Qualcomm is in discussions to provide advanced chip-design services to China’s ByteDance, the parent company of TikTok, in what would represent an early and significant customer win for the U.S. semiconductor giant’s push into custom silicon services, according to four people familiar with the matter cited by Reuters.

The negotiations, if successful, would mark ByteDance as one of the first major clients for Qualcomm’s relatively new chip-design operation, expanding the company’s footprint beyond its traditional stronghold in smartphone modem chips. Qualcomm remains the world’s largest supplier of chips that manage cellular communications in mobile devices, but it is actively working to reduce its heavy dependence on that market amid slowing global smartphone shipments and rising component costs.

The talks center on designing custom chips for ByteDance, with a portion of the technology drawing from AlphaWave Semi, a high-speed connectivity specialist that Qualcomm acquired last year. One source indicated the discussions include video processing units (VPUs), with an eye toward potential mass production by the end of this year.

ByteDance has also been developing its own AI inference chips and custom central processing units (CPUs), according to earlier reports, suggesting the collaboration could complement its in-house efforts.

While the discussions are progressing, the outcome remains uncertain. ByteDance could still pursue alternative partners, and it is unclear whether the talks will result in a finalized design or eventual manufacturing. The sources spoke on condition of anonymity because the matter is private.

Diversification Amid Smartphone Headwinds

A deal with ByteDance would deliver a timely boost for Qualcomm as it navigates challenges in its core business. Global smartphone shipments are on track for their steepest annual contraction on record this year, partly due to surging memory chip prices that have squeezed device makers. By expanding into custom chip design services, an area where rivals like Broadcom and Marvell have already made significant inroads, Qualcomm aims to tap into the booming demand for specialized silicon across data centers, AI, and other high-growth segments.

Qualcomm is already developing three types of data center chips: CPUs, inference accelerators, and application-specific integrated circuits (ASICs). The ByteDance discussions align with this broader strategy to move up the value chain and capture more of the AI infrastructure buildout. ByteDance, one of the world’s most valuable private companies, is aggressively investing in AI and video technologies to support TikTok’s massive global user base and its expanding enterprise ambitions.

The potential partnership also highlights a nuanced reality in U.S.-China tech relations. While Washington has imposed strict controls on advanced AI chips, affecting companies such as Nvidia, AMD, Applied Materials, and Lam Research, business continues in less sensitive areas. Qualcomm’s modem and connectivity expertise falls outside the most restricted categories, allowing it to explore opportunities that many pure-play AI chipmakers cannot.

ByteDance’s Push for Greater Self-Reliance

For ByteDance, working with Qualcomm could accelerate its efforts to build more robust in-house technology capabilities. The company has been investing heavily in semiconductors as it seeks to reduce reliance on foreign suppliers for critical components, particularly in AI and video processing. A custom chip collaboration would provide access to Qualcomm’s design expertise while potentially integrating elements from the AlphaWave acquisition, which specializes in high-speed connectivity solutions well-suited for data-intensive applications.

This fits into a larger pattern among Chinese tech giants. As U.S. export restrictions tighten on cutting-edge AI hardware, companies like ByteDance are pursuing a dual-track approach: developing domestic alternatives while selectively partnering with U.S. firms in permitted areas. The talks also come as ByteDance continues to face regulatory scrutiny in Western markets over data privacy and national security concerns related to TikTok.

Qualcomm’s move into custom design services is believed to be an indication of a maturing semiconductor market where hyperscalers and large tech platforms increasingly prefer tailored solutions over off-the-shelf chips. This trend has created a lucrative new market for design expertise, but it also requires navigating complex geopolitical waters.

Success in the ByteDance negotiations is expected to open doors for Qualcomm with other major Chinese tech players hungry for advanced but compliant silicon. It would also validate the company’s diversification strategy at a time when the smartphone market weakness has weighed on its financial performance.

For the wider industry, the talks also show that commercial interests continue to drive engagement between U.S. chip firms and Chinese tech giants, even as strategic competition intensifies. While advanced AI chips face heavy restrictions, areas like connectivity, video processing, and specialized modems remain active ground for collaboration.

Industry analysts expect the outcome of these discussions to have ripple effects in the market. This is because a partnership might strengthen Qualcomm’s position in the custom ASIC market, where competition is heating up. It could also accelerate ByteDance’s AI and video infrastructure ambitions, helping it maintain a competitive edge in the global digital entertainment and enterprise sectors.

