Tesla is rapidly increasing the number of its driverless Cybercabs authorized for commercial service in Texas, but the pace of expansion remains modest compared with the scale of Elon Musk’s ambitions for the robotaxi business.

In the month since Tesla unveiled its Cybercab service in Austin, the number of its bronze two-seat vehicles authorized for commercial use in Texas has nearly quadrupled, rising from 45 at the Sept. 3 launch to 169 as of Friday, according to public records from the Texas Department of Motor Vehicles. The increase shows that Tesla is steadily building out the fleet, but it also highlights the distance the company still has to cover if autonomous ride-hailing is to become a meaningful new growth engine.

The Cybercab is an unusually stripped-down vehicle. It has no steering wheel, pedals, rear or side-view mirrors, or outside door handles, and relies on Tesla’s camera-based autonomous driving technology and sensors. For paying passengers, however, the novelty has come with a range of practical problems.

Passengers have posted videos and comments describing lengthy waits, vehicles arriving at incorrect pickup or drop-off locations, and difficulties involving the cars’ butterfly doors and trunks. Those problems matter because Tesla is attempting to turn autonomous driving from a technology demonstration into a commercially reliable transportation service.

The company needs that transition to happen at a time when its core automotive business is under increasing pressure. Tesla’s vehicle sales have remained sluggish, while its automotive revenue has declined for consecutive years as Chinese electric-vehicle manufacturers such as BYD and Xiaomi compete with more affordable and innovative models.

Tesla shares were down 18% this year before Friday’s delivery report, making it the only megacap technology stock to have failed to generate a positive return for investors. Better-than-expected third-quarter vehicle deliveries subsequently lifted the stock almost 5%, but deliveries still fell about 2% from a year earlier. That makes the robotaxi operation more important to Tesla’s broader investment narrative. Musk has repeatedly positioned autonomous vehicles as a major source of future growth, but investors now have to assess not only whether Tesla can make its vehicles autonomous, but whether it can operate a large, dependable commercial network.

Waymo’s Head Start Raises The Bar

Tesla is not entering an empty market. Alphabet’s Waymo has spent years developing commercial autonomous ride-hailing and has established a considerable operating lead.

Waymo is now operating in 15 U.S. markets, with another 15 on the horizon, according to the company’s website. It has also announced planned international launches in London, Tokyo and Munich. In Texas alone, 1,154 autonomous vehicles were authorized for commercial use as of Friday, including 359 of its newer Ojai models, which feature a low step and sliding doors.

Waymo says it is conducting more than 500,000 paid rides each week and has completed 270 million fully autonomous miles of commercial driving domestically. Its U.S. commercial operations currently include more than 4,000 driverless vehicles.

Tesla’s Texas robotaxi fleet is considerably smaller. In addition to the 169 Cybercabs, its fleet includes 420 Model Y vehicles equipped with Tesla automated-driving systems that are not yet available to individual car buyers.

Austin is currently the only city where Tesla operates the Cybercab as part of its driverless ride-hailing service. The city has become an important testing ground for autonomous vehicles because of relatively lax state regulations and its large technology community. Amazon’s Zoox is also testing there.

The difference between the companies goes beyond the number of autonomous vehicles on the road. It is the maturity of the underlying service. Tesla is still refining basic elements of the passenger experience while Waymo has already built a network capable of handling hundreds of thousands of paid journeys every week.

Ethan McKanna, a 20-year-old Austin resident, former Tesla intern and computer science student at Texas A&M University, has taken rides in every type of autonomous vehicle available to public riders in the city. He has also created RobotaxiTracker.com, which uses machine learning to analyze data from traffic cameras and other public records to track autonomous services in U.S. cities.

McKanna, according to CNBC, said he has taken about 30 Cybercab rides and has used Waymo since the latter launched in Austin in 2024. He likes the Cybercab’s private cabin, massive screen, media controls and apps, describing the driving experience as “very smooth.”

But the service around the vehicle remains less mature.

The “pick up, drop-off experience,” he said, was not yet refined, while the outward-opening butterfly doors were “less convenient” than conventional or sliding doors. Early demand also produced significant waiting times.

“There was so much demand that wait times were often like 45 minutes plus,” McKanna said. He added that waiting times have fallen in recent weeks as demand has become more “equalized.”

Waymo, meanwhile, benefits from having operated commercially for longer.

