Nvidia and Amkor are partnering to increase superior chip packaging and testing capability for next-generation AI infrastructure. Picture generated by way of ChatGPT
Nvidia is placing $1.5 billion behind one of many much less seen however more and more necessary levels of AI chip manufacturing. The corporate has signed a multiyear settlement with Amkor Expertise to increase superior semiconductor packaging and testing capability in the US.
The partnership will assist Amkor’s Arizona growth and fund new packaging and testing applied sciences for Nvidia’s future AI and accelerated computing platforms.
For companies investing in AI infrastructure, the deal may assist tackle a rising packaging bottleneck, add extra US capability to Nvidia’s provide chain, and enhance the corporate’s means to carry more and more complicated processors, reminiscence, and networking elements into completed techniques at scale.
Nvidia’s prepayment helps Amkor increase US capability
Amkor stated Nvidia would offer a prepayment beneath the settlement, permitting the packaging firm to increase its US capability whereas the 2 companies align their know-how roadmaps. The businesses will concentrate on high-density interconnects and heterogeneous integration, which mixes several types of chips and elements inside one package deal.
Amkor already supplies packaging providers for Nvidia merchandise, together with knowledge heart processors, networking chipsets, and accelerated computing techniques. The expanded partnership is meant to carry future packaging and check applied sciences into large-scale manufacturing as demand for AI infrastructure grows.
Reuters reported that Amkor shares rose 17% in prolonged buying and selling following the announcement. The publication additionally famous that Amkor signed a 10-year US packaging partnership with Taiwan Semiconductor Manufacturing Firm in June and supplies packaging providers for AMD.
Why chip packaging is changing into an AI bottleneck
Packaging is the stage the place particular person chips are assembled, linked, protected, and ready to be used in a bigger computing system. For contemporary AI accelerators, the method should assist high-speed communication amongst processors, reminiscence, and networking elements whereas serving to techniques handle energy consumption and warmth.
The Enterprise Occasions described superior packaging as a rising provide bottleneck as chipmakers attempt to fulfill demand for AI accelerators. Nvidia’s funding may assist safe extra capability for its platforms whereas giving Amkor funding to construct amenities earlier than buyer demand totally materializes.
Superior packaging has additionally change into extra necessary as chipmakers mix specialised processors, reminiscence, and different elements as a substitute of counting on a single piece of silicon. Any scarcity in packaging and testing capability can delay completed techniques even when the person chips can be found.
What the Amkor growth may imply for AI {hardware} provide
Enterprise patrons shouldn’t anticipate the settlement to supply a right away improve in accessible AI servers. Constructing packaging capability and growing new manufacturing processes can take time, and the businesses didn’t present an in depth manufacturing schedule.
The settlement additionally doesn’t assure that each deliberate facility or manufacturing line will arrive on schedule. Amkor warned that its Arizona campus may face adjustments involving timing, prices, specs, or anticipated industrial advantages.
The longer-term impact could possibly be a extra geographically numerous provide chain. Amkor stated its Arizona growth would complement its current manufacturing presence throughout Asia moderately than exchange it, giving Nvidia entry to extra US capability whereas retaining established manufacturing networks abroad.
For infrastructure groups, the primary improvement to observe is whether or not the funding reduces provide constraints for future Nvidia techniques. Extra packaging capability may ultimately enhance availability, however pricing and supply schedules will nonetheless rely on chip manufacturing, reminiscence, networking, server meeting, and knowledge heart demand.
Learn extra about Nvidia and Microsoft’s RTX Spark Home windows PCs for operating AI brokers regionally, with techniques from main PC makers anticipated this fall.
