Accelerating bets on self-developed AI chips! AWS raises third-quarter ASIC server shipment target by up to 30%

``` Amazon AWS is accelerating its bet on a self-developed chip strategy. Supply chain sources indicate that AWS has notified relevant suppliers to increase its third-quarter ASIC server shipments by 20% to 30% compared to the original plan, reflecting a highly optimistic outlook for the market prospects of the Trainium 3 chip. On Monday, according to digitimes citing sources in the Taiwan supply chain, server motherboard-level components equipped with Trainium 3 chips began shipping in May and are currently ramping up monthly. AWS's move to increase shipment plans has been interpreted by supply chain players as a sign of pulling in orders ahead of schedule to capture market share, rather than merely a regular adjustment for increased demand. Several supply chain sources point out that Anthropic is one of the key drivers behind AWS’s accelerated procurement—this AI company has stated that even with planning based on 10x growth, its computing power is still insufficient and gaps need to be filled as soon as possible. Boosted by this news, many Taiwanese companies in AWS's Trainium 3 server supply chain are expected to benefit, with room for upward revision in operational performance in the second half of the year. Meanwhile, the industry is also watching whether AWS will launch the next-generation Trainium 4 ahead of schedule due to strong market momentum, further strengthening its competitiveness in the ASIC market. Supply Chain Enters Full-scale Production Ramp-up Phase AWS Trainium 3’s Taiwanese supply chain has formed a relatively complete ecosystem. In the field of heat dissipation, Chicony, Mecotec, and Sun Max International are involved; slide rail manufacturers include Chenbro and Nanjun International; case manufacturers are Chicony and QCT, with Accton as the L6 motherboard-level assembly plant. Regarding shipment pace, motherboard-level L6 components began shipping in May and are ramping up month by month; cabinet-level L11 orders and components such as slide rails will enter mass production in July. Overall, shipments are still ramping up and are expected to scale up in the third quarter. With AWS’s increased procurement, related vendors’ shipments are expected to increase more than previously estimated, with further room for upward revision in operational performance in the second half. Anthropic, OpenAI, and Other Major Clients Have Strong Demand Supply chain sources believe Anthropic is one of the main drivers behind AWS’s accelerated procurement this time. AWS’s parent company Amazon is a key investor in Anthropic, and in April 2026 the two sides raised cooperation again and signed a 10-year agreement for Anthropic to expand AWS computing power procurement. Anthropic has publicly stated that there is still a significant gap in its planned computing power based on 10x growth, which needs to be filled quickly. Besides Anthropic, OpenAI and Uber are also important customers of AWS Trainium. Amazon’s generative AI managed platform Bedrock, which targets enterprise customers, now has 125,000 clients. Its inference workloads mainly rely on Trainium, and continuously expanding enterprise demand also provides strong support. Amazon CEO Andy Jassy said in the recent financial report call that Trainium 2 has sold out, Trainium 3 is almost fully booked, some customers have already pre-ordered the next-gen Trainium 4, and the company has started developing Trainium 5 in parallel. Competitive Landscape Tightens, AWS Needs to Strengthen ASIC Scale Quickly Competitive pressures in the cloud computing market are prompting AWS to act faster. AWS remains the leader in the public cloud market, with Microsoft and Google right behind. In the ASIC chip sector, Google's TPU is the market leader in shipment volume and puts direct pressure on AWS, forcing it to accelerate the strengthening and scaling up of Trainium. In addition, Meta recently announced plans to open its proprietary AI computing power for external use, allowing businesses and developers to pay to use its AI models or rent bare-metal GPU computing power, openly competing with AWS, Microsoft, and Google for enterprise customers. This intensifies market competition and increases AWS’s sense of urgency to expand. DIGITIMES Research analysis predicts that in 2026, high-end AI server chips will see GPU server shipments grow annually by about 43.8%, while ASIC server shipment growth will be even more significant, reaching 64.2%. With both AWS Trainium and Google TPU servers entering the scale-up phase in the second half of 2026, the overall ASIC server market is expected to accelerate its expansion. Risk Warning and Disclaimer The market carries risks, and investments should be made cautiously. This article does not constitute personal investment advice and does not take into account individual users' specific investment objectives, financial situation, or needs. Users should consider whether any opinions, viewpoints, or conclusions in this article are suitable to their particular situation. You bear sole responsibility for any investment made on this basis. ```