t have that. If Toyota’s driving experience gets better, Toyota gets the credit — and Toyota, not Momenta, owns the customer. Momenta is the supplier behind the curtain.
That’s a perfectly viable business. Plenty of automotive suppliers have done very well operating exactly like this. But it’s a different risk profile from owning the full stack, and I don’t think the two should automatically be valued the same way. It also raises a question I keep circling back to through the rest of this piece: how difficult would it actually be for one of these automakers to replace Momenta once the current generation of vehicles is designed?
If autonomous driving defines the brand, why outsource it?
This is the part I struggle with most.
Mercedes-Benz already uses Chinese technology in its vehicles, so working with a Chinese software company isn’t unusual by itself. But I don’t think all software is equally strategic.
A display system is one thing. The autonomous driving system that could eventually determine how safely, smoothly and intelligently a Mercedes drives is something else entirely. If several automakers can all buy similar capability from the same supplier, what exactly differentiates one brand from another?
Tesla’s FSD is already part of Tesla’s identity. XPeng is building towards the same thing. If autonomous driving eventually becomes one of the main reasons consumers choose one car over another, I’d expect Mercedes, Toyota and every major automaker to think hard before permanently outsourcing that differentiation to Momenta.
I owned a Mercedes-Benz EQA 250 for a while. I tried the auto-parking feature once. It felt unsafe, and I never used it again.
I don’t know whose system was actually behind that feature, and that’s partly the point. As a customer, I blamed the Mercedes experience, not whichever supplier may have provided the underlying technology. If autonomous driving becomes much more central to why people choose a car, I would expect Mercedes to want tight control over that experience rather than simply becoming one of several automakers buying a similar stack from the same supplier.
This doesn’t mean every automaker will successfully build its own system. World-class autonomous driving technology is enormously expensive and difficult, and some may decide outsourcing is the only economical choice. But that’s precisely the tension I see in Momenta’s business model.
The more valuable its technology becomes, the more strategically important it becomes to its customers — and potentially the more uncomfortable those customers become depending on it.
That’s very different from supplying seats, batteries or display software.
Can Momenta actually learn across its customers?
The bull case for Momenta rests partly on a data flywheel: more automakers using its software means more driving data across brands, which in theory compounds into better models than any single automaker could build alone.
“Who owns the data” isn’t quite the right question, though. The more important question may not be who legally owns the raw data, but what Momenta is contractually allowed to learn from it.
The question that actually matters is closer to the one enterprise AI vendors deal with: does the contract let Momenta train its shared models on data generated from Toyota’s fleet, or does Toyota insist on an isolated setup where its driving data never touches the model also learning from Mercedes or BYD? Most enterprise LLM agreements default to no cross-customer training unless the customer explicitly opts in, precisely because no company wants its own data improving a system that’s also getting smarter for its competitors.
For Momenta, that distinction is the whole thesis. The flywheel story only works if automakers have granted broader pooling and training rights than a typical enterprise software deal would include. If each automaker has instead insisted on siloed training, keeping its own fleet’s data walled off from the model serving its competitors, then Momenta is really running several separate systems under one brand, and the cross-automaker learning advantage the bull case depends on may not exist in practice. I haven’t found a clear answer on which of these it is in the public disclosures I’ve reviewed.
Geopolitics caps the realistic market
Momenta has close to two dozen automaker partnerships, and the theoretical addressable market — every automaker without its own leading autonomy stack — looks huge. I don’t think the realistic market is anywhere near that size.
Autonomous driving software touches cameras, location data, driving behaviour and increasingly the core intelligence of the vehicle. It’s a different category from sourcing seats or displays. This isn’t just a hypothetical risk either. A US Commerce Department rule bans Chinese-developed software from connected and autonomous vehicles sold in the US starting with the 2027 model year, with a hardware ban following from 2029. The Trump administration confirmed in April it would retain these bans, with only case-by-case exceptions granted so far.
China is the obvious growth market. Parts of Europe may open up. Momenta also has a partnership with Grab aimed at deploying robotaxis in Southeast Asia, and that’s a market I’m watching closely given GRAB and UBER are already in my basket. But the broader point remains: Momenta’s realistic addressable market may look quite different from the global platform story implied by having relationships with nine of the world’s ten largest automakers. That matters when thinking about valuation.
But what if Momenta becomes better than everyone else?
There is another possibility I can’t dismiss.
I’ve been looking at Momenta mainly from the perspective of the automakers. If autonomous driving becomes central to why someone buys a Tesla, XPeng or Mercedes, why would these companies want to outsource that technology? But perhaps I’m asking the question the wrong way around. What if Momenta becomes so good that automakers have little choice but to use it?
Tesla learns from Tesla vehicles. XPeng learns from XPeng vehicles. But if Momenta can genuinely learn across Toyota, Mercedes-Benz, BYD and other manufacturers, its supposed weakness — not owning the customer or manufacturing its own cars — could eventually become its greatest advantage. A cross-automaker data flywheel could theoretically become larger and more diverse than anything a single automaker can build alone. If that produces a system demonstrably safer and more capable than Tesla’s FSD or XPeng’s autonomous driving technology, the moat shifts: Tesla’s advantage stops being autonomous driving the moment Toyota, Mercedes and ten other manufacturers can simply buy something better from Momenta.
In that scenario, Momenta doesn’t just participate in the autonomous driving theme. It potentially commoditises everyone else’s autonomous driving advantage — including the two names, Tesla and XPeng, whose full-stack ownership I’ve used throughout this piece as the standard Momenta doesn’t meet.
I don’t think we’re there today. Autonomous driving remains a genuine differentiator, and I don’t see Tesla, XPeng or other leading players giving up their own systems anytime soon. But over time, perhaps AD follows the same path as batteries: once a major competitive advantage, eventually something many manufacturers can source at a sufficiently high standard from specialist suppliers.
There’s also another possible outcome. If Momenta’s technology becomes genuinely superior, perhaps the endgame isn’t remaining independent at all — a major automaker could decide it’s easier to acquire the technology than spend years trying to replicate it. A carmaker might not acquire Momenta simply because it wants the technology. It might acquire Momenta because it doesn’t want that technology becoming available to every competitor. The problem is that an acquisition by one automaker could weaken the very neutrality that makes Momenta attractive to everyone else. If Toyota owned Momenta, would Mercedes or BYD still want to depend on it? I don’t know.
But it makes the Momenta story more interesting than simply deciding whether it’s a good or bad supplier. The real question may be whether Momenta remains one layer behind the automakers — or eventually becomes the layer that makes their own autonomous driving moats less valuable.
Where this leaves me
“Exposed to a theme I believe in” and “a stock I want to own” are two different tests.
I didn’t build my robotaxi basket around a clean framework from day one. PONY, WRD, XPEV, GRAB, UBER and HSAI were different ways of buying into the same theme, and I’m only now developing a more rigorous way to tell them apart. Momenta is useful precisely because it’s forcing me to ask harder questions about who actually captures the value when autonomous driving succeeds.
For now, I still see a structural disadvantage in being the supplier rather than owning the vehicle and customer relationship. But I can’t treat that conclusion as settled either. If Momenta can genuinely learn across multiple automakers and build a system demonstrably better than what Tesla, XPeng and others can develop alone, then its supposed weakness could become its greatest strength — and the autonomous driving moat I value in some of those companies may not be much of a moat after all.
I haven’t seen enough evidence of that yet.
So I’ll keep watching Momenta. It’s not a position for now.
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.