Elon Musk was right? Personal Autonomous Vehicles are coming

Pony.ai's Q2 earnings, WeRide secures Grab investment, Tesla Recruits Test Drivers in NYC

👋 Welcome to another packed week in the autonomous driving industry

This week brought significant developments across the autonomous vehicle landscape. Pony AI delivered their Q2 earnings demonstrating clear commercial progress, GM announced its re-entry into autonomy following the Cruise shutdown, and startup Tensor unveiled plans for a luxury self-driving car targeting private ownership.

⏱️ ~3200 words, ~14 min read

---

🚗 The Personal Autonomous Vehicle Revolution

This week delivered two significant announcements in the personally-owned autonomous vehicle space. General Motors revealed its pivot toward consumer AVs after shuttering its Cruise robotaxi division, while the startup Tensor unveiled plans for a luxury self-driving car targeting private ownership.

Most autonomous vehicle companies have historically started with robotaxis for strategic reasons. A robotaxi needs only to operate reliably within one city to create a viable business, returning to controlled depots each night for maintenance and updates. The company maintains complete control over the operating environment, routes, and vehicle conditions.

For personally-owned vehicles, the bar is dramatically higher—owners expect their cars to drive everywhere they want to go, from familiar neighborhood streets to unfamiliar highways across the country, in all weather conditions and traffic scenarios.

GM's Strategic Reset: From Fleet Dreams to Consumer Reality

Let's start by examining GM's announcement, which represents one of the most dramatic strategic reversals in recent automotive history. GM had been deeply invested in the robotaxi business through Cruise, which the automaker acquired in 2016 for its autonomous vehicle ambitions. For years, Cruise operated robotaxi services in San Francisco and other cities, positioning GM as a leader in autonomous mobility services.

However, the robotaxi dream turned into a financial nightmare. Cruise was burning through massive amounts of cash—$1.7 billion in 2024 alone—while struggling to achieve sustainable operations. The breaking point came when a Cruise vehicle seriously injured a pedestrian in San Francisco, triggering regulatory intervention and public scrutiny that exposed fundamental problems with the company's safety protocols and operational oversight.

Faced with mounting losses and regulatory pressure, GM made the dramatic decision to pull the plug on Cruise entirely. The shutdown will save GM up to $1 billion annually, according to CEO Mary Barra, who announced the decision during an earnings call. GM took a $500 million one-time charge just to exit the robotaxi business, on top of broader restructuring costs. In total, GM had invested nearly $10 billion in Cruise since 2016.

Now GM is making a fresh start with a completely different approach. The company has hired Sterling Anderson, the former Tesla Autopilot chief who also worked at autonomous trucking company Aurora, to lead a new autonomous vehicle program focused exclusively on personally-owned vehicles rather than robotaxis. Anderson, brings both academic rigor and practical experience to GM's renewed autonomy efforts.

Rather than attempting to leap directly to full autonomy like Cruise did, Anderson's team plans to advance incrementally. The first step involves developing hands-free, eyes-free driving capability with a human still present in the vehicle. This builds on GM's existing Super Cruise technology while incorporating lessons learned from Cruise's technical development. The ultimate goal is progressing to vehicles that can operate with no one at the wheel, but GM is taking a much more cautious, step-by-step approach this time.

However, the technical integration might pose challenges. GM faces the complex task of merging Cruise's L4 technology stack with its existing Super Cruise L2+ system—two different approaches that may not be compatible. Or do we see a new licensing deal with someone like Wayve or Nuro to really start from scratch?

GM also plans to selectively integrate former Cruise employees who worked on the most valuable aspects of the robotaxi program. But the sudden shutdown left many Cruise employees blindsided and has created skepticism about GM's commitment to autonomous technology. They lost trust with both the markets and critically, with engineers themselves. So I think GM will have a hard time convincing those people to return.

The open question: Did GM Exit Too Early?