“Waymo’s been doing commercial operations for longer, so they’re just more mature with 24/7 service, better support, things like that,” McKanna said. He noted that Waymo itself experienced problems in its early Austin operations, including “harsh braking” issues.

For Tesla, the challenge is turning the Cybercab from an impressive autonomous vehicle into a transportation system that passengers can rely on repeatedly.

Expansion Brings Technical and Regulatory Risks

Scaling that system beyond Austin could expose Tesla to another layer of difficulty: regulators.

The National Highway Traffic Safety Administration initiated an audit query following the Cybercab launch to determine whether the vehicles comply with federal safety standards. NHTSA initially asked Tesla to respond by Sept. 30 to a long list of safety questions, although Tesla obtained an extension, according to an agency spokesperson.

The regulatory scrutiny comes as Tesla continues to confront the difficult edge cases that autonomous vehicles encounter outside controlled demonstrations.

One of the latest involves something far smaller than a conventional road hazard. Musk said Saturday that Tesla was extending its Austin robotaxi operating hours from 10 p.m. to 11 p.m., while also identifying nighttime visibility as a problem the company is trying to solve.

“The main thing we’re trying to solve is making sure that we don’t run over pets when they’re hard to see at night,” Musk wrote in an X post. “Literally trying to avoid grey kittens on grey tarmac in the dark.”

The issue underpins the difficulty of autonomous driving: a system capable of handling ordinary traffic still has to respond correctly to rare, unpredictable situations. Cats and dogs can move suddenly into the roadway, creating a different challenge from the predictable flow of vehicles and pedestrians.

Tesla says its Cybercabs use “camera vision and sensors” to navigate. Its vehicle manuals acknowledge that low or limited light, including unlit or poorly lit roads at night, can limit a camera’s ability to provide accurate visual information.

Musk has argued that AI-assisted image processing can improve nighttime performance. Responding to a comment on his X post, he said: “Visual spectrum cameras can see very well in the dark when AI is doing the photon count analysis.”

But the issue is already attracting regulatory attention. NHTSA is investigating Tesla’s camera system for full self-driving vehicles and whether it can adequately identify road hazards under less-than-ideal visibility conditions.

Waymo takes a different technical approach, using LiDAR, which relies on laser beams to scan the surrounding environment. That technology is not without its own challenges. Late last year, a Waymo vehicle struck and killed a 9-year-old tabby cat named KitKat, prompting an outpouring of tributes from residents in the neighborhood.

Tesla has also encountered an animal-related incident in its Austin robotaxi operation. In September 2025, a Tesla Model Y operating as part of the service struck a dog that ran into the street. Tesla reported the incident to NHTSA, saying the dog ran away from the vehicle and appeared to be uninjured.

These incidents are a reminder that the robotaxi race is not merely a contest over fleet size or city launches. Each additional vehicle and each additional hour of operation increases the number of real-world situations autonomous systems must handle safely.

That is challenging for Tesla because Musk’s ambitions require scale. The Cybercab fleet has grown substantially in a short period, but the company is still operating in a single city with a relatively small number of purpose-built vehicles, while Waymo is already operating thousands of autonomous vehicles commercially across the United States.

Therefore, Tesla faces a difficult sequencing problem. Analysts, including some Tesla bulls, believe the EV giant needs to expand quickly enough to demonstrate that robotaxis can become a significant business, while improving reliability and satisfying regulators before expansion exposes shortcomings at a much larger scale.

The U.S. government plans to lend about $4.2 billion to Vistra Corp. to increase electricity production from its existing nuclear power fleet, in a move that underscores Washington’s growing focus on nuclear energy as demand from artificial intelligence data centers and other power-intensive industries accelerates.

Energy Secretary Chris Wright is expected to announce the financing on Monday at one of Vistra’s nuclear facilities along Lake Erie in Ohio, according to a person familiar with the matter cited by Bloomberg.

The financing would support power uprates at at least three of Vistra’s four nuclear power stations. An uprate increases the amount of electricity a reactor can generate without constructing an entirely new plant, potentially allowing the company to add capacity more quickly than through a new-build project.

The proposed loan is significant because it targets existing nuclear assets rather than relying solely on the much slower process of developing new reactors. The approach could become an important part of the U.S. response to a rapidly emerging power constraint as electricity consumption begins to rise after decades of relatively modest growth.

