Artificial intelligence is already reshaping the ride-hailing business in a way that would have seemed futuristic only a few years ago: some Uber drivers are now relying on Tesla’s Full Self-Driving system to handle part of the driving task, and Uber’s chief executive is openly acknowledging it.

In remarks on Peter Diamandis’ Moonshots podcast, Uber CEO Dara Khosrowshahi said the company has “tens of thousands” of Teslas operating on its platform and confirmed that some drivers are using Tesla’s Full Self-Driving, or FSD, while transporting passengers.

“We’ve got tens of thousands of Teslas on our platform now,” Khosrowshahi said. “And some of our drivers use FSD. Sure. So we’ve got a lot of data. It’s a great car. It’s a safe car.”

The disclosure offers a striking glimpse into how AI-assisted driving is quietly entering the mainstream ride-hailing economy, long before fully autonomous robotaxis become widespread. While Tesla markets FSD as an advanced driver assistance system rather than a fully autonomous platform, its growing use by Uber drivers effectively means passengers are already experiencing AI-assisted rides in human-supervised settings.

This is not, strictly speaking, illegal.

At present, there are no US laws that explicitly prohibit ride-hail drivers from using advanced driver assistance systems during trips, provided they remain fully attentive and retain control of the vehicle. Tesla itself states that active safety and driver-assistance features are designed to assist, not replace, the driver, and that the human behind the wheel remains responsible for safe operation at all times.

Uber’s own policy aligns with that standard. The company requires drivers using such systems to keep at least one hand, and normally both hands, on the steering wheel while the technology is engaged.

However, the development raises fresh questions about liability, safety, and public trust. Tesla’s driver-assistance systems have long been under legal and regulatory scrutiny. Several incidents involving Autopilot and FSD have intensified debate over whether the branding and real-world use of the technology outpace its actual capabilities.

Tesla has come under the National Highway Traffic Safety Administration (NHTSA)’s investigation following the growing number of crashes involving the EV maker’s vehicles.

One notable case cited in earlier reporting involved an Uber driver using Tesla’s FSD who crashed into an SUV while the system was active. That case underscores a central tension in the emerging AI mobility market: these systems are sophisticated enough to take over large portions of the driving task, but not advanced enough to absolve the human driver of responsibility.

For Uber, however, the shift is expected to evolve into its broader robotaxi strategy. Khosrowshahi’s comments indicate that Uber is treating today’s human-supervised AI driving as a bridge to tomorrow’s autonomous network.

He went further, saying Uber would welcome Tesla’s future robotaxis onto its platform once the company’s camera-only autonomous system is proven sufficiently safe for fully driverless deployment.

“When the day comes when those Teslas are safe enough with a camera-only approach, we’d love to have those Teslas on our platform as well,” he said.

This is consistent with Uber’s broader strategy of rebuilding its position in autonomous mobility through partnerships rather than in-house development. After exiting its own self-driving programme years ago, Uber has increasingly aligned itself with external autonomous vehicle players, including Waymo, Zoox, Pony.ai, and WeRide. More recently, it has expanded its ambitions through fresh robotaxi rollout plans in multiple global cities.

The company appears to be positioning itself as the distribution layer for whatever form autonomy ultimately takes, whether human-supervised Teslas today or fully driverless fleets tomorrow. That approach is commercially pragmatic. Uber no longer needs to win the technology race itself if it can remain the platform through which autonomous mobility is monetized at scale.

For Tesla, meanwhile, the development offers a preview of how its FSD ecosystem may evolve commercially beyond private ownership. Even before the company’s dedicated Robotaxi service achieves broad regulatory clearance, Tesla vehicles are already functioning as what some analysts describe as “proto-robotaxis” on platforms like Uber, where AI handles parts of the trip under human oversight.

The broader implication is that the line between assisted driving and autonomous ride-hailing is becoming increasingly blurred. Passengers booking an Uber may still see a driver in the front seat, but in a growing number of cases, AI is already doing part of the work.

That may be the most consequential transition underway for the ride-hailing industry: not the sudden arrival of driverless cars, but the gradual normalization of shared control between human drivers and machine intelligence.

Chinese semiconductor companies posted blockbuster revenues in 2025, turning years of U.S. export restrictions into an unexpected tailwind that has accelerated Beijing’s push for homegrown technology and reshaped supply chains across the industry.

Semiconductor Manufacturing International Corp. (SMIC), the country’s flagship foundry, reported full-year sales of $9.3 billion — a 16 percent jump and the highest in its history. Analysts at LSEG expect the figure to climb above $11 billion this year.

