When Elon Musk talks about Tesla Robotaxi, he is rarely describing a product in isolation. He is outlining an alternate future for transport, one where cars drive themselves, owners earn money while they sleep, and Tesla evolves from an automaker into a dominant artificial intelligence and mobility platform.

By the end of 2025, that future was meant to be largely visible on American roads. Instead, what exists today is something more tentative: a limited rollout, heavy regulatory oversight, and a widening gap between ambition and execution that carries consequences far beyond Tesla itself.

Elon Musk framed 2025 as the year Tesla’s long-promised Robotaxi vision would finally crystallize into a mass-market reality. Fully autonomous paid rides, rapid expansion across US cities, and coverage for half of the American population were all placed firmly on the calendar.

As the year closes, according to a report by Business Insider, what exists on the ground is far more modest: limited pilot services in Austin and the San Francisco Bay Area, a small fleet size, and continued reliance on safety monitors. The gap between ambition and execution has once again become central to the Tesla story, raising deeper questions about the pace of autonomy, regulatory limits, and Tesla’s position in an increasingly competitive global EV and autonomous driving market.

Tesla Robotaxi today operates less like a transformative ride-hailing network and more like a controlled experiment. In Austin, where the pilot launch began, just over 30 vehicles are offering paid rides, with a safety monitor seated in the front passenger seat. In the San Francisco Bay Area, where regulations are stricter, a safety monitor remains behind the steering wheel. These operational details matter. They underline that Tesla is still operating within the bounds of supervised autonomy, even as Musk publicly touts progress toward full autonomy.

This outcome contrasts sharply with Musk’s own projections earlier in the year. In April, he said Tesla was “on track” to begin fully autonomous paid rides in Austin by June, followed by many other US cities by year-end. That timeline slipped almost immediately. By October, Musk was forecasting Robotaxi operations in eight to ten metro areas, including Nevada, Florida, and Arizona, subject to regulatory approval.

While Tesla has secured permits in some of those states and is hiring for Robotaxi-related roles nationwide, paid services remain confined to just two regions.

Fleet size has been another reality check. Musk suggested Tesla could have around 500 Robotaxis operating in Austin and more than 1,000 in the Bay Area by the end of the year. Crowdsourced tracking data points to a far smaller footprint, with roughly 35 vehicles in Austin and about 130 in the Bay Area. Tesla has registered more than 1,600 vehicles and nearly 800 drivers for its ride-hailing service, suggesting preparation for scale, but not yet execution at the level Musk described.

These shortfalls are familiar territory for Musk watchers. “Elon Time” has become shorthand for his chronic optimism around timelines, a trait he has openly acknowledged. For years, Tesla investors have tolerated this tendency on the assumption that delayed delivery is better than no delivery at all. The question now is whether that tolerance is beginning to erode as autonomy becomes the linchpin of Tesla’s long-term valuation story.

Robotaxi is not a side project. It sits at the heart of Tesla’s claim that it is fundamentally different from other electric vehicle makers. While many automakers sell EVs as products, Musk has long argued that a Tesla is a platform capable of generating recurring revenue through autonomous ride-hailing. This idea has helped justify Tesla’s market valuation, even as competitors close the gap in EV technology and manufacturing scale.

What makes Tesla’s approach distinctive and risky is its insistence on solving general autonomy using a camera-only system. Unlike rivals such as Waymo, which rely heavily on lidar and operate within tightly geofenced areas, Tesla is betting that vision-based systems trained on data from millions of consumer vehicles can scale more cheaply and broadly. If Tesla gets this right, the payoff could be enormous, allowing Robotaxi services to expand rapidly without expensive hardware retrofits.

However, the slower-than-promised rollout suggests that the last mile of autonomy is proving as difficult as critics have long argued. Edge cases, unpredictable human behavior, and the challenge of achieving regulatory trust are slowing progress. Regulators, particularly in California, continue to require human oversight, reflecting broader concerns about safety and liability. Any serious incident involving a Robotaxi would not only set Tesla back but could also tighten rules across the entire autonomous driving sector.

This caution is visible even in Musk’s more recent claims. While he said a Tesla without a safety monitor took him around Austin with “perfect driving,” those experiences remain anecdotal and limited. The company has not yet removed safety monitors from its paid public service, underscoring the difference between internal testing and regulated deployment.

The implications extend well beyond Tesla. Globally, the EV market is entering a more competitive and less forgiving phase. Chinese manufacturers such as BYD, XPeng, and others are pushing aggressively into both EVs and assisted-driving technologies, often supported by faster regulatory approval processes at home. In parallel, technology companies and automakers are forming alliances to spread the cost and risk of autonomy development, rather than betting everything on a single, vertically integrated vision.

Tesla’s Robotaxi delays also sharpen the contrast with companies like Waymo, which have achieved fully driverless operations in limited areas. While Waymo’s model is expensive and geographically constrained, it has crossed a regulatory and technical threshold that Tesla has yet to achieve at scale. The market is now weighing two competing philosophies: Tesla’s ambition to solve autonomy everywhere at once, and rivals’ preference for incremental, tightly controlled deployment.