Vistra operates six nuclear reactors across four U.S. plants with combined generating capacity of more than 6.5 gigawatts, enough electricity to supply roughly 3.25 million homes.

Nuclear plant uprates have been used in the United States for decades to extract more electricity from operating reactors.

The Nuclear Regulatory Commission says utilities can increase reactor output through several approaches, including refueling with slightly more enriched uranium, increasing the proportion of new fuel used during refueling, or carrying out major modifications such as replacing main turbines.

The NRC had approved more than 170 uprates as of 2022, according to the agency.

The significance for the current U.S. power market is speed. Building a new nuclear reactor requires substantial capital, lengthy construction periods, and regulatory approvals. Increasing output from reactors that are already operating can potentially add electricity to the grid without requiring an entirely new generation project.

The proposed Vistra financing is therefore consistent with the broader nuclear policy being pursued by the Trump administration. In May 2025, Trump issued executive orders directing the government to facilitate the expansion of U.S. nuclear capacity from roughly 100 GW in 2024 to 400 GW by 2050. The orders specifically instructed the Energy Department to prioritize efforts supporting power uprates at existing reactors, as well as restarting closed plants and developing new reactors.

The administration’s nuclear strategy has been linking reliable electricity generation to the country’s technological and industrial ambitions. The White House has specifically identified AI and other energy-intensive computing applications as areas requiring reliable, high-density power.

That connection is becoming more important as utilities confront an electricity-demand outlook that is markedly different from the one that prevailed for much of the past two decades.

U.S. power demand is growing as companies build increasingly large data centers to support AI models and cloud computing. Electrification of transportation and cryptocurrency mining are also contributing to demand.

The AI buildout is becoming crucial because data centers require large quantities of electricity around the clock. Unlike some industrial loads, advanced computing facilities can require highly reliable power, increasing the value of generation sources capable of operating continuously.

Nuclear plants are particularly suited to that role because they can provide steady baseload electricity without the fuel-price volatility associated with fossil-fuel generation.

The investment race in AI is therefore increasingly becoming an investment race in electricity infrastructure. Semiconductor manufacturing, data centers and other digital infrastructure can be built relatively quickly compared with power-generation projects, creating a potential mismatch in which computing capacity comes online faster than the grid can provide the electricity required to operate it.

That is helping change the economics of the U.S. nuclear industry. Existing reactors, which already have grid connections, operating workforces, and established infrastructure, can offer a potentially faster way to increase supply than waiting for new reactors to be completed.

The DOE’s May 2025 nuclear industrial-base order explicitly directed the department to facilitate 5 GW of power uprates at existing reactors and to prioritize financing for increasing output from operating plants. The Vistra loan would therefore fit into a broader policy shift rather than representing an isolated financing decision.

The federal government has also made hundreds of billions of dollars in potential financing available through its loan programs, while Wright has said nuclear power would account for a substantial portion of that capacity.

The policy challenge is that expanding nuclear generation is capital-intensive even when existing reactors are being upgraded. The government is attempting to reduce financing barriers while simultaneously changing the regulatory environment to make new nuclear projects faster to approve and construct.

Trump’s target of increasing nuclear capacity from about 100 GW to 400 GW by 2050 would require a substantial expansion from the current fleet. That makes the Vistra project a useful test of a more incremental approach: rather than waiting for hundreds of new reactors, the United States can first seek additional output from plants already operating.

The financing could turn Vistra’s existing nuclear infrastructure into a source of additional growth at a time when electricity demand is strengthening. For Washington, it offers a way to address near-term power requirements while pursuing a much larger nuclear expansion over the next two decades.

A former OpenAI safety employee has publicly criticized the company’s approach to artificial intelligence development after quitting the company, arguing that its rapid launch cycle is creating risks that cannot be adequately addressed through safeguards added after systems are deployed.

David Robinson, who spent three and a half years at OpenAI and worked on the company’s preparedness framework, said the company and the wider AI industry are not being “nearly careful enough” as they develop increasingly capable models.

Writing in The Atlantic under the headline “I Quit OpenAI Because Its Culture Is Broken,” Robinson said that advanced AI development requires a fundamentally different approach to safety, with greater reliance on specialized expertise and research before increasingly powerful systems are built and deployed.

“The time for trial and error is over,” Robinson wrote, arguing that AI development should adopt safety standards more comparable to those used in high-risk industries such as nuclear power and aviation.