Hua Hong Semiconductor notched a record fourth-quarter revenue of $659.9 million and guided for similar strength in the current period. Domestic GPU hopeful Moore Threads forecast 2025 sales between 1.45 billion and 1.52 billion yuan ($209–220 million), an eye-popping 231 percent to 247 percent increase from 2024. Even more striking was the performance in memory chips.

ChangXin Memory Technologies (CXMT) more than doubled revenue to over 55 billion yuan ($8 billion), according to people familiar with the numbers quoted by CNBC.

The surge reflects a perfect storm of exploding demand for AI infrastructure inside China, a persistent global shortage of memory, and Washington’s tightening export controls that have forced local tech giants to buy Chinese wherever possible.

Paul Triolo, a partner at Albright Stonebridge Group who has tracked the sector for years, called the restrictions “rocket fuel” for Chinese chip demand. What started as an attempt to slow Beijing’s technological rise has instead poured money and urgency into domestic fabs.

Mature-node chips used in electric vehicles and industrial equipment have provided a steady baseline, while advanced components for AI data centers are “through the roof,” Triolo told CNBC.

The memory segment offers the clearest example. High-bandwidth memory (HBM), the specialized DRAM that powers large AI models, remains tightly controlled by the United States. With Samsung, SK Hynix, and Micron largely blocked from selling their latest generations to China, CXMT has stepped into the breach. Its HBM2 and HBM2e products may trail the industry leaders by several generations, but Chinese buyers have embraced them with open arms simply because they are available and produced locally.

Morningstar senior equity analyst Phelix Lee noted that “even the technologically inferior HBM2 or HBM2e are met with enthusiasm” now that foreign supply is restricted. CXMT is on track to begin producing HBM3 this year, a significant step forward.

“After HBM is restricted into China, CXMT is picking up as the only homegrown alternative, so even the technologically inferior HBM2 or HBM2e are met with enthusiasm,” Lee told CNBC.

The skills being honed in memory manufacturing are already spilling over. Triolo pointed out that China’s memory fabs have become unexpected “incubators for advanced process technology” in ways that would have been unthinkable before the October 2022 export controls. The precision required for high-volume DRAM production is sharpening expertise that can eventually be applied to logic chips and GPUs.

Huawei has emerged as the most visible buyer, quietly building an entire domestic semiconductor ecosystem to keep its servers, smartphones, and AI systems running.

“While China does not yet lead in peak GPU performance, these homegrown solutions are filling the domestic ‘compute gap’ and driving record revenues,” Parv Sharma, senior analyst at Counterpoint Research, told CNBC.

However, the picture is far from uniformly rosy. Chinese foundries like SMIC and Hua Hong remain unable to produce the most advanced chips at the scale and yields achieved by Taiwan’s TSMC. They lack access to the latest extreme ultraviolet lithography machines from Dutch supplier ASML, and domestic alternatives are still years from matching that capability. The gap in cutting-edge technology persists even as revenues soar.

There is also the growing risk of overcapacity in older nodes. As Sharma noted, much of the current growth is driven by “import dependence replacement.” Once that wave crests, sustaining momentum will require Chinese firms to climb the value chain into advanced HBM and next-generation logic processes — a far harder task.

Still, the momentum is real and self-reinforcing. What began as a defensive scramble has evolved into a comprehensive, state-orchestrated effort to rebuild entire segments of the semiconductor supply chain from the ground up. Beijing has poured hundreds of billions of dollars into the sector through subsidies, talent programs, and national champions.

The record 2025 revenues show that, in the areas where policy and market forces align, Chinese chipmakers are no longer just catching up — they are scaling at a pace that is forcing the rest of the world to take notice.

Removing a difficult manager is rarely as straightforward as filing a complaint with senior leadership, according to former Amazon vice president Ethan Evans, who says employees often underestimate the institutional resistance that protects poor leadership.

In remarks on The Peterman Pod, Evans offered a blunt assessment of how workplace complaints are typically received at the top: senior managers may have strong incentives, even if subconscious, to discount concerns raised about one of their direct reports.

“If you come to me with a weakness in one of my employees,” Evans said, the instinctive calculation is whether to dismiss the complaint as oversensitivity or accept it and trigger a much larger operational problem.

That problem, he explained, can quickly escalate into a leadership crisis. If the complaint is deemed credible, the senior executive may need to manage the person out, oversee a replacement process, and absorb the disruption that follows, often while carrying the additional workload themselves.

