MediaTek says its next generation program will use only Intel’s EMIB T packaging technology, a move that gives Intel a meaningful win as it tries to rebuild its position in advanced chip manufacturing. The statement was made during Goldman Sachs Taiwan Day, where MediaTek said tape out is targeted for the fourth quarter of 2026, with mass production planned for the fourth quarter of 2027.
The announcement matters because advanced packaging has become one of the most important parts of the AI chip race. Modern AI processors are not just single chips. They often combine compute dies, memory, and other parts inside one package. How those pieces are connected can affect performance, cost, yield, and how quickly companies can scale production.
Intel’s EMIB T is being positioned as an alternative to TSMC’s CoWoS packaging, which is widely used for high end AI hardware. CoWoS relies on a large silicon interposer to connect components inside a package. EMIB instead uses smaller silicon bridges to connect key parts. The pitch is that this can reduce complexity and cost while still offering the kind of high density connections needed for advanced AI chips.
| Packaging technology | Basic idea | Why it matters |
|---|---|---|
| Intel EMIB T | Uses silicon bridges between components | Could reduce cost and complexity |
| TSMC CoWoS | Uses a large silicon interposer | Strong bandwidth and widely used in AI accelerators |
| FCBGA | Traditional packaging benchmark | Used as a yield comparison point |
| HBM packaging | Connects compute chips with high bandwidth memory | Essential for advanced AI systems |
MediaTek’s statement also matters because the company is reportedly tied to Google’s next generation TPU custom AI chips. If those programs lean on EMIB T, Intel could gain a stronger role in the AI supply chain at a time when TSMC’s CoWoS capacity remains heavily demanded by Nvidia, AMD, and other AI hardware companies.
That does not mean EMIB T has already won the market. Advanced packaging decisions depend heavily on yield, cost, performance, and supply reliability. Analyst Ming Chi Kuo has reportedly said EMIB T’s wider adoption will depend on whether Intel can hit strong yield targets. A higher yield means more usable chips from each production batch, which helps reduce the cost of each finished processor.

The yield target being discussed is important. Intel is said to be aiming for a level that would make EMIB T competitive with mature packaging methods such as flip chip ball grid array. If Intel can reach that level, it would make EMIB T more attractive to companies building custom AI chips where cost matters as much as performance.
The timing also gives the announcement more weight. Tape out in late 2026 means the design would be finalized and sent into manufacturing preparation. Mass production in late 2027 points to a long term project rather than a short term experiment. That suggests MediaTek is making a serious platform decision, not simply testing Intel’s technology on the side.
For Intel, this is exactly the kind of customer signal it needs. The company has been trying to prove that its foundry and packaging business can compete for advanced AI work. Winning trust from a major fabless chip company like MediaTek would support that effort, especially if the program is connected to large scale AI infrastructure.
There are still open questions. MediaTek did not clearly identify which chips the next generation program covers. It could involve custom AI silicon, CPUs, or other high performance processors. The final importance of the move will depend on the size of the project, whether Google’s TPU roadmap is involved, and how well Intel’s packaging performs at scale.
The broader picture is that AI chip competition is moving beyond raw compute. Packaging, memory integration, yield, and supply chain capacity are now just as important. TSMC’s CoWoS remains a major force, but demand is so high that companies are looking for alternatives that can lower cost or ease supply pressure.
MediaTek’s EMIB T commitment gives Intel a chance to prove that it can be one of those alternatives. If the technology delivers strong yields and competitive costs by 2027, Intel could become a more important player in AI packaging. If it struggles, the market may continue to favor existing TSMC based routes.
For now, the announcement is a strategic win for Intel. MediaTek’s decision gives EMIB T credibility, puts Intel deeper into the AI chip packaging conversation, and shows that major companies are willing to look beyond CoWoS as the AI hardware race grows more expensive and supply constrained.



Discussion (0)
Be the first to comment.