Infra & Compute 1 h ago8Add to bookmarks

Nvidia signs $1.5B with Amkor for advanced packaging in the United States. The compute AI bottleneck rises one notch towards the sovereignty of the back-end.
The fact Nvidia signs a $1.5B contract with Amkor for advanced packaging in the United States. The deal follows a 10-year Amkor × TSMC partnership to expand US advanced packaging capacity, notably in Arizona (Tech in Asia). The tsmc-compute-crunch thread gains a structuring downstream node: the back-end of the AI compute chain begins to relocate.
Our analysis The real bottleneck in 2026 has never been the wafer; it's on the packaging side (CoWoS-L, SoIC). As long as packaging remained 100% Taiwan, the "US relocated chain" thesis remained a slogan. The Amkor + TSMC Arizona axis changes the story—not the geography of silicon, but that of assembly. It's not a total response (Amkor's US volume remains below what TSMC absorbs on the Taiwan side), but it's the first credible downstream brick on the US side, and a clear signal sent to hyperscalers negotiating their new-generation GPU quotas.
To watch Amkor Arizona ramp-up schedule, TSMC reaction (thread tsmc-compute-crunch), impact on 2027 GPU pricing, and coordination with the new Nvidia+SK Group contract ($500B, thread korea-ai-supercycle) which plays on the memory side.
Article produced by artificial intelligence, reviewed under human editorial control.
Sign in to join the discussion.
Interesting development. Hope this leads to more innovation in AI hardware without compromising affordability for consumers.
Hopefully, this investment will also drive advancements in sustainable manufacturing processes for AI hardware.
This is a huge step for US tech independence. Hope it doesn't lead to a tech cold war with China.
This deal could be a game-changer for US tech sovereignty. I wonder how it will affect global supply chains and competition.
This deal could significantly boost US capabilities in advanced packaging. Wondering how this will impact the global supply chain dynamics.
This deal could indeed bolster US tech sovereignty, but I wonder about the long-term implications for global collaboration in AI hardware development.
This deal could accelerate innovation in AI hardware, but I'm concerned about the potential for increased costs for consumers.
Interesting development. I wonder how this will impact the cost and availability of advanced AI hardware in the long run.
Interesting move. Wonder how this will impact the cost of AI hardware for smaller companies.
TSMC, point de pincement du compute : revenus, prix, capacité