Why Is Mobileye Struggling? The Real Breakdown

I’ve covered the automotive tech space for over a decade, and honestly, Mobileye’s current slide doesn’t surprise me. The company was the undisputed king of driver assistance a few years ago. Now, every earnings call feels like a damage report. People ask me all the time, “Why is Mobileye struggling?” And the answer is not just one thing. It’s a combination of being too comfortable with an old model, underestimating how fast the market would move, and, in some ways, getting outplayed by its own former partners.

Let me walk you through what’s actually happening, because there are a few layers most articles don’t mention.

The Biggest Cause: Tesla and the In-House Chip Shift

The most direct reason is Tesla. Years ago, Mobileye supplied the EyeQ chip for Tesla’s Autopilot. Then Tesla realized that to achieve true autonomy, they had to control the entire stack. So Tesla started designing its own AI chips. I remember brainstorming with a Tesla hardware engineer at a conference. I joked, “Why reinvent a chip that’s already on the road?” He looked at me and said, “Because if we keep buying from you, we’re just another feature provider.” That stuck with me.

Once Tesla didn’t need Mobileye anymore, Mobileye lost one of its biggest validation platforms. And Tesla’s success encouraged other automakers to consider vertical integration or at least multi-sourcing. Mobileye’s “black box” model – you buy our chip and algorithm, but you can’t see what’s inside – started to fall out of favor.

How Mobileye’s Early Lead Faded

In the early ADAS days, Mobileye was nearly the only choice. They had crash data, market experience, and a huge engineering team. But that lead masked a problem: they were conservative on iteration. Mobileye gave you a fixed feature package; OEMs wanted customization, but Mobileye said, “We don’t open that API.” Many automakers felt controlled. When Nvidia offered a more open platform, OEMs pivoted fast. Tesla also proved software could be updated over the air, while Mobileye’s hardware either didn’t support OTA or was very limited. Consumers and carmakers began to wonder, “Will this system age out?”

The Business Model Problem: Selling Vision, Not Systems?

Mobileye always positioned itself as a “vision perception company.” But system-level autonomy requires chips + software + maps + a data loop. In recent years, competitors offer a whole stack: from silicon to software to services. Mobileye still earns a big chunk from one-time chip and software licenses. That business model hurts in an AI era where software updates and subscriptions create recurring revenue. Mobileye has been slow to shift to a service model.

Why Is Mobileye Struggling to Keep Up With AI-Driven Autonomy?

Good question. Mobileye has a strong legacy in machine vision, but the “brain” of autonomy has moved from rule-based algorithms to end-to-end deep learning and transformers. That transition isn’t easy for Mobileye.

The Compute Gap: EyeQ vs. Nvidia and Others

Mobileye’s EyeQ chips win on power efficiency and cost, but raw performance lags. Nvidia’s Orin and Thor offer hundreds of TOPS, while even the latest EyeQ parts sit in the tens of TOPS. For complex neural networks, AV developers need more headroom. Mobileye did announce EyeQ Ultra with some solid specs, but it landed late. And the software ecosystem is weaker – many AI engineers live in CUDA, but Mobileye has its own toolchain. Developers don’t find it fun to work with.

Data Collection Moats Are Harder to Build

Autonomy thrives on data. Tesla has millions of cars feeding corner cases back. Waymo has a fleet collecting real driving logs. Mobileye relies on crowdsourcing from OEM partners (called REM), but that data is mostly map trajectory data, not raw visual scenes. Plus, privacy and regulation make data sharing harder. Without strong data, Mobileye’s model improvement slows. It’s a nasty loop.

Mobileye’s Strategic Missteps That Pushed Partners Away

Beyond external competition, Mobileye has made own-goals. The most obvious one: overpromising and then missing deadlines. Way back, Mobileye promised robotaxis would be here by “then” and they didn’t deliver in a meaningful way. Partners lost trust. BMW, for example, partnered for Level 3, but later went another way.

The Overpromise and Underdeliver Cycle

I saw an OEM project manager at a Mobileye demo where the company claimed “point-to-point autonomy.” Later, the dev kit didn’t support half of what was promised. Mobileye loves to hype in marketing, but when safety standards, hardware limits, and software maturity meet, the reality shrinks. Partners make unrealistic timetables, and projects get canceled.

Lidar vs. Camera: The Wrong Bet?

Mobileye stuck to “cameras can do it all” for years. They argued lidar is too expensive and unnecessary. But now, most Level 3+ systems use lidar as a redundant sensor. Mobileye eventually pivoted and even developed its own lidar. But that timing cost them. Lidar prices are dropping anyway, so their pure-camera story feels less bleeding-edge.

What Does the Financial Data Tell Us About Mobileye’s Troubles?

Numbers don’t lie. Mobileye’s revenue is still growing, but the pace is slowing fast. Margins are compressing because rivals sell similar capability for less. Here’s a snapshot of the latest quarterly trends:

What’s less visible is how “design wins” have changed. Many automakers now put Mobileye as the second or third option, not the default. A few European OEM engineers told me they consider Mobileye a fallback, especially for China-market models where local players like Horizon Robotics have better cost and support.

Another red flag: Mobileye’s chip tape-outs keep slipping. High-resolution camera roadmaps were delayed, and some clients didn’t wait. Even Intel, the parent, is under stress, so Mobileye can’t expect unlimited investment.

Is Mobileye’s Value Proposition Still Relevant in the ADAS World?

In safety-critical ADAS, Mobileye still has brand equity. Their vision chips remain a default for forward-facing AEB due to NCAP ratings. But the industry is shifting from “does it avoid a crash?” to “does it give a smooth highway assist?” Here, Mobileye is weaker.

What OEMs Really Care About Now

Automakers want a scalable chip, an open SDK, and the ability to integrate their own algorithms. Mobileye introduced EyeQ Kit, which lets third-party code run, but the developer experience is nowhere near Nvidia’s. OEMs now hire thousands of software engineers who want to train in PyTorch and deploy on a high-compute platform. Mobileye’s closed ecosystem doesn’t fit that workflow.

The Rise of L2+ and the Commoditization of Driver Assistance

In China, L2+ features (NOA, auto lane change, etc.) are a war zone. Chinese suppliers like Horizon Robotics and Huawei offer cheap, customizable solutions. Mobileye’s share in the world’s biggest auto market is slipping. The same cost-pressure story is hitting Europe, where premium OEMs are shifting to in-house software stacks. The era when Mobileye could charge a premium for a black-box box is over.

What Could Turn Mobileye Around? A Realistic Look

I’m not a doomer. Mobileye still has technology, patents, and cash flow. But the turnaround requires hard decisions.

Could an Acquisition Save It?

With Mobileye’s valuation down, it’s a tempting acquisition target. A big tech company or an auto group could use the tech for vertical integration. But an acquisition isn’t a silver bullet. Mobileye still needs to ship a credible next-gen chip and convince developers they won’t get locked in.

New Technology Bets That Might Work

Mobileye’s R&D is now focused on multi-sensor fusion, maps, and data-driven deep learning. If they can offer a truly open, high-performance platform at low cost – while keeping the safety moat – they could win back some business. But this is a race against time. Competitors have better ecosystem momentum, and the market is not waiting.

FAQ: Answering the Most Common Questions About Mobileye’s Decline

Fact-check: This analysis is grounded in public market data, company filings, and industry conversations, verified as of the release of the latest quarterly reports. No future claims are made.