GM's decision to abandon robotaxis came just as the market appeared to be gaining momentum, raising questions about whether the company pulled the plug prematurely. Waymo has expanded operations to multiple cities and completed millions of paid rides, while Chinese companies like WeRide, Pony and Baidu are scaling globally and have deployed hundreds of robotaxis in cities like Wuhan and Shenzhen. The robotaxi market that seemed perpetually "five years away" is finally here.

However, GM's experience suggests the capital requirements and operational complexity of robotaxis may exceed most companies' capabilities. Even Waymo, the industry leader, operates in only a handful of cities after more than a decade of development.

Despite the challenges, there's also hope for GM's new direction. By targeting the broader consumer market rather than complex urban robotaxi operations, GM can leverage its existing strengths in manufacturing, dealer networks, and consumer financing. Anderson's methodical approach and proven track record at Tesla and Aurora provide confidence that GM can execute this strategy successfully.

Most importantly, GM's pivot comes at an opportune time. While the robotaxi market is hot right now, the personally-owned AV market is just beginning to emerge. By focusing on this segment now, GM positions itself to be a leader rather than a follower in what could become a significant market opportunity.

Tensor's Market Entry: Privacy-First Luxury Autonomy

Transitioning to this week's second major announcement, Tensor's unveiling of its luxury autonomous vehicle represents a different approach to personal ownership. The Silicon Valley startup, which plans to launch its vehicle in 2026, promises the first consumer-available L4 autonomous car designed specifically for private ownership rather than fleet operation.

Source: Tensor

Tensor's vehicle is engineered as what the company calls a true "robocar" with "eyes-off" self-driving capability. The design includes several futuristic features that emphasize the transition from human to machine control. When operating in autonomous mode, the steering wheel folds away into the dashboard and is replaced by a screen, while the pedals retract into the footwell. This transformation creates a spacious, uncluttered interior that signals to passengers they are no longer in control of the vehicle.

The car features an extensive sensor array including 37 cameras, 5 custom-designed lidars, 11 custom radars, 22 microphones, 10 ultrasonic sensors, 3 IMUs, 16 collision detectors, 8 water-level detectors, 4 tire pressure sensors, 1 smoke detector, and triple-channel 5G connectivity. This maximum-hardware approach contrasts sharply with Tesla's camera-centric philosophy, suggesting Tensor believes early adopters will pay premium prices for sensor redundancy and safety assurance.

The privacy focus addresses a key consumer concern that robotaxi companies have largely ignored. While most autonomous vehicles maintain constant cloud connectivity and upload driving data, Tensor promises that owner data remains with the vehicle in self-driving mode. This approach could appeal to consumers uncomfortable with the surveillance aspects of connected vehicles, potentially creating a competitive advantage in privacy-conscious markets.

Tensor's emergence reflects the growing geopolitical complexity of autonomous vehicle development. The company appears to be a strategic rebranding of AutoX, a Chinese autonomous vehicle developer that has operated robotaxi services in China while maintaining a presence in California. This connection raises questions about technology transfer and regulatory compliance as the U.S. implements restrictions on Chinese software in vehicles.

AutoX has reportedly shut down all operations in China and divested from its domestic business, positioning Tensor as an independent American entity with majority non-Chinese investment. This restructuring appears designed to navigate regulatory restrictions while preserving access to U.S. markets, but it also demonstrates the challenges facing internationally-focused AV companies in an increasingly fragmented regulatory environment.

Tensor declined to reveal pricing for its vehicle, though it will likely cost more than existing luxury EVs like the Lucid Air that lack self-driving capability. This pricing reality highlights the fundamental challenge facing personally-owned autonomous vehicles: the technology remains too expensive for mass market adoption.

The extensive sensor suite that enables Tensor's L4 capabilities likely costs tens of thousands of dollars, placing the complete vehicle well into luxury territory. For personally-owned AVs to achieve broad adoption, hardware costs must decline dramatically while software capabilities continue improving.

The manufacturing arrangement adds another layer of complexity. Tensor has partnered with Vietnamese automaker VinFast for production, meaning vehicles will be manufactured in Vietnam. For U.S consumers, this means the vehicles are potentially subject to 20% tariffs that could significantly impact pricing in an already expensive luxury segment.

