Tuesday, November 17, 2026 · Week 47 DE · EN · FR · ES Dark
AIData Centers

Ahead of Computex: AMD expands packaging in Taiwan

AMD is expanding packaging partnerships in Taiwan. What the announcement and Nvidia's Vera Rubin ramp-up mean for cloud teams.

By Alec Chizhik May 26, 2026 3 min read
Ahead of Computex: AMD expands packaging in Taiwan

AMD plans to expand its partnerships in Taiwan’s semiconductor ecosystem and support additional advanced packaging capacity. A few days after the announcement, Nvidia CEO Jensen Huang met TSMC CEO C.C. Wei in Taipei. Nvidia subsequently announced the production ramp-up of its Vera Rubin platform. For cloud teams, firm delivery dates, regional availability and the suitability of the software stack remain decisive.

Key takeaways

  • AMD is expanding its collaboration with Taiwanese packaging partners. ASE and SPIL are working with AMD on a wafer-based EFB interconnect technology, while PTI is working on a panel-based variant.
  • Nvidia has announced the ramp-up of Vera Rubin. At the end of May, the company said production shipments were expected to begin in autumn 2026.
  • Investment plans do not translate into specific cloud capacity. Companies should assess offerings by region, delivery date, software compatibility and cost.

AMD expands its packaging ecosystem

On 21 May 2026, AMD announced investments in its Taiwanese ecosystem. The company announcement describes the expansion of strategic partnerships and advanced packaging for future AI infrastructure. It does not specify what share of the total investment is allocated to packaging alone.

According to AMD, it is developing and qualifying a wafer-based 2.5D interconnect technology called Elevated Fanout Bridge, or EFB, with ASE and SPIL. AMD cites upcoming EPYC processors codenamed Venice as an application for the technology. The company also says it has qualified a panel-based EFB interconnect with PTI.

Advanced packaging connects multiple semiconductor components into a single unit. For procurement, the key questions are which products actually use a technology and when it will result in systems available for delivery. The announcement alone confirms neither available GPU capacity nor a specific reduction in pressure on TSMC.

Nvidia ramps up Vera Rubin

Ahead of Computex, Huang met TSMC CEO C.C. Wei for dinner in Taipei on 26 May. Nvidia’s account of the meeting mentions no specific agreements on production capacity.

On 31 May, Nvidia announced the ramp-up of Vera Rubin into full production. According to the announcement at the time, production shipments were expected to begin in autumn 2026. This is a manufacturer timeline and does not in itself confirm that deliveries have taken place or that instances can be booked in a specific European region.

What cloud teams should check when procuring capacity

A second provider can expand procurement options if applications and the operating model are a good fit. Teams should therefore test their inference or training stack on the hardware offered in each case. Areas to check include model support, the frameworks in use, drivers, the runtime environment and operational effort.

AMD describes Helios as a reference design on which partners can build their own systems. AMD says it expects deployments at volume in the second half of 2026. Whether and when a specific offering becomes available depends on the provider and location.

For a reliable comparison, teams need performance figures for their own workloads alongside price, contract duration and the committed provisioning date. For cloud offerings, they also need to establish the region, reservation terms and capacity actually available for use. The manufacturers’ announcements do not support blanket claims about rising prices or guaranteed delivery times.

Frequently Asked Questions

What does AMD’s announcement mean for GPU deliveries?

AMD plans to expand partnerships and packaging capacity. The announcement does not specify binding delivery volumes or dates for individual cloud regions.

Is Vera Rubin available in a cloud region in Germany, Austria or Switzerland?

Nvidia’s 31 May announcement said production shipments were expected to begin in autumn 2026. It does not confirm that capacity can be booked in a specific region in Germany, Austria or Switzerland. Companies need to ask the relevant provider for the current date.

What should teams test before switching GPU platforms?

Their own models and applications are decisive. A test should cover software compatibility, performance, operating costs and available capacity in the intended deployment region.

Image source: AI-generated (June 2026)

Translated from the German original using artificial intelligence. The German version is authoritative.

Also available in

FrançaisEspañolDeutsch
MBF Media Newsletter

The monthly briefing for decision-makers

Once a month, the MBF Media Newsletter gathers what matters from cloudmagazin, MyBusinessFuture, Digital Chiefs and SecurityToday, curated by the editorial team.

Around 23,500 IT and business decision-makers read this newsletter. Read along.

Subscribe for free
MBF Media Newsletter, aktuelle Ausgabe auf dem iPhone
A magazine by Evernine Media GmbH