There is also a growing tension between Robotaxi and Tesla’s consumer-facing Full Self-Driving system. FSD remains supervised for private owners, despite years of promises that unsupervised driving is imminent. Musk said earlier this year that FSD Unsupervised would arrive in many US cities before year-end, but that milestone has not been publicly achieved. Incremental changes, such as allowing limited phone use during certain driving conditions, hint at Tesla’s direction, but also highlight how far it still has to go.

As 2025 ends, Tesla Robotaxi stands as a symbol of both progress and restraint. The service exists, customers are paying for rides, and the technology continues to improve. At the same time, the bold milestones Musk outlined remain unmet, reinforcing skepticism about near-term autonomy at scale.

For Tesla, Robotaxi is not just about ride-hailing; it is about redefining the economic value of a vehicle and securing the company’s lead in a future shaped by software, data, and automation.

Some analysts believe that whether Musk’s optimism ultimately proves visionary or premature will depend on how quickly Tesla can turn cautious pilots into widespread, regulator-approved services. But for now, Robotaxi remains a promise in motion, inching forward, but far from the sweeping transformation Musk once said would define 2025.

Chinese internet search giant Baidu has taken a decisive step toward unlocking value from its artificial intelligence hardware ambitions, confirming that its AI chip unit Kunlunxin has confidentially filed for a listing on the Hong Kong stock exchange.

The Chinese search and AI giant said Kunlunxin submitted its application to the Hong Kong Exchanges and Clearing on January 1, laying the groundwork for a spin-off and separate public listing. While the offering size, valuation range, and timetable have yet to be finalized, the filing formalizes months of market expectations around Kunlunxin’s capital-raising plans.

Reuters previously reported that Kunlunxin was preparing for a Hong Kong IPO after a fundraising round that valued the business at about 21 billion yuan ($3 billion), suggesting that Baidu and its backers see growing investor appetite for domestic AI hardware platforms. The confidential filing route gives Baidu flexibility to test market sentiment while shielding sensitive financial and operational details during the early stages of the process.

Kunlunxin was founded in 2012 as an internal Baidu unit focused on developing custom AI chips to support the company’s search, cloud computing, and artificial intelligence workloads. Over time, it evolved into a more independently run business, although Baidu has retained a controlling stake and said Kunlunxin will remain a subsidiary after the proposed spin-off.

That structure mirrors a broader trend among Chinese internet and technology groups, which are increasingly carving out capital-intensive hardware units to give them clearer governance, access to dedicated funding, and strategic focus. For Baidu, separating Kunlunxin also allows investors to more directly value its chip ambitions, which require sustained spending on research, manufacturing partnerships, and ecosystem development.

While Baidu remains Kunlunxin’s largest customer, the chipmaker has expanded external sales over the past two years, supplying processors to other technology companies and data-center operators. That shift is important in the context of an IPO, as it signals a move beyond captive demand toward a more diversified revenue base. It also positions Kunlunxin as a potential national supplier at a time when Chinese companies face uncertainty over access to advanced foreign chips.

The listing plan comes against the backdrop of escalating U.S. export controls on advanced semiconductors, which have limited Chinese firms’ access to high-performance AI processors from companies such as Nvidia. In response, Beijing has intensified policy support for domestic alternatives, encouraging investment, public listings, and consolidation across the chip sector.

Several Chinese AI and semiconductor firms have recently moved toward public markets. Earlier this week, AI startup MiniMax said it expects to raise up to HK$4.19 billion ($538 million) in its Hong Kong offering. Semiconductor designer Shanghai Biren Technology raised HK$5.58 billion in its IPO, according to exchange filings. Other chip specialists, including OmniVision Integrated Circuits and GigaDevice Semiconductor, have begun bookbuilding, each targeting roughly $600 million.

This wave of listings has helped revive Hong Kong’s equity capital markets. According to LSEG data, the city raised $36.5 billion from 114 new listings in 2025, its strongest year since 2021 and more than three times the $11.3 billion raised in 2024. The rebound reflects a combination of pent-up demand from issuers, improved liquidity, and renewed interest in Chinese technology assets after a prolonged downturn.

The IPO could provide fresh capital for Kunlunxin to accelerate chip design, expand production capacity through foundry partners, and invest in software ecosystems that support AI workloads. Competition in China’s AI chip market is intensifying, with players such as Huawei’s Ascend unit and a range of startups vying to fill the gap left by restricted U.S. exports. Sustained funding will be critical for Kunlunxin to keep pace in performance, energy efficiency, and developer adoption.

The move underscores what central AI infrastructure has become to Baidu’s long-term strategy. As generative AI applications proliferate across search, cloud services, and enterprise software, control over underlying hardware is increasingly viewed as a strategic advantage rather than a cost center. Spinning out Kunlunxin allows Baidu to pursue that ambition while sharing the financial burden with public-market investors.

Although the timeline remains uncertain, the confidential filing signals confidence that market conditions are supportive enough to move forward. If successful, Kunlunxin’s listing would add to a growing roster of Chinese AI and semiconductor companies turning to Hong Kong to fund the next phase of development, reinforcing the city’s role as a key financing hub for China’s technology ambitions.

