The robotaxi market is moving from years of testing and ambitious promises into a more commercially meaningful phase, with a growing number of companies now operating paid autonomous ride services across the United States, China and, increasingly, other international markets.

Uber’s launch of autonomous rides in Zagreb on Wednesday, in partnership with Chinese self-driving company Pony.ai and Croatian mobility startup Verne, is the latest sign that the industry is beginning to build a global operating footprint. It also highlights an unexpected shift in the competitive landscape: Tesla, once widely expected to dominate the robotaxi market because of its enormous vehicle fleet, software capabilities and manufacturing scale, is not currently among the companies leading the commercial rollout.

In Zagreb, passengers can book an autonomous vehicle through Uber’s app, making it the first European city where Uber has offered self-driving rides. The service initially covers selected parts of the Croatian capital, including the city center, with a licensed operator aboard to supervise the vehicle as the companies work toward fully autonomous operations.

Pony.ai supplies the autonomous driving technology, Verne owns and operates the fleet, while Uber provides the customer network and booking infrastructure. The arrangement shows that technology developers do not necessarily need to build their own ride-hailing platforms, while companies such as Uber can gain exposure to autonomous vehicles without developing the underlying driving system themselves.

The Zagreb launch is notable because the partnership is already moving beyond a single-city experiment. Uber and Pony.ai announced last week that they plan to deploy more than 2,000 robotaxis across five European cities, expanding their existing partnership beyond Zagreb.

Pony.ai is also pursuing a much broader international expansion. The Chinese company said this week that it plans to deploy more than 4,000 robotaxis outside China, targeting markets in Europe, the Middle East and Asia as competition intensifies in its home market. Robotaxi revenue jumped 691.2% year over year in its second quarter, accounting for about one-third of Pony.ai’s total revenue for the first time.

The developments suggest that the robotaxi market is no longer simply a race between Tesla and Waymo, as it was often portrayed during the earlier stages of the autonomous-driving boom. A group of companies is now establishing commercial networks, with Waymo, Baidu’s Apollo Go, Pony.ai and WeRide among the most advanced operators.

Waymo remains the clearest leader in the U.S. market. By June, the Alphabet-owned company was providing more than 500,000 paid robotaxi trips a week across 11 cities with a fleet of about 3,500 vehicles, according to an autonomous-vehicle industry ranking.

China is producing an equally important competitive force. Baidu’s Apollo Go delivered 3.2 million fully driverless rides in the first quarter of 2026, with weekly rides exceeding 350,000 in March. Its cumulative public rides had surpassed 22 million by April, while its global footprint had reached 27 cities by May.

Pony.ai and WeRide are also expanding internationally, including through partnerships with established mobility platforms. Their strategy became necessary because robotaxis require more than autonomous-driving software. Operators need vehicles, fleet management, maintenance, mapping, regulatory approvals, customer acquisition, and a mechanism for matching passengers with available cars.

That is where Uber’s role becomes strategically important.

Rather than trying to become a vertically integrated autonomous vehicle manufacturer, Uber is building what could become a global distribution network for robotaxis. The company has partnerships with multiple autonomous-driving companies, allowing it to potentially deploy different technologies in different markets.

The approach also gives Uber a hedge against the possibility that one autonomous-driving technology ultimately emerges as dominant. If Waymo, Pony.ai, Baidu, WeRide, or another developer succeeds in a particular market, Uber can potentially integrate that operator into its platform rather than being forced to compete against it.

Tesla presents a striking contrast.

For years, Tesla was widely touted as one of the companies best positioned to dominate autonomous transportation. Its enormous installed vehicle base, Full Self-Driving software, vertically integrated manufacturing capabilities, and access to vast amounts of driving data gave it a potentially powerful foundation for a robotaxi network.

Yet the commercial rollout has been much slower than the company’s earlier projections suggested.

