Open App →
Back to News Feed
Digitimes October 28, 2026 By Yunnie Cheng positive

Skymizer partners with Egis to develop 12nm and 6nm AI accelerators

TSMCAI / LLMSemiconductorsNVIDIA / GPUMemoryManufacturingData Center
Taiwanese startup Skymizer has formally announced a strategic partnership with Egis Technology to develop Flow (T12), a new-generation AI inference chip. Fabricated on TSMC's 12nm node and paired with LPDDR5 memory, the solution requires just three PCIe cards and 750W of power to run 3-trillion-parameter large language models (LLMs). This represents roughly one-tenth the power consumption of a comparable Nvidia H100 deployment.
Read original article ↗

Related Articles

TSMC reportedly weighs 20% CoWoS price increase as Nvidia Blackwell and Rubin chips put pressure on packaging capacity — equipment lead times stretch to eight months

TSMC is reportedly weighing a price increase of up to 20% for its CoWoS advanced packaging technology as Nvidia’s Blackw

Tom's Hardware · October 27, 2026

Anthropic's IPO filing puts TSMC, HBM, direct leasing at the center of AI compute risk

In its preliminary IPO registration statement filed with the US Securities and Exchange Commission, Anthropic describes

Digitimes · October 27, 2026

Inside the Package Powering Modern AI

TSMC’s CoWoS-S is critical to the artificial intelligence era because it solves the central systems problem facing moder

SemiWiki · October 26, 2026

Podcast highlights: TSMC's 3nm demand story, Samsung HBM shift, and A14 ramp

Following TSMC's third-quarter 2026 earnings call, DIGITIMES analyst Luke Lin provided in a podcast additional analysis

Digitimes · October 26, 2026