Taiwanese prosecutors on Monday indicted a former deputy manager at Taiwan Semiconductor Manufacturing Company (TSMC) for allegedly stealing trade secrets involving technologies designated as national core technologies, with the intent to use them in China.
The defendant, identified by the surname Chen, faces charges under Taiwan’s National Security Act and the Trade Secrets Act. Prosecutors are seeking a seven-year prison sentence.
According to the indictment, Chen conspired with a Hong Kong national between 2023 and 2024 to establish CSMAC, a semiconductor materials analysis company in China. He is accused of copying 21 TSMC documents containing national core technologies and other proprietary trade secrets for use in the new venture.
TSMC, the world’s largest contract chipmaker and a key supplier to companies such as Nvidia and Apple, detected the unauthorized copying through an internal investigation and successfully recovered all duplicated documents. The Hong Kong co-conspirator was not prosecuted because he had died. Chen remains in custody, and Reuters was unable to reach him for comment. TSMC declined to comment on the case.
Prosecutors described this as Taiwan’s first indictment involving the alleged theft of trade secrets related to technologies officially designated as national core technologies for intended use in China.
The case underscores heightened concerns in Taiwan over China’s efforts to recruit Taiwanese semiconductor talent and acquire advanced chip know-how. In response, Taiwan has significantly tightened protections for technologies deemed critical to its economic and national security.
MacDailyNews Take: TSMC clearly plays a central role in the global semiconductor supply chain, making the protection of its intellectual property a matter of strategic importance far beyond Taiwan’s borders.
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
Apple is quietly shopping for AI chip companies. According to a new report from The Information, the company has been approaching chip startups and consulting with bankers about potential acquisitions — a notable shift for a firm long known for its reluctance to make big hardware buys.
The move comes as Apple grapples with limitations in its current AI infrastructure. Its Private Cloud Compute servers, which power parts of Apple Intelligence, currently rely on internally designed M2 Ultra chips. These have proven insufficient for handling larger, more demanding AI models.
The Siri Overhaul Problem
One concrete example stands out: Apple has been working on a major Siri overhaul powered in part by Google’s Gemini models through a partnership. When engineers tried running these advanced models on Apple’s own servers, the M-series hardware fell short. As a result, more complex Siri tasks are now routing through Nvidia GPUs hosted in Google’s cloud infrastructure.
This dependency clashes with Apple’s long-standing preference for controlling its own silicon and protecting user privacy through on-device and Private Cloud Compute processing.
Baltra Delayed
Internally, Apple has been developing its first dedicated AI server chip, codenamed Baltra. The company had originally targeted a 2026 launch, but the timeline has slipped. Baltra is designed as a purpose-built server ASIC (with Broadcom contributing key networking and interconnect technology) to reduce reliance on third-party GPUs for cloud AI workloads.
The reported pursuit of acquisitions suggests Apple wants to accelerate progress on its next-generation server silicon — either by bringing in specialized talent and IP or by acquiring teams that have already solved similar challenges.
A Rare Shift in M&A Approach
Apple has historically avoided large acquisitions, preferring to build technology in-house. Its most recent notable deal was the January 2026 acquisition of Israeli audio AI startup Q.ai (reportedly valued around $2 billion). With roughly $45.6 billion in cash and cash equivalents as of late March 2026, Apple certainly has the financial firepower if it chooses to move.
This new interest in chip startups marks a pragmatic acknowledgment that the AI hardware race — particularly in data center-scale inference and training — moves extremely fast. While Apple excels at designing efficient mobile and laptop chips, scaling specialized AI accelerators for servers is a different challenge that hyperscalers like Google, Amazon, and Meta have been tackling for years.
Broader Context
This development comes just days after Apple announced a major multi-year expansion of its partnership with Broadcom, worth more than $30 billion. The deal includes custom silicon components and supports U.S. manufacturing expansion in Colorado. It underscores Apple’s broader push to strengthen its custom silicon capabilities while also highlighting how much it still relies on external partners for certain advanced technologies.
What It Means
If Apple follows through on acquisitions, it could help close the gap in its cloud AI infrastructure faster than organic development alone would allow. Success here would support more ambitious on-device and private-cloud AI features without heavy dependence on external clouds and GPUs.
It would also represent a meaningful evolution in Apple’s strategy: from a company that largely designs its own chips from scratch to one willing to acquire specialized expertise to stay competitive in the AI era.
The coming months will reveal whether these discussions lead to actual deals — and whether they can meaningfully accelerate Apple’s ability to run sophisticated AI models on its own terms.
MacDailyNews Take: The mad scramble to catch up continues. Hopefully, new Apple CEO John Ternus can quickly kick things back into high gear after far too many years of coasting along in neutral.
Please help support MacDailyNews — and enjoy subscriber-only articles, comments, chat, and more — by subscribing to our Substack: macdailynews.substack.com. Thank you!
OpenAI has replaced its Mac apps with a confusing 1.45 GB amalgamation, killed its ChatGPT Atlas Web browser after less than a year, and boosted ChatGPT Voice by backing it with more capable models.
Audiofile Engineering’s Fidelia plays your music collection with bit-perfect accuracy and lets you shape the sound through Audio Units, equalization, and headphone virtualization. It’s not for everyone—but audiophiles will appreciate its depth.
Investigadores de la Universidad del Sur de California analizaron las señales neuronales de 24 pacientes con epilepsia y predijeron si recordaban animales, edificios, vehículos u otros objetos. Todavía no pueden reconstruir pensamientos completos, pero el avance demuestra que los recuerdos dejan patrones suficientemente claros como para ser interpretados por una máquina.
A aplicação foi criada pela startup EdTech, criada por quatro alunos do Técnico e já reconhecida em vários prémios. Está disponível para experimentar gratuitamente.
Los datos de DESI y una controvertida revisión de supernovas sugieren que la energía oscura podría estar debilitándose con el tiempo. La hipótesis aún divide a los astrónomos, pero ha devuelto al debate el Big Crunch, un final inverso al Big Bang que parecía descartado.
Investigadores identificaron una especie extinta de hormiga dentro de una pieza de ámbar que perteneció a Johann Wolfgang von Goethe. Gracias a técnicas de imagen avanzadas, el fósil pudo estudiarse por dentro y reconstruirse en 3D por primera vez.
Login
