SK Hynix is facing renewed labor tensions as management and representatives of one of its South Korean unions entered a fifth round of negotiations on Tuesday over a lucrative bonus package.

The talks, taking place at the company’s semiconductor manufacturing complex in Cheongju, center on a proposed overhaul of SK Hynix’s profit-sharing scheme after record earnings driven by soaring demand for AI memory chips.

Last year, SK Hynix agreed to allocate 10% of its annual operating profit to employees under a new bonus formula. With the company posting record profits from sales of high-bandwidth memory (HBM) chips used in artificial intelligence servers, Reuters calculations indicate the programme could deliver an average payout of about 779 million won ($547,127) per employee in 2026, one of the richest bonus packages in South Korea’s corporate sector.

Management has proposed paying more than half of those bonuses in company shares instead of cash while restricting employees from selling the stock for a specified period. According to a union document reviewed by Reuters, the company also wants greater flexibility to reduce bonus payments in years when it records losses.

The union has rejected the proposal, arguing that it transfers market risk from shareholders to employees.

“We would absolutely not accept any arrangement in which union members bear the risk of stock-price fluctuations,” the union said in a note to members.

It warned that if management fails to present “a concrete and forward-looking revised proposal” during Tuesday’s negotiations, “resolving the matter through dialogue alone would be difficult,” adding that it would “take necessary action.”

SK Hynix has become one of the world’s biggest winners from the AI boom as Nvidia’s primary supplier of high-bandwidth memory chips, a technology that has become indispensable for training and running large AI models. Strong demand from hyperscale cloud providers and AI developers has transformed the company’s financial performance, lifting operating profits to record levels and fueling expectations among employees that compensation should rise in tandem.

For management, paying part of bonuses in equity would reduce immediate cash outflows, strengthen employee ownership and encourage longer-term retention. For workers, however, stock-based compensation introduces exposure to share-price volatility while limiting their ability to immediately realize the value of their bonuses.

Those concerns have become more pronounced because SK Hynix shares have experienced sharp swings this year. Although the stock remains up more than 130% since the beginning of the year, it has surrendered nearly half of its value from the record highs reached in June as investors reassessed valuations across AI-related semiconductor companies amid concerns over the sustainability of AI infrastructure spending.

The recent decline has strengthened the union’s argument that employee compensation should not depend on unpredictable market movements.

Part of A Wider Labor Dispute Across South Korea’s Chip Industry

The disagreement at SK Hynix comes against the backdrop of broader labor unrest that has swept through South Korea’s semiconductor sector during the AI boom. The country’s chipmakers have enjoyed unprecedented revenue growth as demand for AI processors and memory chips surged, but employees have been demanding that the financial windfall be shared more generously.

Samsung Electronics, SK Hynix’s largest domestic rival and the world’s biggest memory chip producer, has also been embroiled in labor disputes over wages, bonuses and working conditions over the past two years. Samsung faced a series of strikes and work stoppages after unions said that employee compensation had failed to keep pace with the company’s earnings potential and the strategic importance of its semiconductor business.

Although Samsung and SK Hynix face different labor issues, both disputes highlight a structural shift in South Korea’s semiconductor industry, where organized labor is becoming more assertive after decades in which management largely dictated compensation policies.

The labor pressures also arrive at a critical moment for the industry. South Korean chipmakers are investing tens of billions of dollars to expand production capacity for AI memory chips while competing with rivals including Micron Technology and Chinese semiconductor manufacturers. At the same time, companies are trying to balance higher labor costs, volatile capital markets and massive investment requirements needed to maintain leadership in advanced chip manufacturing.

However, the outcome of the negotiations carries implications beyond employee compensation.

Stable labor relations are important as SK Hynix races to meet strong demand for next-generation HBM chips from customers including Nvidia and other AI infrastructure providers. Any escalation in labor tensions could complicate production planning at a time when advanced AI memory remains in short supply globally.

The Trump administration is preparing new restrictions that would prohibit imports of new Chinese-made optical transceivers used in U.S. data centers, escalating Washington’s efforts to secure critical artificial intelligence infrastructure and reduce reliance on Chinese technology in strategic sectors.

According to Reuters, the Federal Communications Commission is drafting a measure that would block imports of new models of Chinese optical transceivers, networking components that transmit data through fiber-optic cables at extremely high speeds inside data centers that power AI training and cloud computing.

The proposal, which has not been previously reported, is expected to be published later this year if approved. However, the sources cited by Reuters cautioned that the FCC could still revise or abandon the plan before implementation.