Tesla’s robotaxi service is currently operating in only a handful of U.S. markets and remains considerably smaller than Waymo’s network. Reuters reported in July that Tesla had accumulated about 2.5 million robotaxi miles, compared with roughly 220 million autonomous miles for Waymo, while Tesla had scaled back its previously aggressive expansion projections and adopted a more cautious city-by-city approach.

Tesla’s regulatory progress has also been uneven. The company sought approval for 5,000 robotaxis in Nevada but received permission for only 10 vehicles in Clark County, with conditions requiring human supervision.

That does not mean Tesla has been eliminated from the race. It remains one of the industry’s most consequential potential challengers. Tesla has begun preparing its purpose-built Cybercab, a vehicle designed without a steering wheel or pedals, and its autonomous fleet has continued to expand in the United States. Recent reports indicate that Tesla robotaxis in Austin have begun operating without human safety monitors on some rides.

The difference is that Tesla’s advantage remains more prospective than operational.

Waymo and the leading Chinese companies have spent years accumulating real-world autonomous miles and operating commercial services under defined geographic and regulatory conditions. Tesla has spent much of the same period promising that its software and enormous vehicle fleet would eventually allow it to leapfrog competitors.

But autonomous driving is not simply a software problem. A company must demonstrate that its system can handle edge cases reliably, operate safely across changing conditions, and satisfy regulators before it can put thousands of vehicles on public roads without human drivers.

The market is therefore beginning to reward operational experience and deployment density, not just technological claims.

A recent industry ranking illustrates the shift. Waymo ranked first, followed by Baidu Apollo Go, Pony.ai and WeRide, while Tesla ranked fifth. The ranking attributed the strong positions of the Chinese companies partly to their international expansion through partnerships with ride-hailing platforms including Uber, Lyft, Bolt and Grab.

The emerging competitive structure could eventually resemble the broader technology industry, with several layers of companies competing rather than a single winner.

Autonomous-driving companies will provide the “driver.” Vehicle manufacturers will supply the hardware. Fleet operators will manage vehicles and maintenance. Ride-hailing platforms will provide passengers and payments. Cities and regulators will determine where and under what conditions the vehicles can operate.

This structure could accelerate adoption because it reduces the amount of capital any individual company needs to deploy the entire system.

It also explains why Uber’s Zagreb launch matters beyond Croatia.

The immediate number of vehicles is small, and the presence of a licensed operator means the service has not yet reached the fully driverless model that represents the industry’s ultimate objective. But the commercial significance lies in the infrastructure being assembled around the technology.

A passenger does not need a separate robotaxi application. They can open Uber, request an UberX or Comfort ride and, when an autonomous vehicle is available, receive instructions for the self-driving vehicle. That makes autonomous transportation invisible to the customer. The robotaxi becomes another vehicle category within an existing transportation network.

Europe could become an important battleground as this model expands. Pony.ai and Uber’s plan for more than 2,000 robotaxis across five European cities indicates that companies are moving quickly to establish early positions before regulation and market structures become more settled.

SK Hynix has reached a tentative wage agreement with its South Korean workers that would pay at least 60% of this year’s bonuses in company shares, reshaping how employees will receive a windfall generated by the artificial intelligence boom as the chipmaker moves simultaneously to return billions of dollars to shareholders.

The agreement was necessitated by the growing financial and labor pressures created by SK Hynix’s surge in profitability. Demand for high-bandwidth memory, or HBM, used in AI accelerators has driven earnings sharply higher, producing unusually large employee bonuses while also intensifying investor demands for the company to distribute more of its cash.

Workers are expected to receive an average of 779 million won ($547,000) in compensation in 2026, according to Reuters calculations. The scale of the payouts indicates that the AI semiconductor boom is increasingly reaching beyond chipmakers’ income statements and into employee compensation.

Under the tentative agreement, 40% of the bonus would be paid in cash and another 40% in company stock that employees can cash out immediately. The remaining 20% would be provided as deferred stock compensation, with half available after one year and the rest after two years. The agreement still requires approval in a vote by union members.