His criticism comes as leading AI companies face growing pressure over whether their ability to develop more capable systems is advancing faster than their ability to understand their behavior, identify failure modes, and build reliable safeguards.

Robinson said he helped draft OpenAI’s preparedness framework and oversaw safety reports for 12 frontier-model launches during his time at the company. His criticism therefore centers not only on AI safety as a general concern but on the internal processes used to evaluate increasingly capable models before release.

“As the company sprints from one launch to the next, it is failing to achieve the level of care that I believe is needed,” Robinson wrote.

The Limits of “Iterative Deployment”

A central target of Robinson’s criticism is OpenAI’s philosophy of “iterative deployment,” under which models are released, and safeguards are strengthened as problems emerge in real-world use.

The approach has an obvious advantage: deploying systems provides companies with information about how models behave outside controlled testing environments and allows developers to identify problems that may not appear during laboratory evaluations.

But Robinson argues that the approach becomes increasingly dangerous as AI systems become more capable and the consequences of unexpected behavior become harder to contain. His concern is that companies may be treating deployment itself as part of the safety-testing process at a point when the systems involved could have capabilities that make failures substantially more consequential.

The argument goes to a fundamental tension in the AI industry. Companies developing frontier models need real-world feedback to improve their systems, but they also face pressure to ensure that more capable models cannot exploit weaknesses in their safeguards, access unauthorized systems, or behave in ways their developers did not anticipate.

Robinson said AI capabilities are advancing faster than researchers’ understanding of alignment, the field concerned with ensuring that AI systems behave in accordance with human goals and values. That gap is growing wider as AI models move beyond generating text and images toward operating software, using tools, making decisions, and performing longer sequences of tasks with less direct human supervision.

A model that produces an incorrect answer can generally be corrected by a user. A more autonomous system that can interact with external systems presents a different category of risk if it misinterprets an objective, circumvents a restriction, or behaves in an unintended way.

Robinson’s argument is less about whether AI companies should stop developing new models and more about whether their safety processes are advancing quickly enough to match the capabilities they are creating.

OpenAI Says It Can Hold Back Models When Necessary

OpenAI disputed the suggestion that it is simply rushing models into deployment without sufficient safeguards.

“We’re making sure our models don’t become more capable than we can safely manage and secure, and we pause training or hold back models when we need to slow down,” an OpenAI spokesperson said.

That response points to a key difference between the company’s stated safety framework and Robinson’s criticism. OpenAI says development is constrained by its assessment of whether a model can be safely managed, while Robinson argues that the company’s development culture itself is making it difficult to reach the level of caution required.

The disagreement also comes amid a broader debate within the AI industry over the appropriate balance between rapid innovation and precaution.

OpenAI and rival Anthropic have faced scrutiny following incidents involving safety controls and unexpected model behavior. Such episodes have increased attention on how frontier laboratories test models, monitor them after deployment, and determine when a system’s capabilities warrant additional restrictions.

For AI companies, the stakes extend beyond the technical question of whether a model is aligned in a controlled environment. Increasingly capable systems are being integrated into products and workflows where they can interact with users, software, and external data. That makes failures potentially more difficult to isolate once a model is deployed at scale.

Robinson’s warning consequently raises a question that is becoming more important as the industry moves toward autonomous AI: can safety continue to be treated primarily as an iterative process that improves alongside deployment, or should certain capability thresholds require substantially more evidence of reliability before systems are released?

His comparison with nuclear power and aviation underscores the distinction he is trying to draw. Those industries operate on the premise that some failures are too consequential to discover through ordinary trial and error in live environments.

OpenAI’s response indicates that it believes its existing safeguards, preparedness processes, and ability to pause development can manage that risk. However, Robinson’s resignation and public criticism suggest that at least some people involved in those processes believe the pace of development is making that model of safety increasingly difficult to sustain.

The disagreement is unlikely to be resolved by a single model launch or safety report. As frontier systems become more capable, the industry’s credibility will increasingly depend on whether its testing and governance mechanisms can demonstrate that safety standards are keeping pace with capability growth.

For Robinson, the central problem is that the industry is approaching that threshold too quickly.

“The time for trial and error is over,” he wrote.

Intesa Sanpaolo has raised the stakes in its proposed takeover of Monte dei Paschi di Siena, offering MPS shareholders an additional €800 million ($900 million) in cash if they reject the Italian lender’s rival expansion plan.

