Image Source: Pexels
 

Waymo bundles up for cold weather.
 

Robotaxis Head to Las Vegas, San Diego, and Detroit

Please note Waymo is bringing its robotaxis to Las Vegas, San Diego, and Detroit.

Today, Waymo LLC announced it’s bringing its autonomous ride-hailing service to three more cities: Las Vegas, San Diego, and Detroit. The company started driving in each city today with a mixed fleet of its Jaguar I-PACE and Zeekr RT vehicles equipped with its sixth-generation Waymo Driver.

The news comes just a few weeks after Waymo shared its plans to bring its robotaxi services to London in 2026. The company’s other international plans include Tokyo, where it had deployed test vehicles to adapt it autonomous vehicle (AV) technology to local traffic patterns and road features.

In the U.S., the Waymo Driver has already driven over 100 million fully autonomous miles on public roads and provided more than 10 million paid rides. The company said it is now driving more than 2 million fully autonomous miles and providing over a quarter of a million rides per week. The company currently operates services in Phoenix, San Francisco, Los Angeles, and Austin.

In March, Waymo announced plans to bring its robotaxis to Washington, D.C., in 2026. The district gets an average of 13.7 in. of snow a year, according to the National Weather Service.

Waymo has also shown some interest in the city of Boston. In May, it sent a small fleet of robotaxis, driven by human drivers, to map Boston’s unique driving environment.

Additionally, in August, New York City granted Waymo a permit to test a limited number of AVs, equipped with safety drivers, in parts of Manhattan and downtown Brooklyn.

The Waymo Driver uses cameras, radar, and lidar to perceive the world around it, with each sensor providing a complementary field of view that’s especially helpful in inclement weather. Its automated cleaning system — using engineered and heating elements — keeps the sensors clear so the vehicle can continue serving riders without needing to pull over.

The company said its AI can distinguish between snow, slush, ice, and a normal road surface. The Waymo Driver then uses this information to adjust its driving behavior to match the road conditions in real time, allowing the Waymo Driver to navigate based on what it sees and feels, the company said.

Where’s Tesla?

Nowhere near cold, in beta, with backup drivers.

Tesla (TSLA) is currently testing its robotaxi service, which uses autonomous vehicles for on-demand rides, in limited areas like Austin, Texas, with plans to expand to eight to ten metro areas by the end of 2025. The service is still in a beta phase, often with human safety drivers, and relies on the company’s Full Self-Driving (FSD) software. In the future, Tesla plans to introduce purpose-built, fully autonomous vehicles called the Cybercab, which won’t have a steering wheel or pedals.

Grok’s Take Q&A

Mish: Will FSD work without LIDAR in snow, sleet, dust storms?
Grok ….

Performance in Snow and Sleet

• Light to moderate snow: Recent versions (e.g., FSD v13.x in 2024–2025) have shown notable improvements. User reports and videos demonstrate reliable handling on snow-covered or unplowed roads, including gravel paths with 5+ inches of accumulation, following ruts/grooves, avoiding deeper snow, and maintaining traction without interventions in many cases. Traction control integrates well, and the system can infer road position even without visible lane lines.
• Heavy snow, blizzards, or whiteouts: Performance degrades sharply. Cameras often get covered in snow/slush (especially rear and side ones), leading to warnings, disengagements, or unsafe behavior like veering, accelerating too quickly, or failing to detect edges. reddit.com +2 Older videos (e.g., FSD Beta in 2023) show dangerous sliding or inability to navigate. carscoops.com In extreme cases, no sensor system (including human vision) works reliably, and FSD may hand control back or refuse to engage. brownstoneresearch.com
• Sleet (freezing rain/ice) compounds issues with camera icing and slippery surfaces, often requiring manual intervention. @JPPelletier3

Performance in Dust Storms

• Dust severely limits camera visibility, similar to heavy fog or snow. There’s limited real-world data, but reports indicate FSD would struggle in near-zero visibility (e.g., haboobs), as it can’t “see through” opaque air like radar might. @TheTrueTruth777 Elon Musk has noted LIDAR also fails here due to scatter, implying vision-only isn’t uniquely disadvantaged—but in practice, pure camera systems are blinded when light is blocked. gearmusk.com +1

Why No LIDAR Helps or Hurts

• Tesla argues humans drive with eyes only, and billions of miles of training data allow neural nets to infer depth/position reliably. brownstoneresearch.com LIDAR is called a “crutch” that’s expensive and scatters in precipitation/dust. @elonmusk
• Critics (and competitors like Waymo/Cruise) counter that LIDAR + radar provide redundancy: LIDAR for precise 3D mapping (filtered for noise in snow/rain), radar for penetration in fog/dust/snow. electrek.co +2 Waymo operates in snow/fog with these sensors, adapting speed and path in real-time.

