Tesla FSD, Waymo, and Mobileye represent three fundamentally different bets on how self-driving cars should actually work — camera-only end-to-end AI, sensor-redundant geofenced robotaxis, and a chip-and-software supplier model — and in 2026 the gap between their approaches has become a real-world business divergence, not just a technical debate.
I've followed all three closely through their 2026 announcements, ridden in Waymo vehicles, and tracked Tesla's FSD rollout data. Here's the honest state of the race — including a major strategic pivot from Mobileye that happened this week and changes the shape of this comparison entirely.
Three Different Bets, Not Three Competing Products
It's worth being precise about what's actually being compared. Waymo operates its own commercial robotaxi service — you can hail a fully driverless ride today in multiple US cities. Tesla FSD is sold as a feature on consumer vehicles that still legally requires driver supervision in most contexts, alongside a smaller, separately operated robotaxi pilot. Mobileye, until this week, was purely a technology and chip supplier to automakers — it didn't operate any consumer-facing service at all. As of June 16, 2026, that changed: Mobileye announced it will become a vertically integrated robotaxi operator too, launching its own driverless service in a US city in 2027. All three are now competing in the same eventual market, but they're starting from completely different positions.
Quick Overview of Each Approach
Tesla FSD uses an end-to-end, vision-only approach — cameras and neural networks, no LiDAR, no radar in current hardware. The bet is that a system trained on enough real-world driving data can match or exceed sensor-redundant systems at a fraction of the hardware cost — Tesla's camera-based sensor suite costs roughly $400 per vehicle versus $10,000–12,000 for a comparable LiDAR-equipped Waymo vehicle. FSD remains classified as SAE Level 2 (driver assistance) in its consumer release; Tesla's separate robotaxi pilot in Austin and San Francisco operates with reduced geofencing but, as of mid-2026, the majority of rides still include a human safety monitor. More at tesla.com/autopilot.
Waymo, Alphabet's self-driving unit, uses a sensor-redundant approach combining cameras, radar, and LiDAR, paired with detailed pre-mapped routes within defined geofenced areas. As of mid-2026, Waymo provides roughly 500,000 paid driverless rides per week across 10 US cities — Phoenix, San Francisco, Los Angeles, Austin, Atlanta, Miami, Dallas, Houston, San Antonio, and Orlando — operating a fleet of roughly 2,500+ vehicles with no safety driver on board. The company raised $16 billion in early 2026 at a $126 billion valuation and is targeting over 1 million weekly rides by the end of 2026, with expansion into 20+ additional cities including international launches in Tokyo and London. More at waymo.com.
Mobileye built its business as the dominant ADAS chip and software supplier to traditional automakers — its EyeQ system-on-chip and SuperVision/Chauffeur platforms power driver-assistance and automated driving features in vehicles from Volkswagen, Porsche, Polestar, and others, with deployment slated across 17 new models from multiple global brands starting in 2026. Until this week, Mobileye's entire business model was B2B technology licensing, not consumer-facing robotaxis. On June 16, 2026, the company announced a strategic pivot: it will launch its own vertically integrated robotaxi service, starting with roughly 100 driverless vehicles in an unnamed US city in 2027, while continuing its existing automaker supply business in parallel. More at mobileye.com.
Comparison Table
| Feature | Tesla FSD | Waymo | Mobileye |
|---|---|---|---|
| Core approach | End-to-end vision-only AI | Sensor fusion (camera, radar, LiDAR) + HD maps | Camera-primary + radar, supplier-built stacks |
| Business model | Consumer feature + emerging robotaxi pilot | Commercial robotaxi operator | Tech/chip supplier + new robotaxi operator (2027) |
| Current SAE level (consumer) | Level 2 (supervised) | Level 4 (driverless in geofenced areas) | Level 2–3 (SuperVision/Chauffeur, varies by automaker) |
| Driverless commercial service today | Limited — most rides include safety monitor | Yes — ~500,000 rides/week, no safety driver | Not yet — planned for 2027 |
| Cities with paid driverless rides | 2 (Austin, San Francisco — partial) | 10 US cities, expanding to 20+ globally | 0 (planned: 1 US city in 2027) |
| Sensor cost per vehicle | ~$400 (cameras only) | ~$10,000–12,000 (LiDAR-equipped) | Mid-range (camera + radar, no LiDAR in SuperVision) |
| Geofencing | No — operates on highways and city streets broadly | Yes — defined service areas per city | TBD for robotaxi service; supplier products vary |
| Price per mile (robotaxi) | ~$0.81/mile | ~$1.36–1.43/mile | Not yet operational |
| Average wait time | 15+ minutes | ~5.7 minutes | Not yet operational |
| Automaker partnerships | None (Tesla-only) | Zeekr (vehicle supply), Uber (platform) | Volkswagen, Porsche, Polestar, Ford, Mahindra, and others |
| Regulatory approval scope | Limited — driverless permits in few markets | Driverless permits in 20+ markets | Operates under existing ADAS/L3 regulatory frameworks |
| Safety data published | Limited public crash-rate data | 127M+ autonomous miles; 90% reduction in serious-injury crashes vs. human drivers (company-reported) | Responsibility-Sensitive Safety (RSS) framework; supplier-level data |
Tesla's Bet: Cheap Hardware, Software-Defined Scale
Tesla's central argument is economic, not just technical: if a vision-only system trained on enough real-world driving data can reach acceptable safety performance, the cost advantage over LiDAR-equipped competitors is enormous and compounds at scale — both in hardware bill-of-materials and in the simplicity of deploying the same system across every vehicle without geofencing or HD-mapping a city in advance. Tesla has continued shipping FSD updates rapidly through 2026, with versions like FSD v14.2.2.2 reported to improve lane-change smoothness and decision-making.