The shift marks a significant departure from last year’s arrangement, under which SK Hynix agreed to distribute 10% of annual operating profit to workers in cash under a 10-year agreement. Some employees initially resisted management’s proposal to move more than half of the bonuses into shares, citing concerns over the volatility of SK Hynix’s stock.

That concern became relevant after the shares reached a record high in June on expectations for sustained AI-related demand before falling sharply as investors began questioning whether the enormous amounts being invested across the AI industry would generate adequate returns quickly enough.

The new structure gives SK Hynix and its employees a way to share the upside from the company’s performance while reducing the immediate cash burden on the business. Kim Yong-jin, a management professor at Sogang University, described the arrangement as a “win-win solution”, saying an all-cash payment could have strained the company’s cash reserves and generated public backlash over the size of employee payouts.

The agreement also includes a 6.3% increase in base wages. SK Hynix will have the ability to defer up to 3% of wages if the company records losses, introducing a degree of flexibility into its compensation structure should the semiconductor cycle turn.

“Labor and management came up with their own breakthrough without relying on external mediation or systems,” SK Hynix said.

The timing of the labor agreement is significant because SK Hynix is simultaneously embarking on one of the most aggressive shareholder-return programmes in South Korea.

On Wednesday, the company announced plans to buy back and cancel 40 trillion won ($28.6 billion) of its own shares, the largest such share cancellation by a South Korean listed company. The programme covers about 24.07 million shares and is scheduled to run from Aug. 20 through Nov. 19. SK Hynix said it believes its current market valuation does not adequately reflect the company’s underlying business strength, cash-generation capacity and longer-term growth prospects.

The company also raised its shareholder-return target, saying it plans to return more than 50% of cumulative free cash flow generated between 2025 and 2027 through a combination of share buybacks, cancellations and dividends. Its previous policy had targeted returns of up to 50% of cumulative free cash flow.

Investors responded immediately. SK Hynix shares jumped about 13% on Thursday following the buyback announcement, recovering some of the ground lost during the recent sell-off in technology stocks.

The rally shows how the company’s capital-allocation strategy has become almost as important to investors as its operating performance. SK Hynix has been one of the biggest beneficiaries of the AI infrastructure build-out, supplying advanced HBM memory to companies developing and deploying large-scale AI systems. Yet the extraordinary expectations surrounding AI have also made semiconductor stocks vulnerable to sharp swings whenever investors question the pace or profitability of AI spending.

That tension has placed SK Hynix in an unusual position. Its strong cash generation gives it room to reward both employees and shareholders, but the company must also preserve enough capital to fund the enormous investment required to maintain its technological lead in memory chips.

The employee stock component effectively links part of workers’ compensation to that same valuation cycle. Employees who receive immediately saleable shares can convert them to cash, while the deferred portion gives them a longer-term financial interest in the company’s performance. For SK Hynix, the arrangement limits the immediate cash outflow associated with record bonuses and potentially aligns employees more closely with shareholders.

The development also comes as South Korea’s major chipmakers face growing pressure to distribute more of the profits generated by the AI boom. SK Hynix and Samsung Electronics have both benefited from surging demand for advanced memory, while investors have increasingly pushed companies to return excess cash rather than allow large cash balances to accumulate.

Samsung is also reportedly preparing a shareholder-return programme worth more than 100 trillion won, potentially including a special dividend and a commitment to return 50% of free cash flow to shareholders, according to media reports. This comes after months of a face-off with employees over bonuses.

However, the combination of employee stock compensation and a record buyback points to a broader change in how SK Hynix is managing the proceeds of the AI semiconductor cycle. Rather than treating the boom purely as a source of higher wages or larger shareholder payouts, management is attempting to distribute the gains across both groups while retaining the financial capacity needed to compete in an increasingly capital-intensive industry.

The immediate market reaction is seen as an indication that investors welcomed the strategy. The longer-term test will be whether SK Hynix can sustain its exceptional profitability as AI memory demand expands, while avoiding the overinvestment and valuation pressures that have recently caused sharp swings in semiconductor stocks.