
Microsoft's Maia 300 chips offer 40% more tokens per megawatt, set to power AI by 2027
The tech giant unveiled its third-generation custom silicon, aiming to power its AI services with gigawatt-scale deployments by 2027.
Everything Threat Times has reported about TSMC: 8 stories, newest first. Part of our Tech News coverage.

The tech giant unveiled its third-generation custom silicon, aiming to power its AI services with gigawatt-scale deployments by 2027.

Microsoft unveils its third-generation custom AI accelerator, targeting greater efficiency and reduced reliance on external GPU vendors for cloud inference.

The new EPIC Center aims to cut semiconductor development cycles by 30% through collaborative work with major industry partners.

The new chip uses a dataflow architecture to process 40% more tokens per megawatt than leading GPU-based systems.

The third-generation processor, built on TSMC’s N2P node, will serve as Microsoft’s primary system for artificial intelligence inference tasks.

TSMC officials in charge of overseas operations have started more serious discussions with Texas authorities in the first half of 2026, following over ten years of intermittent talks.

A five-year manufacturing deal will see GlobalFoundries produce silicon interposers for TSMC’s advanced packaging, enabling US-branded AI chip production.

GlobalFoundries agreed to a $2 billion, five-year agreement to supply silicon interposer capacity to TSMC from its Malta, New York facility.