“So you can see why, even if it’s subconscious, I have a lot of reasons not to listen or not to believe very easily,” he said.

His comments offer a rare insider look into the political mechanics of workplace hierarchy, where the process of addressing bad management often collides with incentives around continuity, optics, and workload.

At its core, Evans’ argument is less about whether complaints are justified and more about how organizations process risk.

For skip-level managers, acknowledging that a subordinate manager is failing does not merely validate an employee’s grievance. It can also raise uncomfortable questions about their own judgment in hiring, promoting, or supervising that person. In many organizations, this creates an institutional bias toward delay, denial, or minimization.

This is why, Evans argues, lone complaints frequently go nowhere. His advice is deliberately collective: “Never mutiny alone.”

Rather than escalating concerns in isolation, employees should first compare experiences with trusted colleagues to determine whether the issue is systemic or simply a mismatch in management style.

That “sanity check,” as he describes it, serves two purposes. First, it helps separate subjective frustration from an objectively harmful pattern. Second, it transforms a personal grievance into a team-level operational concern, which senior leadership is far more likely to take seriously.

When multiple employees raise the same issue, the complaint ceases to look like a personality clash and starts to resemble a leadership failure. Evans said that in one case involving a problematic leader, he would likely not have acted on a single complaint. But once several consistent reports surfaced, it became clear that intervention was necessary.

In a follow-up email to Business Insider, he expanded on that point, stressing that documentation is often the difference between being heard and being dismissed. The most effective approach, he said, is to present at least three clear, fact-based examples, ideally corroborated by others, and frame the issue in terms of business impact rather than personal emotion.

One of the most common mistakes employees make, according to Evans, is allowing frustration to dominate the complaint. Bitter or emotionally charged grievances can be easily reframed by leadership as interpersonal conflict or hypersensitivity.

A more effective strategy is to acknowledge what the manager does well before outlining specific shortcomings and their consequences for team performance, morale, retention, or delivery.

This approach shifts the conversation from accusation to risk management. Senior leaders are far more likely to intervene when they believe the manager’s behavior is creating measurable organizational harm, particularly if top talent is leaving or if the conduct introduces legal, ethical, or compliance risks.

Absent those factors, complaints often remain easy to ignore.

Play Chess Not Checkers

Evans also outlined an alternative route for employees who may not want to confront leadership directly: reposition the issue as a business case for internal mobility. Rather than criticizing the manager, he suggests making a case for moving to another team by emphasizing where one’s skills can better serve the organization.

“Don’t even bring up the manager,” he said. “Just say, ‘hey, I was looking at this other role, and I think I could do so much more for you and the org over here because of A, B, and C.'”

Ultimately, Evans admitted navigating these situations requires careful strategy: “You’ve gotta play chess, not checkers.”

This is, in effect, corporate diplomacy.

It allows employees to exit a damaging reporting line without forcing leadership into a defensive posture. In many companies, this route may be more practical than seeking to have a manager removed, especially where power structures are entrenched.

Evans’ remarks also tap into a broader management problem that extends well beyond any one company.

Across corporate America and the wider global workplace, ineffective managers are often the by-product of flawed promotion systems. High-performing individual contributors are frequently elevated into leadership roles without the training, emotional intelligence, or coaching skills required to manage people effectively.

This structural issue has been widely documented by management experts and labor economists, who argue that technical excellence does not automatically translate into leadership competence. As organizations continue to flatten hierarchies and expand spans of control, many managers now oversee larger teams with fewer resources, intensifying the pressure and often exposing weaknesses in communication and people management.

Evans draws an important distinction between bad managers and underdeveloped ones. Some, he says, are simply untrained and can improve with coaching and support. The more troubling category consists of leaders who view questioning as insubordination and default to rigid top-down authority, especially under stress. These are often the cases where escalation proves most difficult, particularly if higher-level leaders share the same blind spots.

Evans describes this as “stacked flaws” — situations where the leadership chain reinforces the same dysfunctional management style, making recognition and correction unlikely. In such environments, he argues, employees may need to make a harder calculation.

If colleagues are unwilling to corroborate concerns and the culture discourages dissent, the most strategic move may not be escalation at all, but exit. Sometimes, he suggests, the problem is not just the manager but the organization’s culture itself.

In that sense, his advice is ultimately less about office politics and more about institutional reality: changing a bad boss requires evidence, allies, and timing. Without those, the smarter play may be to protect one’s career by moving on.