The Broader Personal AV Movement

Tensor and GM aren't alone in pursuing personally-owned autonomous vehicles. Tesla has promised consumer-available autonomy for years. While competitors pursued robotaxi-first strategies, Musk consistently argued that personally-owned autonomous vehicles would dominate the market. "The fleet wakes up with an over-the-air update," he famously said, describing how Tesla owners would suddenly possess valuable autonomous assets.

The consensus favored the robotaxi model: controlled environments, predictable routes, centralized maintenance, and professional oversight. Tesla's vision seemed impossibly ambitious by comparison.

Yet Musk's core insight—that consumers want to own their autonomous vehicles rather than summon them—may prove prescient. To be fair, Tesla hasn't delivered on its autonomous promises yet, with Full Self-Driving still requiring human supervision despite years of "next year" predictions. But the industry's pivot toward his original strategy suggests the vision was sound, even if the execution timeline proved optimistic.

Waymo has also begun exploring personal ownership through partnerships with Toyota, though these efforts remain in early stages compared to the company's robotaxi focus. The AV company's approach involves licensing autonomous driving systems to traditional automakers rather than developing complete vehicles independently.

Perhaps most intriguingly, ride-sharing company Lyft has outlined a vision for seamlessly integrating personally-owned autonomous vehicles into its platform. The company's "Lyft-ready" concept envisions owners earning money from their vehicles when not in personal use, similar to how Airbnb created a generation of host-entrepreneurs.

This model could address criticism that personally-owned AVs will exacerbate urban problems like traffic congestion and parking shortages without reducing overall vehicle ownership. If personal AVs can generate income through ride-sharing when idle, owners might need fewer vehicles while still accessing autonomous capability. The shared utilization could improve urban efficiency while maintaining the convenience and control that personal ownership provides.

Hurdles Ahead

Despite this week's announcements, significant obstacles remain before personally-owned AVs achieve mainstream adoption. Vehicle costs must decline substantially to reach mass market price points, requiring breakthrough innovations in sensor technology and manufacturing processes.

Insurance represents an equally complex challenge. Current automotive insurance assumes human drivers bear responsibility for vehicle operation, but autonomous vehicles without steering wheels or pedals create unprecedented liability questions. Tesla and other manufacturers may need to provide their own insurance products, as they're best positioned to understand and price the risks of their autonomous technology.

Consumer acceptance may prove the highest hurdle of all. While summoning a robotaxi involves limited financial risk, purchasing an autonomous vehicle represents a major financial commitment based on trust in technology most consumers don't understand. High-profile incidents like the Cruise pedestrian injury reinforce public skepticism and could trigger consumer backlash that devastates sales across the entire category.

The announcements from GM and Tensor mark important milestones in the evolution toward personally-owned autonomous vehicles, but they also highlight how much work remains before this vision becomes reality. Success will require not just technological advancement but fundamental changes in business models, regulatory frameworks, and consumer attitudes toward transportation itself.

🔗 Bloomberg / Forbes / The Verge / Lyft / TechCrunch / Road to Autonomy / WEF / LinkedIN /

---

💰 Pony.ai Q2 2025 Earnings

Pony.ai’s Q2 results capture the tension between rapid top-line growth and the heavy investment burden of scaling.

Revenue Composition

The top-line revenue grew 75% YoY to $21.5 million. Robotaxi revenue surged 158% year-over-year to $1.5 million, driven by explosive growth in both fare-charging services (up over 300% YoY) and project-based engineering solution services.

CEO James Peng attributed this growth to three specific drivers: "expanding user adoption and demand in tier-one cities," "increased fleet of deployed Robotaxi vehicles," and "ongoing optimization of our pricing and operational strategies across diverse user segments."

Fare-charging services are benefiting from the rapid deployment of Gen-7 vehicles, the expansion to 24/7 operating hours in Guangzhou and Shenzhen, and a larger service footprint across 2,000 square kilometers in China's tier-one cities—more than 20 times the size of San Francisco, as CFO Leo Wang noted. The company's registered user base surged 136% year-over-year while maintaining satisfaction rates "well above 4.8 out of 5," suggesting genuine product-market fit.

The 158% surge in robotaxi revenue is encouraging—it shows that fleet expansion, extended hours, and user adoption are translating into meaningful gains. But even with fare-charging services up over 300% YoY, total robotaxi revenue is still just $1.5 million for the quarter. At that level, it's far from being able to support the business on its own, particularly when viewed against the company's strategic retreat from its historically dominant robotruck business.

Historically, robotruck services dominated the company's revenue mix, generating $40.4 million in 2024 (54% of total revenue) compared to just $7.3 million from robotaxi services. This Q2 marked an inflection point where management deliberately retreated from trucking to focus on higher-margin revenue.

CFO Leo Wang's explanation that the 10% robotruck revenue decline to $9.5 million reflects "proactive operation optimization to focus on high-margin revenues" reveals a strategic about-face.

The explosive 902% growth in licensing and applications revenue to $10.4 million stems from "increased orders and deliveries for autonomous domain controller sales, driven by both new and existing robot-delivery clients."

Unit Economics and the Path to Profitability

The gross margin swing from -0.3% to 16.1% YoY represents the most critical development in these results, demonstrating that Pony AI's focus on operational efficiency is paying dividends. This improvement was helped by specific cost structure improvements: an 18% reduction in vehicle insurance premiums (reflecting their "impeccable safety track record" with over 2 million kilometers driven) and sequential gains in remote-operator efficiency toward the year-end target of 1:30.

CTO Tiancheng Lou emphasized that their insurance costs are already "at just half of the typical costs for traditional human-operated taxis," with the additional 18% reduction reflecting "growing recognition by insurers of our safety track record." The remote assistance efficiency gains are equally important—achieving a 1:30 ratio means one operator can monitor 30 vehicles simultaneously.

Source Goldman Sachs

Goldman Sachs estimates that the Chinese robotaxi industry will achieve a 1:36 remote assistant ratio by 2030E, with costs declining from $2,600 per vehicle annually to $1,400 by that timeframe. However, Pony AI's management projects reaching a 1:30 ratio by the end of this year. Fairly exceeding this forecast.

The rapid production progress—over 200 Gen-7 vehicles in just two months—shows that Pony AI is executing on its manufacturing timeline. The 70% reduction in bill-of-materials costs for Gen-7 vehicles, combined with the operational efficiency gains, creates what CEO Peng described as "a clear path towards positive unit economics."

Still, as former COO Haojun Wang told Bloomberg, breaking even will likely require "tens of thousands" of vehicles, making this year's 1,000-unit milestone just an intermediate checkpoint.

If the Gen-7 fleet reaches full deployment and these operational levers fully materialize, margins should improve further in the coming quarters. However, the magnitude of improvement will depend critically on how quickly fare-charging volumes scale and whether the company can maintain pricing power as competition intensifies.

Investment Burden Reflects Scaling Ambitions

The heavy investment burden is evident in the substantial increases in both R&D and SG&A expenses. R&D expenses surged 69% year-over-year to $49 million, driven by two primary factors: accelerated investments in Gen-7 vehicle development and mass production partnerships with GAC and BAIC, and increased employee compensation to strengthen technological capabilities for scaling operations.

SG&A expenses jumped 97% to $15.7 million, reflecting increased personnel expenses in preparation for large-scale commercial deployment and higher professional service fees related to their public company status and international expansion.

Cash Position Provides Strategic Flexibility

Despite the heavy investment burden, Pony AI maintains a robust financial position with $747.7 million in combined cash, short-term investments, and long-term debt instruments as of June 30, 2025. This represents a decline from $825.1 million at year-end 2024, primarily due to operational cash burn and increased capital expenditures for Gen-7 production scaling.

The quarterly free cash flow of -$35 million (compared to -$19.8 million in Q2 2024) reflects the capital intensity of the scaling phase, but at current burn rates, the company maintains many quarters of financial runway.

CFO Wang noted that "we believe our current cash reserves are well-positioned to support our operational needs” while remaining "proactive in exploring additional opportunities to ensure long-term financial resilience."

Looking ahead

We still lack detailed disclosure on the gross margin profile of each revenue stream—transparency that would significantly help investors judge the sustainability of this improvement. Given that the 10% decline in robotruck revenue was explicitly justified by "optimizing for high-margin revenues," detailed segment-level margin disclosure becomes crucial for understanding the logic behind such strategic shifts.

While the 16% blended gross margin represents a significant step up from negative territory, it remains far from levels typically required for long-term viability in capital-intensive businesses.

As discussed above, if the Gen-7 fleet reaches full deployment and operational levers like remote operator ratios fully materialize, margins should improve further in the coming quarters.

However, the magnitude of improvement will depend critically on how quickly fare-charging volumes scale and whether the company can maintain pricing power as competition intensifies.

🔗 Bloomberg (1) / Bloomberg (2) / Pony.ai (1) / Pony.ai (2) / LinkedIN

---

💡 Quick Takes:

🖋️ Waabi Recruits Uber Freight CEO to Scale Commercial Operations

Waabi hired Lior Ron, CEO of Uber Freight, as chief operating officer ahead of its planned driverless highway launch later this year. Ron will lead go-to-market strategy while Rebecca Tinucci, former Tesla charging network head, takes over Uber Freight. The move signals Waabi's transition from development to commercialization, with plans to launch in Texas

🔗 TechCrunch

💰 WeRide Secures Grab Investment for Southeast Asia Push

WeRide received tens of millions in equity commitments from Southeast Asian super app Grab to accelerate robotaxi deployment across the region. The investment, expected to close by H1 2026, builds on their March MoU and includes plans to integrate WeRide's robotaxis into Grab's network.

🔗 CNEVPost

🚛 DOT moves to clear the road for self-driving trucks

The Trump administration published a Request for Information seeking industry feedback on nationwide autonomous vehicle deployment requirements over the next 60 days. The initiative covers six key areas including data standards, edge-case identification, and human-machine interfaces, building on the first Trump administration's AV "Comprehensive Plan." Capitol Hill pressure includes legislation to codify FMCSA interpretations that don't require human drivers, potentially accelerating commercial AV timelines significantly.

🔗 Freightwaves

🚗 Tesla Recruits Test Drivers in NYC but hasn’t applied for Permits

Tesla is hiring "vehicle operators" in Queens to conduct "dynamic audio and camera data collection" for automated driving systems, despite not applying for New York City or state AV testing permits.

🔗 CNBC

🇭🇰 Hong Kong Expands Autonomous Vehicle Testing Programs

Hong Kong approved new trials at Cyberport involving 10 driverless cars from Baidu Apollo International, expanding beyond existing Northern Lantau and Tung Chung operations. The West Kowloon Cultural District program will advance to passenger testing this year, while Baidu plans additional trials in Kai Tak urban area due to "relatively lower traffic flow compared with other urban areas."

🔗 SCMP

🧹 Chinese Sanitation Fleet Goes Level 4

Jijing Intelligent Technology began nationwide sales of Level 4 autonomous sanitation vehicles in China, with production orders scheduled through Q4 2025. The Wuhan-based company's vehicles operate at 20 km/h with automatic obstacle avoidance.

🔗 Gasgoo

---

📚 Worth Reading/Listening:

Phil Koopman: Independence & Safety

🔗 Phil Koopman

---

📊 Weekly Performance

Note: Stock performance data as of August 17, 2025. Past performance does not indicate future returns.

---

Thanks for reading!

If you found value in this newsletter, please consider sharing it with a friend or colleague who might benefit from these insights.

And if you haven't already, subscribe to stay updated on the latest developments in the autonomous driving industry