The honest gap between Tesla's marketing and its regulatory status is significant. As of mid-2026, FSD's consumer release remains classified as SAE Level 2 — meaning the driver is legally required to supervise the system at all times, regardless of how capable it appears in practice. Tesla's separate robotaxi service, launched in limited form in Austin and expanded toward San Francisco and other cities through 2026, operates a much smaller fleet than Waymo's, and independent reporting indicates the majority of those rides still carry a human safety monitor in the vehicle. Tesla has targeted seven cities for robotaxi service in the first half of 2026 but had several markets still pending as of May.
Where Tesla's approach shows a genuine edge: cost and flexibility. The roughly $400 camera-only sensor suite versus $10,000+ for LiDAR systems is a real, structural cost advantage that scales across every vehicle Tesla sells — not just a dedicated robotaxi fleet. Tesla also doesn't pre-map or geofence its operating areas the way Waymo does, meaning FSD in principle works on any road the car encounters, including highways, which Waymo's robotaxi service does not currently service.
Waymo's Bet: Prove It Works Before You Scale It
Waymo's strategy has been the inverse of Tesla's: move slowly and conservatively in a small number of well-mapped, sensor-redundant environments, prove the safety case with extensive data, then expand methodically. That approach is now paying off commercially. Weekly paid rides grew roughly tenfold in under two years — from approximately 50,000/week in May 2024 to approximately 500,000/week by early-to-mid 2026 — across a fleet that has scaled past 2,500 vehicles. The company has completed over 20 million lifetime rides and reports more than 127 million fully autonomous miles driven.
The expansion pace through 2026 has been aggressive specifically because the unit economics appear to be working: Waymo is both deepening its presence in existing cities (Phoenix, San Francisco, Los Angeles, plus newer additions like Austin, Atlanta, Miami, Dallas, Houston, San Antonio, and Orlando) and launching in entirely new markets (Detroit, Las Vegas, San Diego, Denver, Nashville, Washington D.C.) while preparing its first international markets in Tokyo and London. Co-CEO Tekedra Mawakana has publicly targeted 1 million weekly rides by the end of 2026.
The honest trade-offs: Waymo rides cost roughly 70% more per mile than Tesla's robotaxi pricing (~$1.36–1.43 vs ~$0.81), the LiDAR-equipped sensor suite remains expensive on a per-vehicle basis, and the service area is geofenced — Waymo doesn't currently run on freeways or service some major airports in its original markets. Regulatory scrutiny is also real: NHTSA and NTSB have opened investigations into specific incidents, including reports of Waymo vehicles failing to stop for school buses in Austin and Atlanta. Waymo's response has leaned heavily on its published safety statistics, citing a 90% reduction in serious-injury crashes compared to human drivers across its autonomous mileage — a company-reported figure that external safety researchers continue to evaluate independently.
Mobileye's Pivot: From Supplier to Operator
Until June 16, 2026, Mobileye's position in this comparison was straightforward: it doesn't compete with Tesla or Waymo directly because it doesn't run a consumer-facing service at all. Its business has been building and licensing the underlying chips and software — EyeQ system-on-chips, SuperVision (hands-free, eyes-on ADAS), Chauffeur (more advanced automated driving), and the Road Experience Management mapping system — to automakers like Volkswagen, Porsche, Polestar, Ford, and Mahindra, who then integrate that technology into their own vehicles.
That changed with the company's announcement of a vertically integrated robotaxi business, set to launch with approximately 100 driverless vehicles in an unnamed US city in 2027. CEO Amnon Shashua framed the move explicitly as a response to industry concentration: "the industry has become increasingly dependent on a small number of technology providers and business models," and Mobileye believes there's room for a new approach built on two decades of autonomous-driving development plus operational ownership. The new robotaxi unit will use Mobileye's existing Moovit mobility platform for ride-booking and fleet coordination, and is explicitly designed to complement rather than replace its existing automaker supply business.