The move represents the latest step in the Trump administration’s broader efforts to prevent Chinese technology from becoming deeply embedded in America’s AI supply chain, where officials fear it could create long-term national security vulnerabilities.

Optical transceivers are essential components of modern data centers, enabling ultra-fast communication between thousands of servers and AI processors.

As investment in AI infrastructure accelerates, U.S. officials are increasingly scrutinizing every layer of the hardware stack, extending beyond advanced semiconductors to include networking equipment that keeps massive AI clusters connected.

The proposed restrictions are intended to reduce the risk that Chinese-made networking equipment could be used to steal sensitive data, introduce malicious software or disrupt the operation of facilities that host advanced AI models.

“Transceivers definitely pose a risk,” said Divyansh Kaushik, an AI policy expert at Beacon Global Strategies.

“As the data center buildout scales up, you want to make sure the data center supply chain is secure from the get-go.”

The U.S. national security policy treats AI infrastructure as critical infrastructure comparable to telecommunications networks or power grids.

China Warns of Retaliation

China criticized the reported proposal, warning that it would respond if its interests were harmed. A spokesperson for the Chinese embassy in Washington urged the United States to “heed the objective and rational voices of the business communities in both countries” and “stop smearing Chinese companies and threatening them with sanctions.”

The embassy added that “China will take all necessary measures in response to any action that causes material harm to its interests.”

The latest tensions come even as Washington and Beijing have sought to stabilize parts of their economic relationship following last year’s easing of trade tensions.

Officials involved in the proposal are reportedly determined to avoid repeating what many in Washington consider the Huawei experience.

During previous administrations, equipment from Huawei Technologies became deeply integrated into U.S. telecommunications infrastructure before national security concerns prompted extensive efforts to remove it.

Replacing Huawei equipment ultimately proved costly and time-consuming, amplifying the view among U.S. policymakers that potential security risks should be addressed before foreign technology becomes entrenched. The proposed restrictions on optical transceivers mirror that preventive approach by targeting components before they become more deeply embedded in rapidly expanding AI data centers.

Chinese Manufacturers Could Be Hardest Hit

A ban would primarily affect Zhongji Innolight, one of the world’s largest suppliers of optical transceivers. According to Counterpoint Research, Innolight controls approximately 27% of the global data center transceiver market, making it the industry’s largest supplier.

The company was added to the Pentagon’s list of alleged Chinese military-linked companies in June, a designation that often precedes broader U.S. restrictions, although it does not itself impose sanctions.

Industry research cited by Reuters indicates that roughly 90% of Innolight’s revenue is generated outside China, underscoring its importance to global cloud infrastructure providers.

Analysts expect the restrictions to increase costs for major U.S. cloud operators, including Amazon Web Services, by limiting access to one of the industry’s largest hardware suppliers. American manufacturers including Coherent and Lumentum Holdings are expected to benefit from any shift away from Chinese suppliers.

However, analysts note that neither company currently has sufficient manufacturing capacity to fully replace Chinese production.

A report by the Foundation for American Innovation concluded that while U.S. firms produce competitive technologies, scaling output quickly enough to substitute for Chinese suppliers would present a significant challenge.

The proposed action would continue the FCC’s aggressive use of its Covered List, a national security mechanism created by Congress to restrict future imports and sales of equipment produced by companies deemed to pose security risks.

In recent months, the regulator has expanded restrictions beyond telecommunications equipment to include Chinese drones, networking routers, industrial robots and power inverters.

According to Reuters, the optical transceiver proposal would follow a similar approach by prohibiting imports of new Chinese models while allowing many non-Chinese manufacturers to continue introducing new products into the U.S. market.

The initiative also reveals the FCC’s growing role in U.S. technology policy. Reuters reported earlier this year that the U.S. Department of Commerce paused several planned restrictions on Chinese technology imports after a trade détente was reached with Beijing last October, including measures affecting data center equipment.

The FCC has since emerged as one of the administration’s most active agencies in limiting Chinese technology, particularly after a Supreme Court ruling in June strengthened President Donald Trump’s authority over certain independent regulatory agencies.

That shift has enabled the commission to pursue a series of national security-related restrictions independently of broader trade negotiations.

However, the proposed restrictions signal that Washington’s technology competition with Beijing is expanding beyond semiconductors into the broader infrastructure underpinning artificial intelligence. While export controls have largely focused on limiting China’s access to advanced AI chips, the latest proposal targets the networking hardware that enables those processors to operate at scale.