Elon Musk says Neuralink is preparing to cross a decisive threshold in the development of brain–computer interfaces, with plans to ramp up high-volume production of its brain implants in 2026 and automate much of the surgical process required to implant them in humans.

The move would mark a shift from cautious clinical experimentation to a manufacturing-led phase aimed at rapid scale, potentially transforming both neuroscience and the commercial trajectory of the BCI industry.

In a series of posts on X this week, Musk said Neuralink “will start high-volume production of brain-computer interface devices” this year, while the implantation procedure itself will move to a “streamlined, almost entirely automated surgical procedure” in 2026. He described the change as significant, particularly because the device threads would pass through the dura, the brain’s tough outer membrane, without requiring it to be removed.

The announcement underscores how aggressively Neuralink is trying to solve the two biggest bottlenecks facing implantable brain technologies: surgical complexity and manufacturing scale.

From lab-scale trials to mass production

Neuralink was founded in 2016 with the ambition of creating implantable chips that translate neural signals into digital commands, allowing users to control computers, phones, and other devices with their thoughts. While Musk has often spoken about long-term goals such as human–AI integration, the company’s near-term strategy has been tightly focused on medical use cases, particularly severe neurological conditions where existing treatments are limited.

Those include paralysis, spinal cord injuries, neurodegenerative diseases such as Parkinson’s and Alzheimer’s, and certain forms of vision impairment. In these cases, the potential benefits of invasive brain implants can outweigh the risks, making regulatory approval more feasible.

After years of animal testing and regulatory scrutiny, Neuralink implanted its first chip in a human patient in January 2024. That patient, Noland Arbaugh, who is quadriplegic, has since demonstrated the ability to control a computer cursor, type messages, and play games using neural signals alone. Arbaugh has said publicly that the implant has restored a sense of autonomy and social connection that he had lost after his injury.

By September 2025, Neuralink said 12 people worldwide had received implants and were actively using the system, a modest number that reflects the slow, regulator-driven pace of early human trials. Musk, however, has made clear that this is only the beginning. He has previously said Neuralink could exceed 1,000 patients by 2026, a target that would require a dramatic expansion in both device production and clinical capacity.

Automating the brain surgery bottleneck

One of Neuralink’s most persistent challenges has been the implantation process itself. The current approach involves a human surgeon removing a small section of the skull before a robotic system inserts dozens of ultra-thin electrode threads into specific regions of the brain. Each thread is about 20 times thinner than a human hair, a design intended to minimize tissue damage while capturing high-quality neural signals.

Musk’s claim that the procedure will become “almost entirely automated” suggests Neuralink is trying to turn neurosurgery into something closer to a standardized industrial process. Allowing the threads to pass through the dura without removing it could reduce surgical trauma, shorten recovery times, and lower the risk of complications, while also making procedures faster and more repeatable.

If successful, automation would address a core constraint: neurosurgeons are scarce, highly trained, and expensive, and scaling a technology that depends on manual brain surgery is inherently difficult. A largely robotic procedure could lower costs and enable Neuralink to treat far more patients, though it also raises questions about safety validation, liability, and regulatory oversight when machines, rather than surgeons, perform critical steps.

Manufacturing push and hiring signals

The production ramp Musk described has been building for some time. In late 2024, Neuralink went on a hiring spree, advertising roles for manufacturing technicians, microfabrication specialists, and robotics engineers. Those hires signaled an internal shift toward industrial processes rather than pure research and development.

High-volume production also implies greater standardization of the implant hardware itself. The Neuralink chip, roughly coin-sized, integrates signal processing, wireless communication, and power management, all of which must meet medical-grade reliability standards. Scaling production while maintaining consistency and safety will be a major test, particularly given the scrutiny applied to implantable medical devices.

However, Neuralink operates in a growing but still narrow field. Rivals such as Synchron and Blackrock Neurotech are also developing implantable BCIs, often with a more conservative approach focused on clinical partnerships and specific medical indications. Neuralink’s strategy is more vertically integrated, combining custom hardware, robotics, and software under one roof, and more expansive in its long-term vision.

That vision has made the company both influential and controversial. Musk has repeatedly suggested that brain implants could eventually be used by healthy individuals to enhance cognition or merge human intelligence with artificial intelligence. Such applications remain speculative and far from regulatory approval, but they shape how policymakers, ethicists, and the public view Neuralink’s trajectory.

Concerns include data privacy, long-term brain health, consent, and the societal implications of cognitive enhancement technologies. As Neuralink moves toward mass production, these debates are likely to intensify, particularly if the company begins to scale beyond small clinical populations.

Neuralink’s planned transition to high-volume production and automated implantation marks one of the most ambitious phases in its history. If the company succeeds, it could redefine what is technically and commercially possible in brain–computer interfaces, moving the field from experimental medicine toward something closer to a platform technology.

At the same time, the risks remain substantial. Scaling invasive brain implants is unlike scaling consumer electronics or software. Regulatory hurdles, unforeseen medical complications, and public trust will all shape how far and how fast Neuralink can go.