The move escalates a high-stakes battle over the future of MPS and adds another layer to Italy’s rapidly consolidating banking sector, where a series of deals over the past two years has reshaped competition among the country’s lenders.

Intesa said on Saturday it would increase the cash component of its offer to €1.25 per MPS share, up from €1. Shareholders would also receive 1.6 newly issued Intesa shares for every MPS share they tender.

The revised terms represent Intesa’s latest attempt to persuade MPS investors to back its €34 billion share-and-cash takeover, which was announced in June and is the largest transaction so far among roughly a dozen banking deals in Italy during the past two years.

The enhanced offer comes ahead of an October 29 shareholder vote at MPS, where Chief Executive Luigi Lovaglio is seeking approval for an alternative strategy that would allow the bank to pursue its own expansion rather than accept Intesa’s bid. Lovaglio unveiled the counter-plan in August. It calls for MPS to make two separate all-share takeover offers for wealth manager Banca Generali and rival lender Banco BPM, effectively positioning MPS as an active consolidator rather than a takeover target.

The vote has become central to the contest because Italian takeover rules require Lovaglio to obtain shareholder approval before advancing the counter-plan.

Intesa said that if MPS shareholders approve either of the two proposed acquisitions on October 29, it would have the right to withdraw its own takeover offer under the conditions attached to the bid.

The bank made clear, however, that it does not intend to waive those conditions.

“Intesa Sanpaolo does not intend to exercise the right to waive the conditions … and will claim for their non-fulfilment,” it said.

That position raises the stakes of the shareholder vote. MPS investors are effectively being presented with competing visions: accept Intesa’s offer and become shareholders in a much larger banking group, or back Lovaglio’s strategy, which would seek to build MPS through acquisitions of Banca Generali and Banco BPM.

The increased cash component gives shareholders an additional financial incentive to choose Intesa’s proposal. Under the revised terms, investors who tender their MPS shares would receive €1.25 in cash for each share alongside the 1.6 newly issued Intesa shares.

The structure means the contest will depend not only on the headline value of Intesa’s offer but also on investors’ assessment of the future value of Intesa shares versus the potential returns from Lovaglio’s expansion strategy.

For MPS, the counter-plan would represent a significant change from its recent transformation. The bank has spent years rebuilding its balance sheet and reducing the legacy problems that made it one of the most troubled lenders in Europe following its 2017 state rescue.

Lovaglio’s strategy suggests management believes MPS can use its improved financial position to participate directly in Italy’s consolidation rather than surrender its independence to a larger rival.

The proposed acquisition of Banco BPM would be particularly significant because it would combine two major Italian banking franchises. Banca Generali, meanwhile, would give MPS a larger presence in wealth management, an increasingly important source of relatively stable fee income for banks seeking to reduce their reliance on lending margins.

But pursuing both transactions would also require MPS shareholders to accept the risks associated with a substantially more ambitious growth strategy.

Intesa’s approach offers a different proposition. Rather than asking MPS investors to finance additional acquisitions and wait for the benefits of integration, the revised bid provides immediate cash alongside shares in Intesa. The €800 million increase in the cash component therefore serves both as a direct enhancement of the offer and as a challenge to Lovaglio’s argument that MPS can create greater value by remaining independent.

The battle also illustrates the changing structure of Italian banking. A prolonged period of consolidation has encouraged lenders to pursue greater scale as they seek to improve efficiency, strengthen their ability to invest in technology and compete for customers and wealth-management assets.

The MPS contest is particularly consequential because of the institutions involved. Intesa is Italy’s largest bank, while MPS, founded in 1472, is the world’s oldest surviving bank and has become a symbol of the country’s difficult banking history.

The outcome will therefore determine more than the ownership of MPS. It could influence the next phase of consolidation in Italy by establishing whether shareholders favor a large-scale combination led by Intesa or a more aggressive independent strategy under Lovaglio. The October 29 vote is now likely to be the decisive battleground. Intesa has raised the immediate financial value of its proposal, while MPS management is asking shareholders to endorse a strategy that would transform the bank into an acquirer itself.

The competing offers now leave investors with a fundamental choice over MPS’s future: join a larger banking group under Intesa or accept the execution and integration risks of an independent expansion plan built around Banco BPM and Banca Generali.