The strategic logic is about data: direct fleet operation gives Mobileye a proprietary feedback loop to refine its SuperVision and Chauffeur platforms using real deployment data, rather than relying solely on what its automaker customers report back. It's a high-stakes pivot — Mobileye reported $558 million in Q1 2026 revenue from its existing supplier business, and the move into capital-intensive fleet operations is a meaningfully different and riskier business than chip licensing. Market reaction has been split: some institutional investors have increased positions on the long-term AI data thesis, while others have reduced exposure over execution risk and rising capital expenditure.
Mobileye's technical positioning is also distinct from both Tesla and Waymo: its SuperVision platform combines cameras and radar without LiDAR (a middle path between Tesla's camera-only and Waymo's full sensor suite), and the company has built its safety case around the Responsibility-Sensitive Safety (RSS) framework, a formal mathematical model for defining what constitutes safe driving behavior — an approach distinct from either Tesla's data-driven end-to-end training or Waymo's empirical mileage-based safety reporting.
The Honest State of the Race in Mid-2026
By the numbers that matter for an actual consumer deciding whether to hail a driverless ride today, Waymo is unambiguously ahead: more cities, more weekly rides, no safety monitor required, and a multi-year head start on regulatory approvals across more than 20 markets. Tesla's robotaxi service remains smaller in scale and, per independent reporting, still relies on human safety monitors for most rides — even as its consumer FSD feature continues to improve and ships to a vastly larger installed base of vehicles than Waymo's entire fleet.
Tesla's structural argument — that vision-only AI at a fraction of the hardware cost will eventually scale faster and cheaper than sensor-redundant systems — remains unproven in commercial robotaxi terms as of mid-2026, even though FSD's consumer-feature trajectory has clearly improved. Whether "eventually" arrives before Waymo's geofenced-but-proven model captures the early commercial market is the actual open question, not whether either technology "works" in isolated tests.
Mobileye's entry adds a third structural bet to watch starting in 2027: that deep automaker relationships, an RSS-based formal safety framework, and a more moderate camera-plus-radar sensor cost can support a vertically integrated robotaxi business without either Tesla's vision-only cost aggression or Waymo's full sensor-redundancy expense. It's the newest and least-tested of the three approaches in actual commercial deployment, but it comes from a company with the longest continuous track record in automotive ADAS of any party in this comparison.
FAQ
Is Tesla FSD considered fully self-driving?
No. As of mid-2026, Tesla's consumer FSD release remains classified as SAE Level 2 driver-assistance technology, requiring constant human supervision regardless of marketing language. Tesla's separate robotaxi pilot operates with reduced supervision in a small number of cities, but independent reporting indicates most of those rides still include a human safety monitor.
How many cities does Waymo operate in?
As of mid-2026, Waymo provides paid driverless rides in approximately 10 US cities, including Phoenix, San Francisco, Los Angeles, Austin, Atlanta, Miami, Dallas, Houston, San Antonio, and Orlando. The company has announced plans to expand to over 20 additional cities in 2026, including its first international markets in Tokyo and London.
What did Mobileye announce in June 2026?
On June 16, 2026, Mobileye announced it will launch its own vertically integrated robotaxi service, starting with roughly 100 driverless vehicles in an unnamed US city in 2027. This marks a strategic shift from its traditional business as a pure technology and chip supplier to automakers into also becoming a direct robotaxi operator, while continuing its existing supplier relationships.
Why doesn't Tesla use LiDAR like Waymo?
Tesla's strategy bets that cameras and neural networks trained on sufficient real-world driving data can match sensor-redundant performance at a fraction of the cost — roughly $400 per vehicle for cameras versus $10,000–12,000 for LiDAR-equipped systems like Waymo's. This cost advantage scales across every vehicle Tesla sells, not just a dedicated robotaxi fleet, but the approach remains commercially unproven for fully driverless operation as of mid-2026.
Is Waymo profitable?
Waymo doesn't break out standalone profitability figures within Alphabet's earnings, but independent analysts estimate a revenue run rate above $400 million annually based on its ride volume and pricing as of mid-2026. The company raised $16 billion in early 2026 at a $126 billion valuation, suggesting investor confidence in the long-term commercial trajectory even if current operations aren't yet profitable on a standalone basis.
Which is cheaper, a Tesla robotaxi ride or a Waymo ride?
Tesla's robotaxi pricing runs roughly $0.81/mile compared to Waymo's $1.36–1.43/mile — Tesla is meaningfully cheaper per mile. However, Tesla's average wait time exceeds 15 minutes compared to Waymo's approximately 5.7 minutes, and Tesla's service covers far fewer cities with a smaller fleet, so direct price comparison doesn't capture the full difference in service availability and reliability.