The Autonomous Driving Stack Wars Just Got Personal
Wayve, the London-based autonomous vehicle software company backed by SoftBank and NVIDIA, is making a calculated push into fleet deployment territory that Mobileye has long treated as its own. The move signals a deeper competition forming at the intersection of AI-native driving software and the commercial fleet market.
What Wayve Is Actually Building
Wayve’s core bet is an embodied AI approach – the idea that a driving model trained across diverse real-world environments can generalize better than rule-based or sensor-fusion systems that need exhaustive manual mapping. Where Mobileye’s EyeQ chips and REM mapping technology depend heavily on a proprietary data layer built over years of hardware integration, Wayve’s software stack is designed to run on existing vehicle hardware from multiple manufacturers. That architecture makes it attractive to fleet operators who want flexibility without being locked into a single silicon supplier.
The company’s LINGO model – its multimodal AI system for driving – marks a departure from how most autonomous driving software is structured. Rather than treating perception, prediction, and planning as separate modules with handoffs between them, Wayve trains a single end-to-end model that handles all three. This approach mirrors what has worked in large language model development: scale the data, scale the model, reduce the handcrafted rules. For fleets running vehicles across varied urban environments, the appeal is obvious – less reliance on pre-mapped roads, faster adaptation to new routes.
Wayve has been testing this architecture with delivery fleets in London, including reported trials with grocery and logistics operators. These aren’t robotaxi deployments aiming for level 4 autonomy on day one. They’re supervised deployments with safety drivers, accumulating the training data and commercial credibility needed to win longer-term contracts. It is an incremental strategy that mirrors how Waymo built trust with regulators before expanding driverless operations – but at a faster commercial tempo.
The funding backing this push is substantial. Wayve raised over $1 billion in a Series C round in 2024, making it one of the better-capitalized autonomous driving startups outside of the U.S. That capital is being used to expand the engineering team, grow the compute infrastructure needed for large-scale model training, and – critically – fund the kind of extended fleet pilots that convert into multi-year software licensing deals.
Why Mobileye Is Exposed Right Now
Mobileye built its dominance by being indispensable to OEMs. Its chips sit in hundreds of millions of vehicles, its ADAS technology became the default advanced driver assistance layer for brands across Europe and North America. But that dominance was built on a hardware-first model, and hardware-first models face a specific vulnerability when the competitive threat comes from software companies that don’t need to sell chips.
Mobileye’s fleet business – including its robotaxi technology under development and its SuperVision hands-free driving system – has faced execution delays and OEM relationship complications. Intel’s long saga of deciding what to do with Mobileye, culminating in a planned partial spinoff that stalled as Intel’s own finances deteriorated, created strategic uncertainty that affected Mobileye’s ability to pursue aggressive new fleet deals. A company managing investor expectations around a contested corporate structure is not in its best posture to fight off a well-funded software challenger.
There is also a generational technology question. Mobileye’s architecture was designed in an era when neural networks were components inside a larger system, not the system itself. Retrofitting an end-to-end AI approach onto a platform built around discrete computer vision modules is an engineering challenge – not impossible, but slow. Wayve does not carry that legacy weight. Its model was built end-to-end from the start, which gives it a structural development speed advantage even if Mobileye retains a deployment scale advantage for now.
Fleet operators themselves are signaling openness to alternatives. Several large logistics companies have begun running parallel evaluations of multiple ADAS and autonomous driving vendors rather than committing exclusively to one stack. The reasoning is straightforward: no operator wants a single-vendor dependency in a technology category that is still evolving rapidly. That procurement behavior creates exactly the kind of opening Wayve needs – not to displace Mobileye overnight, but to win pilot contracts that become references for the next deal.
The commercial dynamics here echo patterns seen elsewhere in B2B software, where a newer entrant captures mid-market clients through agility while the incumbent holds enterprise accounts through integration depth. Wayve is not yet threatening Mobileye’s existing OEM hardware relationships. It is going after the next wave of fleet software contracts before those deals are locked in.
The Contract Gap Wayve Is Targeting
The specific opportunity Wayve is pursuing sits between full Level 4 autonomy – which remains commercially limited to a small number of cities and operators – and basic ADAS, which is already commoditized. This middle tier, sometimes described as hands-free or supervised autonomy for commercial fleets, is where the next wave of deployable contracts will be awarded over the next three to four years. Fleet operators running last-mile delivery, grocery fulfillment, and urban logistics need technology that works today, not a roadmap to full driverless operation in 2030.
Wayve’s pitch for that tier is a software licensing model with an ongoing data feedback loop – fleets pay for the software, the software gets better as it collects more miles, and the improving model justifies contract renewals. Whether that model generates the margins needed to sustain a $1 billion-funded operation at scale is the unresolved question. Mobileye’s hardware margins are thin in consumer ADAS but substantial at scale; Wayve’s software margins could be higher, but only if the licensing fees hold up against competitive pressure from the next well-funded contender that decides to enter the same space.
