TSMC Develops EMIB-Like AI Chip Packaging
On July 30, TSMC was reported to be developing advanced chip packaging comparable to Intel's EMIB, potentially expanding options for AI-chip integration. The Information, cited by Seeking Alpha, reported that the work concerns advanced AI-chip packaging, while Nvidia and AMD currently use TSMC's CoWoS technology.
TSMC is developing an advanced chip-packaging technology comparable to Intel's EMIB, according to The Information. Seeking Alpha, citing The Information on July 30, reported that the work concerns advanced AI-chip packaging.
TSMC's CoWoS, short for Chip-on-Wafer-on-Substrate, is used to integrate large compute dies with high-bandwidth memory and other components in AI accelerators. Seeking Alpha reports that Nvidia and AMD currently use CoWoS for advanced AI chips.
Intel's Embedded Multi-Die Interconnect Bridge, or EMIB, is a packaging approach that connects chiplets through localized silicon bridges rather than a full silicon interposer. The Information's report frames TSMC's development work as a response to Intel's packaging technology.
What remains unreported
The available reports do not provide technical specifications, production timing, named customers, pricing, or a stated manufacturing capacity for the reported TSMC technology. They also do not establish whether the packaging method will be compatible with existing CoWoS designs or represent a separate product line.
For AI hardware teams, advanced packaging is a material systems constraint rather than a back-end implementation detail: accelerator performance and deployable memory capacity depend on interconnect density, power delivery, thermals, and high-bandwidth-memory integration. Across the semiconductor sector, alternative packaging routes can broaden design options for multi-die accelerators, although any practical effect depends on qualification, yield, capacity, and customer adoption.
Key Points
- 1The Information reports that TSMC is developing an EMIB-like packaging technology, extending competition beyond leading-edge wafer fabrication into advanced integration.
- 2Seeking Alpha reports Nvidia and AMD currently use CoWoS, making advanced packaging capacity and technical alternatives relevant to accelerator supply chains.
- 3Across AI hardware, packaging choices affect memory integration, interconnects, thermals, yield, and ultimately the deployability of large multi-die accelerators.
Scoring Rationale
Advanced packaging is a critical constraint for AI accelerators that combine large compute dies with high-bandwidth memory. The report concerns development work rather than an announced product, customer deployment, or capacity commitment, which limits its immediate practitioner impact.
Sources
Public references used for this report.
Practice with real Ad Tech data
90 SQL & Python problems · 15 industry datasets
250 free problems · No credit card
See all Ad Tech problems