Taiwan’s ASE Technology Holding is undertaking one of the most aggressive capacity expansion programs in the semiconductor industry as it positions itself for what it believes will be years of sustained demand from artificial intelligence.

Speaking on Wednesday, Chief Operating Officer Tien Wu said the world’s largest chip packaging and testing provider is adding 15 new facilities this year, reflecting growing confidence that demand for advanced AI chips will continue to accelerate well beyond the current investment cycle.

The expansion includes six greenfield facilities for ASE, seven new sites for its subsidiary Siliconware Precision Industries (SPIL), as well as facilities acquired earlier this year from Innolux Corporation.

The scale of the buildout underscores a significant shift in the semiconductor industry. While much attention has focused on chipmakers such as Nvidia, Advanced Micro Devices (AMD), and Taiwan Semiconductor Manufacturing Company (TSMC), packaging and testing have emerged as one of the industry’s most critical bottlenecks.

Advanced AI processors require sophisticated packaging technologies that allow multiple chips, memory modules, and processors to work together at high speed while managing power consumption and heat generation. As AI models become greater and more complex, demand for these advanced packaging capabilities has surged.

Wu reiterated that ASE’s capital expenditure budget for the year stands at $8.5 billion but indicated spending could exceed that figure as the company races to add capacity.

The investment level places ASE among the biggest beneficiaries of the global AI infrastructure boom, alongside foundries, cloud providers, and data-center operators.

Notably, Wu emphasized that the expansion is not designed merely to address near-term demand.

“The factory expansion is not just for the next two years, but for 2029 and beyond,” he said, signaling management’s belief that AI-related semiconductor demand is entering a multi-year structural growth phase rather than a short-lived investment cycle.

That outlook aligns with forecasts from major technology companies, which continue to commit hundreds of billions of dollars toward AI infrastructure, cloud computing, and next-generation data centers.

ASE’s strategy also highlights the increasing globalization of semiconductor supply chains amid geopolitical pressure to localize production.

Wu said the company has already established two testing facilities in California and plans to add two more factories in the United States. The expansion was spurred by growing demand from customers seeking greater geographic diversification and aligns with Washington’s broader efforts to strengthen domestic semiconductor manufacturing capabilities.

The company is also evaluating potential investments in Arizona, one of the fastest-growing semiconductor hubs in the United States.

While Wu did not provide details, the state has become a focal point for chip investments, attracting major projects from TSMC, Intel, and other semiconductor companies seeking to build capacity closer to U.S. customers.

“At a particular customer’s request,” Wu said, ASE has been evaluating Arizona investment plans but must carefully assess the scale and nature of any future commitment.

ASE has been in a growing relationship with Nvidia. Last year, Nvidia announced plans to help build as much as $500 billion worth of AI server infrastructure in the United States through a network of manufacturing and supply-chain partners that includes Siliconware Precision Industries.

SPIL is one of Nvidia’s key packaging suppliers and plays an important role in preparing the advanced processors used in AI servers and data centers.

Yet despite Nvidia’s massive U.S. infrastructure ambitions, SPIL has not formally announced a major American investment tied to the initiative. That suggests the company remains cautious about balancing customer requirements, capital allocation, and long-term utilization rates before committing to large-scale overseas expansion.

The broader significance of ASE’s expansion lies in what it reveals about the next phase of the AI race. While much of the attention has focused on training ever-larger AI models, the industry’s ability to scale increasingly depends on physical infrastructure and manufacturing capacity. Packaging and testing have become strategic assets because they determine how quickly advanced chips can move from production lines into AI servers.

Industry analysts now see advanced packaging as one of the most valuable segments of the semiconductor value chain. The scarcity of packaging capacity has, at times, constrained shipments of AI accelerators even when sufficient wafers were available from foundries.

ASE’s willingness to spend more than $8.5 billion and potentially exceed that amount is seen as an indication of confidence that demand from hyperscalers, cloud providers, and AI developers will continue to outpace supply for years.

The company’s expansion plans also provide another indication that semiconductor suppliers expect AI spending to remain robust well into the next decade, with capacity decisions now being made based on demand projections extending beyond 2029 rather than merely responding to the current surge in orders.