Current Reality (as of November 2025)FSD (Supervised) is SAE Level 2—you must always supervise and be ready to intervene. @ArAIstotle

It works increasingly well in mild adverse weather but is not reliable in severe snow, sleet, or dust storms without human oversight. Improvements are rapid via over-the-air updates and fleet data, but extreme conditions remain a weak point for vision-only systems. No autonomous tech (LIDAR-equipped or not) is perfect in whiteouts—safe driving often means pulling over. thinkautonomous.ai

In short: It can work in lighter conditions (better than ever), but no, not fully or unsupervised in heavy snow/sleet/dust without risks. Tesla’s page doesn’t claim otherwise for these scenarios.

LIDAR Q&A

LIDAR is not a crutch. It is an aid. It cannot hurt.

Musk did not want to pay for LIDAR. That was a mistake because costs did not stay prohibitively high.

Mish: What is the cost of Waymo’s LIDAR per vehicle?

Grok …

Historical Context and Cost Reductions

• Early LIDAR units (e.g., Velodyne models used in prototypes) cost up to $75,000 per unit. businessinsider.com +2
• By 2017, Waymo’s in-house development (e.g., Honeycomb and Laser Bear series) reduced this by over 90%, to around $7,500 per high-end unit. arstechnica.com +2 Short-range units were likely under $5,000. tech

Current Estimates (5th and 6th Generation Hardware)Waymo vehicles use multiple custom LIDAR units (e.g., 5 in 5th-gen, 4 in 6th-gen) for redundancy and 360° coverage up to 500m. waymo.com

• Full sensor suite (cameras + LIDAR + radar):
  • 5th-gen (e.g., on Jaguar I-Pace): ~$9,300–$12,700 total per vehicle (BloombergNEF and Contrary Research estimates, 2024–2025). reddit.com +1
  • Older estimates for similar suites: $40–50k (pre-2023). tangramvision.com +1
• LIDAR-specific portion:
  • Likely $2,000–$5,000 total across all units in recent gens, inferred from suite breakdowns and Waymo’s emphasis on “significantly reduced cost” in 6th-gen (announced 2024, with fewer LIDARs but higher performance). waymo.com +1 Per-unit costs for mass-produced equivalents (e.g., Chinese suppliers like Hesai) are now under $1,000, trending to $200–500.

Why Costs Have Dropped

Waymo’s vertical integration (in-house design/manufacturing) and scaling production have driven reductions, similar to battery price curves. businessinsider.com +1 The 6th-gen suite (13 cameras, 4 LIDAR, 6 radar) is “optimized for greater performance at a significantly reduced cost” without safety compromises. waymo.com

In robotaxi operations, even $10k+ sensor suites add ~3–5 cents/mile over a vehicle’s life (assuming 300k+ miles), making them viable for fleets. reddit.com

Waymo’s focus on redundancy (LIDAR + radar + cameras) prioritizes safety in adverse conditions over pure vision systems. businessinsider.com

China reduced the cost to under $1,000. OK tariffs will make that price impossible to get.

Current estimates are $2,00 to $5,000. Grok points out “Even $10k+ sensor suites add ~3–5 cents/mile over a vehicle’s life (assuming 300k+ miles), making them viable for fleets.”

Cost Questions

What’s the cost of all these delays trying to get FSD to work? How far behind is Tesla?

And what’s the cost of a lawsuit?

Amusing Flashback

Please recall More Absurd Predictions from Elon Musk: 1 Million Self-Driving Fleet in a Year

Q: When was that?
A: April 23, 2019

Also, “Probably two years from now we’ll make a car with no steering wheels or pedals.”

Flash Forward

July 20, 2025: Motor Trend Tests Tesla’s FSD, Tesla Flunks 3 Ways, Funny Videos

Tesla’s FSD is not ready for prime time and won’t be without LIDAR.

Please click on the above post if you have not see it. The video is hilarious.

More By This Author: