Techin' Geek

During July and August, Washington implemented stricter regulations on foreign-developed advanced robotic technologies and imposed high tariffs on imported drones and their parts, with both actions citing concerns over national security. The drone tariffs will be enforced starting in September, with additional tariffs on components set to begin in 2027.

These actions are part of a wider U.S. initiative to limit foreign technology presence in critical sectors. The FCC’s Covered List, launched in 2021, initially focused on telecommunications and surveillance devices from companies such as Huawei, ZTE, and Hikvision, before being extended to foreign-made drones and, most recently, advanced robotic systems.

The latest developments come as Chinese manufacturers dominate the markets for both drones and humanoid robots, frequently competing at price points that U.S. and European competitors find challenging to meet.

Cumulatively, the restrictions raise a significant question for the global robotics sector: If Chinese drones and humanoids face increasing exclusion from the U.S. market, where will the competition pivot next?

While the restrictions may offer some protection to segments of the American market, they do not directly tackle China’s extensive manufacturing capacity and cost advantages.

Industry analysts and executives who engaged with TechCrunch indicated that the outcomes may lead to less of a clear U.S.-China divide and more of a fragmented global marketplace, where Chinese firms expand into other regions while U.S. and allied manufacturers contend in markets where security regulations are a primary concern.

The Scale Gap

The robotics sectors in the U.S. and China remain closely linked; however, both countries approach the competition armed with markedly distinct advantages. Unlike semiconductors, robotics relies on multiple technologies that no single nation can entirely dominate, noted Ankur Saxena, an investment director at TDK Ventures.

China leads the global market for humanoid robot production, with shipments reaching 22,000 units in the first half of this year, predominantly sourced from Chinese manufacturers, according to a report by Counterpoint. In contrast, U.S. firms are operating on a considerably smaller scale, as highlighted by Soumen Mandal, a principal analyst at Counterpoint Research.

The five largest humanoid robot manufacturers by shipment volume — AgiBot, Unitree, Galbot, UBTECH, and Leju Robotics — are all based in China and collectively represented 86% of global shipments in the first half of 2026, according to Counterpoint.

This advantage may deepen over time. Lower pricing enables Chinese manufacturers to deploy more robots, generating practical data that can enhance their technology. Higher production volumes can further reduce costs, Saxena explained.

Mandal pointed out that Chinese humanoid manufacturers are also lowering costs by internalizing more of the technology and utilizing China’s established manufacturing infrastructure. For instance, Unitree is enhancing more components in-house, while automotive companies like XPeng leverage their expertise in chips and vehicle production as they venture into robotics.

“The United States excels in advanced AI, software, and semiconductor innovations,” Saxena shared with TechCrunch. “China excels in manufacturing scale, supply-chain complexity, and cost efficiency.”

This manufacturing superiority has enabled Chinese firms to decrease humanoid pricing more swiftly than most U.S. competitors can respond.

“You cannot sanction your way out of a cost curve. The only way to overcome it is through building more, and America has not yet embarked on the long-term investments that will necessitate,” Saxena stated.

Where Does China Go Next?

The answer may increasingly lie outside the U.S. Even if Chinese robotics firms lose entry to the American market, they retain access to a sizable domestic market and opportunities for expansion elsewhere, particularly in regions where demand for cost-effective automation is on the rise, Saxena noted.

Chinese robotics firms are actively pursuing price-sensitive markets facing acute labor shortages across Europe, Southeast Asia, Latin America, and the Middle East, according to Mandal.

Mandal anticipates that humanoid manufacturers will follow a trajectory resembling that of Chinese electric vehicle makers: scaling up domestically, entering foreign markets, and ultimately establishing local production. Regions experiencing labor shortages and demographic decline could emerge as early adopters of humanoids, especially in manufacturing, where robots can perform routine tasks.

The drone sector provides an initial look at what this more fragmented robotics landscape might resemble. The industry is progressively bifurcating into two ecosystems: a U.S.-led market centered around American-made, NDAA-compliant technologies, and a China-led market prioritizing low-cost, high-volume production, shared Bentzion Levinson, founder and CEO of Virginia-based drone manufacturer Heven AeroTech.

Levinson remarked that Western manufacturers are unlikely to outpace Chinese companies in the low-cost consumer drone segment, where pricing remains a significant edge. Instead, U.S. and allied firms could find stronger competition in long-range autonomous systems for defense and critical infrastructure, where security specifications carry greater significance.

Levinson perceives the coming competitive landscape shifting from the drones themselves to the technology that powers them and the payloads they transport. “The next arena of conflict centers on who controls next-gen energy and payload frameworks,” he noted, especially highlighting battery limitations. As drones evolve in capability, battery constraints could become an increasingly crucial competitive factor.

Agility Robotics expressed support for the FCC’s July decision, claiming it could mitigate security risks associated with foreign-made advanced robots becoming entrenched within the U.S. market, a situation that has arisen in the drone field. The company highlighted its Digit humanoid, which is designed and produced in the U.S., while also advocating for ongoing access to necessary tools and technologies to advance robotics research.

A More Regional Robotics Market

“The alternative to China isn’t solely a domestically-focused U.S. supply chain; it’s a diversified allied supply chain,” Saxena remarked.

This could open doors in other parts of Asia. Japan boasts decades of expertise in industrial robotics and precision manufacturing, South Korea excels in electronics, batteries, and automobiles, and Taiwan plays a significant role in semiconductors. However, none can simply replace China, as Saxena pointed out, due to the deep integration of Chinese components in the global robotics landscape.

Asian manufacturers might emerge as a middle path between lower-cost Chinese robots and pricier U.S. alternatives, Mandal suggested. South Korea’s Hyundai, owner of Boston Dynamics, and Japan’s Toyota are among the automotive firms investing in robotics, drawing on their backgrounds in vehicles, manufacturing, and autonomous systems as they pivot toward humanoid robots.

Yang Fang of Beagle Technology, a California-based agtech startup applying AI and robotics software to transform standard farming tools into autonomous machinery, shared with TechCrunch that the robotics sector is likely to become more regionalized as firms engineer devices tailored to the labor needs, working conditions, and clientele within their home territories. Chinese robotics enterprises, for instance, might hone in on products suited for China and nearby markets, whereas U.S. firms are more inclined to develop for industries across North America, he noted.

The outcome may not yield neatly separated U.S. and China-led robotics sectors. Rather, the restrictions could hasten the rise of regional markets: Chinese companies competing on price and scale across a significant portion of the globe, U.S. and allied manufacturers gaining traction where security considerations dominate, and manufacturers in Japan, Taiwan, and South Korea striving to stake out ground amid the two.

Upon taking the helm at Grindr in 2022, George Arison stepped into a company that transitioned from Chinese ownership to a mandated divestiture followed by a private-equity intervention — a lucrative enterprise lacking a clear product or strategic direction. Fast forward four years, following a SPAC listing and a contentious return-to-office policy, and Grindr now presents a compelling growth narrative. Revenue is expected to nearly triple, soaring from $195 million in 2022 to a projected $540 million-plus this year, with adjusted EBITDA margins remaining above 40%.

This surge in revenue has primarily been fueled by getting current users to spend more rather than significantly expanding its user base. In Q2 of this year, the platform boasted 1.4 million paying users, making up 9% of its total users, yet the average revenue per user has significantly increased since 2022, with Arison keenly focused on the source of the next growth surge. Part of this vision aims to transform Grindr into a “gayborhood in your pocket” — a service that offers not only dating and hookups but also healthcare solutions (from ED medications to HIV prevention and ultimately, facilitating connections with gay doctors) and travel (assisting users in finding community regardless of their location). This reflects the widespread “everything app” trend currently shaping consumer technology.

However, that’s not all Arison is pursuing; Grindr is set to launch a subscription model later this year that he’s convinced will be well-received, introducing a more expensive “EDGE” tier that has already sparked backlash online (“literally who’s paying for this,” and “we need 2012 Grindr back”).

Arison, who founded and led Shift Technologies, an online marketplace for used cars that he took public via a SPAC in 2020, is actively courting media attention to argue that institutional investors are undervaluing Grindr’s stock merely because it operates as a gay dating platform. In a Zoom interview with TechCrunch on Thursday, he recounted an instance where an investor presented him with a financial model that included a distinct “Grindr discount” line item, subtracting 25% from a fair-value assessment.

Not everyone agrees that such a discount is warranted. Morgan Stanley, Goldman Sachs, and Raymond James have all elevated their price targets for the stock this year, with Morgan Stanley upgrading it to “overweight” in July, referencing the EDGE tier and Grindr’s foray into telehealth as contributing factors — part of a stock appreciation that has seen shares increase approximately a third in the past six months. Nonetheless, the discount is not a figment of Arison’s imagination — the stock currently trades at around 11 times 2027 EBITDA, representing a roughly 35% discount compared to competitors — though the reasoning for this remains unclear.

Arison — friendly, with traces of his Georgian roots evident in his accent — was open to discussing all these matters in a Q&A that has been edited for coherence and brevity.

You joined Grindr in 2022 when it was somewhat aimless. What was your initial priority?

Three main priorities arose. Firstly, building the company. Many employees had been hired during the height of COVID when in-office productivity expectations were minimal, resulting in an average of perhaps three to four genuine work hours daily. We reinstated a two-day in-office schedule in the summer of 2023 — a move that attracted both significant media coverage and considerable dissent. Consequently, our employee count dropped to about 70. Currently, only around 25 individuals who were here when I arrived remain employed at Grindr, and we maintain a streamlined workforce: 175 employees in the U.S. alongside a team in Colombia, achieving $540 million in projected revenue for this year.

The second focus was generating revenue growth by delivering products that users would willingly pay for — thus, pay conversion surged from below 6% to over 9%, and ARPU nearly doubled. Finally, establishing a long-term vision: the “gayborhood in your pocket” concept, along with all supplementary elements — healthcare, travel — which emanate from it.

You’ve claimed Grindr’s engineering culture is underrated. How small is the actual engineering team?

Roughly 94 or 95 individuals across various technical positions. A CEO from a major tech firm once told me that AI would enable me to achieve with far fewer individuals what I thought would require 300 to 350 — he was correct. We’re accomplishing what equates to about 350 people’s output with around 100. Approximately 80% of our code is now AI-generated, and we’ve recorded a 2.5x boost in engineering productivity over the past year.

Grindr trialed an AI-enhanced premium tier, EDGE, with pricing in Canada translating to about $350–375 monthly in U.S. dollars — a figure that faced ridicule online for being pricier than dating itself. What was the outcome there?

We have not yet launched EDGE — it’s currently in testing, and select users have access as part of that process. It is positioned above our existing XTRA ($23.99) and Unlimited ($44.99) tiers. We are not selling AI itself; we’re offering features made possible by it — leveraging user behavior and intent data, with consent, to create much better matches than an incomplete profile could. Retention rates for these features thus far have surpassed anything we’ve previously experienced.

The prices mentioned were one test scenario among several — we explored a range to gauge elasticity, not establishing final pricing. EDGE is anticipated to launch by late this year or early next, and we’ll have clarity on its position by then. We perceive it similarly to a Tesla Model X or S: a premium flagship product now, with its underlying capabilities gradually extending to the broader offering over time.

Regarding matching — you’ve indicated plans to use AI to suggest partners outside users’ localities, given the limited gay-dating pools even in cities like San Francisco. What evidence supports that long-distance matches are viable?

Even in San Francisco, where the proportion of the gay population is among the highest in the country, the total number of gay individuals may only range from 50,000 to 60,000. That’s a limited pool to seek potential partners, contributing to the overall challenges facing gay men in dating. So what if AI could entirely remove geographic barriers and reveal someone in St. Louis who truly aligns with your desires, based on actual behaviors rather than what’s presented in a profile?

As for whether it leads to real relationships: we don’t monitor users’ post-interactions — that’d be overstepping. However, we do know that Grindr is the platform most gay men report using to meet others for relationships, and there’s a notable shift among younger generations desiring different outcomes compared to older ones: about 50% of gay men under 35 express a wish for long-term monogamous relationships, while 25% desire children — figures that would have seemed inconceivable for my generation. When asked why they’re single, the frequent reply is difficulty in finding partners. While I can’t guarantee this solves that problem, exploring new methods is essential, as the existing approaches have evidently failed.

There are numerous avenues to broaden your potential market beyond just the dating application. The healthcare initiatives are particularly noteworthy — could you elaborate on those efforts, and whether you’re developing these products internally or serving as a gateway to external providers?

We initiated cash-pay offerings under the line we call Woodwork — ED medications, GLP-1s, peptides, etc. — as cash-pay was the most straightforward avenue to kickstart our initiatives, and we’ve recently launched an AI bot that facilitates transactions directly within the app instead of directing users to Woodwork.com. The second category involves HIV prevention and treatment: we’ve pledged to provide 10 million people with direct access to information on acquiring PrEP, both in the U.S. where we already supply this resource within our in-app health center, and internationally.

The third category — actual clinical care, such as connecting users with gay doctors through telehealth — is a longer-term vision. Currently, that’s not something we’re building, but I genuinely believe there’s a future, a decade ahead, where healthcare could surpass our existing revenue streams at Grindr.

As it stands, though, non-subscription revenue — including ads and healthcare initiatives — constitutes a minor portion of the business.

Currently, subscriptions represent about 83% of revenue, slightly down from around 86% in 2022, although subscription revenue itself has significantly grown — indicative of the overall expansion of the base. The emerging segments are indeed small right now. My aspiration is to create a company that, 10 years from now, boasts a robust subscription business, a solid advertising arm, a legitimate healthcare division, and an established travel vertical alongside it. Presently, however, those final two aspects are still in early stages.

You’ve pointed out that investors still apply what one termed a “Grindr discount” to the stock due to the company’s nature. Despite the stock’s recent sharp rise, Morgan Stanley’s upgrade, and it trading at a premium compared to Match Group, doesn’t the market imply that the discount has diminished?

I hope we’re now perceived as a growth entity — we’ve consistently increased revenue by over 25% for 16 consecutive quarters under my leadership, so that’s a reasonable expectation. The stigma discussion is very much real; we’ve encountered a consulting firm refusing to collaborate with us due to reputational concerns, along with a bank denying service during the Silicon Valley Bank crisis, even though major institutions like Goldman and Morgan Stanley have been strong allies. Much of this seems to revolve around Grindr being a gay dating platform rather than dating itself being inherently contentious — no one applies that label to Tinder, which features a “free tonight” button prominently on its homepage. However, I agree the market’s perception of us has evidently improved.

In European cities, vehicles are becoming more compact than before. Yet, as the demand for microcars in Europe has increased, the charming Italian ‘yoghurt pots’ have been mostly replaced by diminutive Chinese electric vehicles. Even Smart, the renowned ultra-compact car manufacturer, has shifted production to China.

Spanish startup Liux believes it can make its mark in a saturated market by introducing a petite electric vehicle centered around sustainability.

Taking inspiration from the Microlino out of Switzerland, Liux aims to establish a niche for itself with its forthcoming microcar, the Liux Big. The name “Big” is meant ironically. It is compact enough to park perpendicularly to the curb; however, the name also symbolizes the lofty aspirations of a team that often deviates from conventional paths.

“The concept of a European car is non-existent,” asserted Liux co-founder Antonio Espinosa de los Monteros during an interview with TechCrunch. This unexpected perspective from the CEO of a startup whose vehicles are already making news for being “made in Spain” reveals much about the company’s focus.

The batteries in Liux’s vehicles are not produced in Europe, but they can be recharged at home, even with energy generated from solar panels. The vehicle is also designed for easy maintenance to mitigate the rapid obsolescence often seen in contemporary cars. Notably, its fiber body is crafted from a unique linen-based biocomposite, allowing for the material to be reclaimed and recycled later.

“It is clear to both David and me that recycling goes beyond theoretical concepts. You can recycle nearly anything in a laboratory. The recyclability of a product is tied to its construction,” stated Espinosa. “When creating a product, one must strive to maintain the material and component integrity so that a second life can be achieved.”

“Genuine circularity” is where Espinosa’s experience lies; he previously co-founded Auara, a Spanish B Corp that sells natural mineral water in bottles that are both recycled and recyclable. Following a buyout of this successful brand by a larger entity, he and Sancho embarked on a new venture together.

In terms of vehicles, Sancho is at the wheel, focusing on minimizing emissions. His expertise involves engineering electric vehicles that can compete with gas-powered counterparts. Prior to Liux, he achieved notable success with the Bóreas, a hybrid supercar revealed at the 24 Hours of Le Mans in 2017. After parting ways with his previous partners, he collaborated with Espinosa to establish Liux.

Liux’s initial prototype, the Animal, merged their skills: The fully electric five-seater was constructed almost entirely from recycled or plant-based materials. However, after unveiling the SUV to the public in 2022, the co-founders decided to pivot. They recognized that their chances of obtaining homologation would be significantly improved with a smaller vehicle.

Fast forward to 2026, and Liux has obtained Europe-wide homologation for the Liux Big, which it anticipates launching in the first half of next year. Meanwhile, the company has expanded to 65 employees and is preparing to scale production across three plants in Spain.

These include the facility that TechCrunch visited in Azuqueca de Henares, situated roughly an hour’s drive from central Madrid.

Liux’s Azuqueca plant is compact as it employs Toyota’s “lean management” practices and handles only the final stages of production, according to its production head Beatriz Belda González, a Spanish engineer previously with BMW in Munich. However, its size should not be underestimated: Liux claims its production capacity could reach 20,000 vehicles annually by 2030.

It is still premature to evaluate market interest, but over 7,500 individuals have signed up for the waiting list for a Liux Big. Joining incurs no fee, but the list has provided the startup insights into its likely customers. The predominant demographic is urban residents aged 55 to 60, and Liux anticipates that the Liux Big will commonly serve as a secondary vehicle for households.

This might reduce expectations that microcars could rival conventional vehicle ownership, but as Espinosa stated, Liux must choose its challenges wisely. Instead of attempting to predict market trends and their pace, the startup remains open to collaborations with businesses managing B2B fleets and others that could facilitate the development of autonomous vehicles.

For Espinosa, the Liux Big has the potential to make a positive impact by delivering a more sustainable and affordable alternative. The startup has yet to finalize the price, but indicated it will be under €18,000 — approximately $21,000 — before any possible EV incentives. This positions it at the upper end of the microcar price spectrum, yet Liux aspires for it to exceed expectations — both in weight and performance.

Celso Fernández Llorens, head of R&D at Liux, noted that weight and size restrictions pose significant challenges in this sector. He believes most microcars share similar traits, even though European regulations distinguish ultralight L6e four-wheelers from slightly heavier L7e models. However, Liux has navigated these limitations to optimize its L7e homologation.

Through a series of thoughtful choices, the startup successfully incorporated a 260-liter trunk into the vehicle. However, a larger focus has been ensuring that users feel as though they are driving a car rather than a two-wheeled vehicle. This aspect is closely linked to safety, according to Sancho: one would not want a vehicle that is lightweight simply because its structure cannot endure an impact or that risks toppling during turns.

With this perspective, even though its category does not demand crash tests, Liux has been rigorously testing and demonstrating the Liux Big’s abilities to navigate slaloms, brake, and execute various maneuvers. The startup showcased some of these functionalities to TechCrunch during a short ride and test drive of its forthcoming off-road model.

Currently, its primary model will be available in two variants: 15 kWh and 20 kWh. A cargo version is also in the pipeline, and with Sancho aboard, the inclination to develop a supercar is always present. In a LinkedIn update, the company mentioned it does not aim to be “a one-car brand.”

First, however, Liux plans to utilize the €16 million it has raised so far (about $18.5 million, including European grants) to launch the urban variant of the Liux Big in collaboration with car dealerships throughout Europe.

The showroom where we convened also serves as a precursor to Liux’s future customer interaction experience, according to head of brand Ana Terrado Leyva. She pointed out features such as textile screens, 3D displays illustrating the Liux Big’s three color options — which is two more than the Ford T — and a linoleum floor as a tribute to linen. These design choices, she explained, signify another way Liux hopes to differentiate itself from its Chinese rivals.

Perhaps the notion of a European car truly does exist.

Elon Musk claims he has discovered a method to resolve one of AI’s major obstacles by personally creating a challenging-to-produce turbine component.

On Saturday, Musk validated the purpose of a clandestine facility SpaceX has been erecting in Bastrop, Texas, appearing in response to a report that was already narrowing in on the specifics. Earlier that day, The Information released a story referencing job postings that clearly mention a “blades and vanes foundry,” alongside information from Corey Trinetti, a diligence expert who provides comprehensive evaluations of AI infrastructure sites in his newsletter, reporting that SpaceX acquired approximately 830 acres adjacent to its existing Starlink factory in Bastrop between March and June.

“SpaceX and Tesla are each rapidly constructing 100GW/year of solar generation capacity,” Musk stated on X on Saturday, “but natural gas will still be essential to support and initiate solar for several years. The limiting element for natural gas turbine production is the casting of the blades & vanes. By conducting in-house casting at SpaceX, we can expedite the availability of natural gas turbines by as much as 18 months, which is a significant game-changer.”

The rationale behind all this relates to one of the most substantial hurdles currently confronting the AI sector. GPU shortages remain problematic — for instance, Nvidia’s latest Blackwell chips still face lead times stretching several months — but an additional limitation has arisen, specifically regarding the physical power grid. The International Energy Agency anticipates that electricity consumption in global data centers will approximately double by 2030, and gas turbine manufacturer GE Vernova indicates that it has effectively exhausted its production capacity through 2030, primarily due to the demands of AI infrastructure.

This scarcity is prompting hyperscalers, like Amazon, Google, Meta, OpenAI, and Microsoft, to develop private gas-fired plants adjacent to data centers, rather than relying on the grid. After years of focusing on wind and solar, they are all now investing in natural gas to bring data centers online more rapidly.

Regarding the casting bottleneck itself, as reported by The Information, the blades within a gas turbine’s hottest section operate at temperatures around 3,000 to 3,600 degrees Fahrenheit, which is approximately 800 degrees higher than the melting point of the metal alloy from which they are constructed. This is feasible only because of the blades’ internal cooling channels and thermal-barrier coatings, in addition to the specialized method employed for casting each blade. Merely four companies worldwide have acquired enough expertise in the casting process to manufacture them on an industrial scale, and all of them are currently at capacity.

What complicates matters further is that each blade must be cast as a single, uninterrupted crystal, cultivated slowly within a vacuum furnace, devoid of microscopic seams that allow regular cast metal to fracture under pressure. This process is already intricate for smaller blades utilized in jet engines; the blades in power-plant turbines are significantly larger, making defect-free production at that scale even more challenging.

Should SpaceX succeed in this endeavor — and it is undoubtedly easier said than done — it would imply that a Musk-controlled organization possesses a manufacturing capacity that all other AI infrastructure builders are currently reliant on a small oligopoly for, giving SpaceXAI a competitive advantage that is hard for any well-funded yet non-manufacturing rival to replicate quickly.

However, this would also mean an increase in gas turbines being deployed swiftly, and existing turbines are already facing federal lawsuits and peer-reviewed health studies regarding the pollution they release.

In Memphis, where SpaceXAI has been operating gas turbines to power its Colossus data centers since 2024, the NAACP has consistently accused the corporation of running turbines without the necessary permits or pollution controls mandated by federal regulations. The organization is concerned that such turbines release smog-forming compounds and hazardous substances like formaldehyde, pollutants associated with asthma, respiratory illnesses, and certain types of cancer. (The location is near neighborhoods that already experience significant industrial pollution, and researchers from the University of Memphis indicated that in their own limited analysis, air pollution had worsened “slightly” due to the data center.)

But Memphis is merely the most conspicuous instance. The same struggle is unfolding wherever gas turbines have become the standard solution for power shortages in data centers. In Virginia’s “Data Center Alley,” a study commissioned by the Piedmont Environmental Council, utilizing the EPA’s COBRA health-impact model, found that emissions from a single facility’s eight full-time gas turbines could impact over 2.5 million individuals across various counties — with the most severe effects falling on already marginalized communities — resulting in an estimated 3.4 to 6.5 additional premature deaths per year, equating to $53 million to $99 million in annual health-related damages.

The list of concerns and complaints continues.

Welcome back to TechCrunch Mobility, your central hub for the evolution of transportation, especially regarding the increasing impact of AI. To receive this in your inbox, subscribe here for free — simply click TechCrunch Mobility!

Supporters of autonomous vehicle technology have consistently claimed that robotaxis and other self-driving cars will lessen accident rates and enhance roadway safety. Some evidence supports this. However, it does not negate the fact that there have been consequences for those employed (or previously employed) by the firms creating this technology. 

Sean O’Kane, senior reporter for special projects, examined data reported to the Occupational Safety and Health Administration and discovered that test drivers for Waymo and Zoox experienced over two dozen injuries in 2024 and 2025 due to sudden braking or other abrupt actions taken by the autonomous vehicles. In certain instances, these employees faced months off work after sustaining injuries such as whiplash from the AVs halting unexpectedly and forcefully.

Read his complete story, which features interviews with both former and current personnel. 

There’s a crucial point to contemplate. You may be wondering, what about other AV developers? Surely, this issue isn’t exclusive to Waymo and Zoox? You’re likely correct. It is probable that test drivers for other AV developers are facing injuries too, but the companies they represent could be exempt from the reporting obligations of OSHA.

Have a suggestion for us regarding this story or others? Contact Kirsten Korosec at [email protected] or reach her via Signal at kkorosec.07, and Sean O’Kane at [email protected].

Deals!

I have been following Gatik, the autonomous vehicle company recognized for its self-driving box trucks, since 2019 when it emerged from its “stealth” phase. At that time, I was uncertain about the startup’s viability. The hype cycle had been rapidly discarding AV startups, and no one appeared near to commercializing their technology. 

Gatik not only persevered but has transitioned from research and development to commercial operations, though on a smaller scale compared to conventional human-operated delivery firms. (The company’s box trucks are unmanned and transport goods from distribution hubs to retail outlets such as Walmart.) With $200 million in new funding, it’s now aiming to expand. The funding round was led by Qatar Investment Authority and Koch Disruptive Technologies, with contributions from Millennium Management, ARK Invest, and Intact Private Capital. 

This represents Gatik’s most significant funding round thus far. However, I would argue that its multi-year commercial partnership with PepsiCo, signed in June and part of a total of $600 million in contracted revenue, is even more substantial. 

Other noteworthy deals …

Airbound, an Indian company focused on autonomous drones, secured $37 million in a Series A funding round led by Greenoaks, with participation from DoorDash, Lachy Groom, Lightspeed, and Humba Ventures.

Mubadala Capital, the alternative asset management division of Mubadala Investment Company, agreed to acquire a majority equity stake in Arrive Logistics, a truckload brokerage located in Austin.  

Regent Craft, a startup from Rhode Island creating and producing electric seagliders, raised $120 million in a Series B round co-led by Mare Liberum and AE Ventures. (The seaglider falls under a category of vehicle known as a wing-in-ground effect vehicle, or WIG.) Regent also obtained around $120 million in debt funding from Erebor Bank, established by Palmer Luckey, the founder of Anduril Industries. A point of interest about Regent, a company I have tracked for some time: The startup, which recently completed its 255,000-square-foot seaglider factory, is evidently advancing deeper into the defense sector, likely due to the available funding and partnership opportunities.  

Vista Global Holding, a private aviation entity based in the UAE, is contemplating a European IPO that might generate over $1 billion, as reported by Bloomberg.

Noteworthy reads and other updates

As per a recent YouGov survey, a larger proportion of Americans oppose police license plate cameras than support their use. What’s your take? Send me an email with your thoughts. 

Any, a startup producing electric two-wheelers from Belgium, is betting on cargo capacity. 

General Motors is under heightened examination from U.S. safety officials following numerous incidents, including over 20 crashes or fires, and at least six injuries linked to brake malfunctions in its electric vehicles.

Ford has appointed Dave Carroll as president of its energy division. Carroll, who previously worked with ENGIE North America, will take over from longtime executive Lisa Drake.

Rivian CFO Claire McDonough is stepping down and will depart by the end of October. Her tenure at Rivian coincided with a challenging yet exhilarating (ahem IPO) phase for the company. Her exit occurs at another pivotal juncture for Rivian as it embarks on some of its largest endeavors to date, including the robotaxi agreement with Uber and scaling production of its R2 SUV. 

Uber is rolling out a fresh live video streaming function allowing parents to monitor their kids during rides. 

