The AI infrastructure race is entering a phase where access to silicon is no longer the only constraint. As compute demand expands into multi-gigawatt projects, the industry is increasingly confronting a second bottleneck: how to finance hardware and data-center capacity at the scale frontier-model developers now require.
The AI infrastructure race has been marked by a near-fixated drive to expand AI-powered buildouts, as demand for increasingly complex chip configurations pushes against the physical and economic limits of scalability.
Inergy Technology said strong demand for AI server cooling and power management pushed the share of server-related applications to 47% of revenue in the second quarter of 2026, with shipments of system-on-chip (SoC) cooling fan driver ICs set to ramp up in the second half of 2026 and battery backup unit (BBU) demand expected to keep rising as AI moves toward high-voltage direct current power architectures.
Nvidia's AI memory orders helped drive a sharp split between SK Hynix and Samsung Electronics in the first half of 2026. According to Chosun Biz, Nvidia became SK Hynix's biggest customer in the first half of 2026, while it did not rank among Samsung's top five revenue sources.
Xiaomi is expanding its in-house semiconductor push from smartphones into AI acceleration and autonomous driving, unveiling three Xring processors that deepen its reliance on TSMC manufacturing while reducing dependence on external chip suppliers.
The memory industry is undergoing a structural shift as AI demand surges, and Etron Technology chairman Nicky Lu said the sector's boom will stretch beyond 2027, with shortages likely lasting through 2028 and possibly into 2030. He said supply is tightening further as AI and edge computing create rigid demand, pushing the supply chain to focus on higher-value applications.
TPK Holding is accelerating its move from consumer touch panels into semiconductors, with its Through Glass Via (TGV) pilot line in Taoyuan, Taiwan set to begin operations in August or September. The company is also working with leading outsourced semiconductor assembly and test (OSAT) providers on the project, while its new Thailand production base aims to complete trial runs by year-end and begin volume production in early 2027.
Rising demand for artificial intelligence (AI) computing is driving a new wave of technological transformation in the semiconductor industry, while energy is emerging as the next bottleneck for the AI sector. Applied Materials group vice president and Taiwan president Eric Yu pointed out that semiconductor technology competition in the AI era can no longer focus solely on performance and transistor counts. How to support more computing with less energy will become the core of the next stage of semiconductor innovation.
As hardware demands for artificial intelligence continue to surge, beyond increased size and complexity in chip packaging, advanced packaging substrates face mounting challenges: larger package footprints, higher layer counts, lower warpage, and higher accuracy. However, relying solely on incremental upgrades to traditional manufacturing processes is hitting a physical wall. This bottleneck is compelling the industry to evaluate next-generation core layer materials.
Nvidia reportedly struck a US$6 billion deal with startup Poolside to license its technology and hire much of its staff. The technology and personnel will reportedly support the chipmaker's Nemotron models in a bid to compete with China's strength in open-source models, but the deal also puts Nvidia in competition with some of its biggest customers.
SK Hynix is moving on several fronts as it pushes deeper into system-level memory competition. From high-bandwidth memory (HBM) and high-bandwidth flash (HBF) to a recently published roadmap proposing to extend co-packaged optics (CPO) to the memory interface, the company increasingly appears to be competing not just over the next generation of HBM, but over how future artificial intelligence (AI) systems organize and access memory.
More coverage
