MediaTek has completed pricing on US$3.9 billion of overseas convertible bonds, more than NT$120 billion and what the company says is the largest overseas CB ever issued in Taiwan's capital market. Nvidia subscribed for US$3.5 billion of it — roughly 90% of the offering.

Marvell said on its fiscal second-quarter of 2027 earnings call that customers are actively planning to deploy scale-up optical solutions as early as 2027, with demand for near-package optics (NPO) now revised higher. The company said copper cabling is increasingly constrained by distance and bandwidth limits as computing systems grow larger, pushing customers to raise the share of optical transmission in their designs.

Nvidia reported second-quarter revenue for fiscal year 2027 of US$96.22 billion, a 106% year-over-year increase, driven by a 117% surge in data center revenue to US$89 billion. With third-quarter revenue of fiscal year 2027 projected to hit US$108 billion and full-year fiscal year 2028 revenue expected to grow around 70%, Nvidia revealed that current supply capabilities meet only about 70% of customer demand. This indicates that the primary constraint on Nvidia's sustained hyper-growth has officially shifted to the supply side.

Chip design company Agentrys announced last week that it raised US$24.5 million in funding from backers including MediaTek. It is one among an emerging crop of startups using agentic tools to power the design of semiconductors, a step in the chip process that still involves considerable amounts of time and labor.

MetaX posted its first half-year profit since listing, with rising GPU shipments and the mass production of its C600 accelerator strengthening its commercial position while China's domestic AI chip sector increasingly splits between profitable vendors and peers still absorbing heavy R&D costs.

The AI infrastructure race has spent the past few years fixated on high-bandwidth memory as one of the most indispensable building blocks of accelerated compute. Yet as model sizes, context windows and inference workloads continue to expand, the bottleneck is increasingly shifting from how much memory can be stacked toward how efficiently data can move between memory and compute.

The four major US cloud service providers (CSPs) have already rewritten their 2026 capital expenditure records, and ASIC vendors are increasingly confident that demand will remain strong through 2028 as customer spending and supply-chain visibility both extend farther out.

Apple on August 25 unveiled the M6 and M5 Ultra chips without holding a launch event, with the M6 becoming the company's first processor built on TSMC's 2nm process and debuting in the new Mac mini. But industry watchers say the M6's short-term significance lies mainly in showcasing 2nm design technology, not in delivering immediate commercial gains.

In an effort to accelerate its custom silicon initiative and curb reliance on external hardware suppliers, artificial intelligence developer Anthropic held discussions to acquire AI chip startup MatX for roughly US$7 billion, according to Reuters. While the acquisition talks have since evolved into potential partnership discussions, the move underscores Anthropic's ambition to expand its proprietary hardware capabilities.

Huawei's Ascend 910C is already sold out in China, but domestic chips still cannot fully replace Nvidia's H200s. The shortage matters beyond China because AI compute constraints, pricing, and supply chain shifts are reshaping how quickly companies worldwide can deploy advanced inference systems.

According to Bloomberg, Nvidia has formed a federal political action committee (PAC), expanding its Washington presence as governments debate artificial intelligence rules that could affect businesses, workers, and consumers worldwide. The move underscores how major technology firms are preparing for a longer policy battle over AI, competition, and market access.

Nvidia's latest earnings and forecast suggest AI demand remains far ahead of available supply, with implications for memory, power, cooling, and networking worldwide. The chipmaker's guidance points to continued price pressure and a broader buildout of infrastructure that will shape AI adoption across markets and industries.

Samsung Electronics' packaging technology lead, Lee Hee-seok, outlined the company's advanced semiconductor packaging roadmap at Advanced Semiconductor Packaging & Chiplet Show (ASPS) 2026. Beyond sharing target production timelines and specifications for fan-out panel-level packaging (FOPLP) and glass substrates, Lee emphasized AI's pivotal role in driving packaging R&D forward.

Surging AI chip demand continues to outstrip TSMC's advanced packaging capacity. At the same time, massive AI server shipments are driving up demand for CPUs, networking equipment, power management ICs (PMICs), and peripheral silicon, rapidly soaking up traditional packaging capacity. IC design insiders reveal that several outsourced semiconductor assembly and test (OSAT) providers have already warned clients that traditional packaging capacity could face a deficit exceeding 20% by 2027. New capacity cannot be added quickly enough, with the supply of wire bonding equipment emerging as the primary bottleneck.

Renesas Electronics has opened a physical AI and robotics lab in Beijing, a move that could speed the development of safer, smarter machines for factories, logistics, and homes worldwide. The facility is designed to help customers test, validate, and co-develop robotic systems as physical AI shifts from concept to real-world deployment.

Marvell Technology used its second-quarter fiscal 2027 earnings call on August 27 to lift its revenue outlook for a second consecutive quarter, and the composition of the raise says more about the AI infrastructure cycle than the headline number does. Management now expects fiscal 2027 revenue of roughly US$12 billion, up from approximately US$11.5 billion a quarter ago, and fiscal 2028 revenue of approximately US$18 billion, up US$1.5 billion from the US$16.5 billion guided three months earlier. Growth is accelerating off a larger base: fiscal 2028 is now guided at about 50% growth, up from roughly 45% previously.

Marvell Technology's second quarter of fiscal 2027 was the point at which its data center business stopped being the growth engine attached to a diversified infrastructure chipmaker and became substantially the whole company. Revenue for the three months ended August 1, 2026, reached a record US$2.74 billion, up 37% from a year earlier, with the data center end market contributing US$2.17 billion, or 79% of the total, against 74% a year ago. Everything else - the carrier, enterprise networking, consumer, and automotive businesses grouped as communications and other - grew 10% and shrank as a share of the company.

Taiwan is charting a new direction for its biomedical industry, drawing on its strengths in semiconductors, AI hardware, precision manufacturing and healthcare. AI-powered drug discovery, smart medical devices and biomedical chips are emerging as priority areas.

China's CMOS image sensor industry is gaining ground beyond smartphones, with robotics emerging as one of the latest areas where domestic suppliers are linking sensor technology with higher-level perception systems.

Efforts to curb the illicit trade of advanced AI hardware to China have intensified following a series of indictments and probes targeting the supply chain.

Rebellions is pursuing deployment of more than 100 NPU-based computing racks in the UK through a partnership with British AI infrastructure startup Callosum, giving the South Korean chipmaker a potential entry point into Europe's emerging sovereign computing market.

Nvidia has made a point of it in three consecutive earnings decks: year-on-year growth has accelerated for four straight quarters, which the company calls unprecedented at its scale. It's true. But split the data center business into its two halves, and the acceleration has a single source, and it isn't the customers everyone assumes.

The five largest cloud operators spent US$181.5 billion on property and equipment in the June quarter, 87% more than a year earlier. Nvidia's hyperscale revenue was US$48.7 billion — 26.8 cents of every dollar they spent, down from a peak of 33.2 cents three quarters ago.

Nvidia said in its fiscal 2027 second-quarter results on the 27th that it sold a small number of H200 artificial intelligence (AI) processors to Chinese customers last quarter, marking its first renewed AI chip sales to China since shipping about US$60 million of H20 chips in early 2025. But shipments fell short of the total approved by US President Donald Trump and licenses cleared by the US government in January, amid opposition from Beijing.