Waymo shared ten insights gained from having their vehicles cover over 200 million autonomous miles. The foremost insight — that multimodal sensors are essential — garnered significant attention as it directly counters Tesla’s focus on camera-only systems. There were additional insights that caught my eye, including Waymo’s endorsement of vision language models (which are increasingly prominent). 

In the meantime, Waymo is advancing internationally, announcing plans to establish operations in Munich, Germany. 

One more thing …

A significant clarification regarding what I mentioned last week. You may recall that Waymo provided details about its custom silicon chip — specifically a 5 nm ASIC chip designed to manage the extensive influx of raw data prior to reaching the core “brain” of the autonomous system. Waymo stated in its blog entry:

“While these ASICs alone deliver over 1,000 TOPS of ML performance dedicated to front-end processing and ML models, we optimize across the full stack to maximize achieved performance, especially in the low-batch regimes we often operate.”

That statement led me, along with others, to conclude that a single chip could deliver 1,000 TOPS (trillions of operations per second) of computing performance. I compared it to Nvidia’s Drive AGX Thor automotive processor, noting the similar capabilities. 

However, one attentive reader reached out with inquiries, leading me to seek official clarification from Waymo. A spokesperson for Waymo indicated, “It’s for the system, not a single chip.”

This is a vital distinction.

Almost every organization aiming to implement artificial intelligence encounters a similar issue: integrating the technology into routine practices is challenging. Industrial giant Caterpillar has spent many years tackling a version of that dilemma in the physical realm, and now it’s leveraging that experience to integrate AI.

Caterpillar’s venture into autonomy initiated with mining, where labor shortages and dangerous conditions create a significant need for automation. Currently, it offers automated haul trucks, drilling equipment, underground loaders, dozers, remote-controlled construction machinery, and more. Additionally, it provides software command centers, fleet management, and remote terrain intelligence as part of its autonomous technology suite.

“We are now in a thrilling era where we can apply all that knowledge from mining and extend it to more dynamic environments like jobsites, quarries, and construction locations,” stated the company’s CTO, Jaime Mineart, to TechCrunch during the Ai4 conference in Las Vegas earlier this month.

The industrial behemoth is now extending AI applications more broadly, including in tools utilized by technicians and its internal team. One instance is the Cat AI Assistant, allowing field technicians beside a machine to use voice commands for accessing repair procedures, troubleshooting issues, and pinpointing necessary parts prior to starting a repair. Mineart noted that the tool is currently used by customers, operators, and technicians alike.

The assistant relies on Caterpillar’s exclusive data, drawn from insights produced by its connected machinery. Mineart mentioned that Caterpillar has approximately 1.6 million connected assets worldwide and over 16 petabytes of structured information.

The organization is also utilizing AI to enhance software for site scanning and creating digital twins in manufacturing to scrutinize operations, she added. Similar to nearly all companies, Caterpillar is employing AI across its business processes, as well as in software development. “We implement AI agents to modernize outdated code, generate and test new software, and detect faults early on,” Mineart stated.

However, Mineart emphasizes that constructing the technology is merely part of the hurdle, as deploying an autonomous machine differs from altering a site to incorporate AI. Organizations must also reconsider how personnel collaborates with the technology and how existing procedures need adjustment.

“The challenging aspect of autonomy and physical AI is integrating that technology into the customer jobsite and into the workflows,” she explained.

Mineart mentioned that the company relies on seasoned operators to assist in training AI systems, harnessing institutional knowledge accumulated over years. As machines grow more autonomous, some operators may transition from managing a single machine to overseeing multiple machines from a remote command center.

This transition, however, presents a new obstacle for Caterpillar: training its 118,000 workforce. Mineart indicated that the organization plans to invest $100 million over the next five years to educate its employees in AI, autonomy, and robotics.

This investment is likely aimed at helping the company capitalize on the wider surge in AI infrastructure, which is already benefiting its revenue. Caterpillar’s quarterly earnings hit a record $20.5 billion in the second quarter, buoyed by robust demand for power-generation equipment utilized in data centers. Its power-generation segment experienced a 72% increase in sales, reaching $3.10 billion, with CEO Joe Creed stating that “no one is slowing down” regarding the demand for cloud computing and generative AI infrastructure.