Amazon Web Services has instructed partners across its AI server supply chain to increase planned third-quarter shipments by 20% to 30% compared with previous schedules, according to sources within the supply chain. The directive reflects stronger-than-expected demand for AWS's Trainium 3 chip, which entered shipping in May 2026 and has seen production volumes expand month to month.
A range of Taiwan-based suppliers are preparing for the higher volumes. Companies named by industry sources include Asia Vital Components, Microloops, Cooler Master, King Slide, Repon, Chenbro Micom, Accton, and Wiwynn. Component manufacturers supplying motherboard assemblies reported they began shipping in May and have been increasing monthly volumes. In parallel, orders for racks and slide rail components are scheduled to move into mass production in July.
At a recent earnings call, AWS CEO Andy Jassy said Trainium 2 had sold out and that Trainium 3 is close to being fully reserved. Those remarks align with the supply-chain signals of accelerated shipment plans and supplier preparations for rising demand.
Industry sources indicate Anthropic is a principal factor behind the accelerated order flow. In April 2026 AWS expanded its commitment to Anthropic through a 10-year partnership under which Anthropic will increase purchases of AWS compute capacity. Other prominent customers for Trainium capacity include OpenAI and Uber, while Amazon's Bedrock platform is reported to serve 125,000 customers whose inference workloads are handled primarily by Trainium chips.
Forecasts from DigiTimes Intelligence analyst Jim Hsiao anticipate continued growth across high-end AI server chip shipments in 2026. His projections call for GPU server shipments to rise 43.8% year-over-year and for ASIC server shipments to increase 64.2%.
The combination of large customer commitments, ramping production, and scheduled mass production milestones for rack components underscores a rapid scaling phase for both AWS and its hardware partners. Suppliers in Taiwan that provide chassis, motherboards, and other server subsystems are positioned to see materially higher volumes in the near term as Trainium 3 adoption expands.
Contextual note: The details above are based on reporting from participants and observers within the AI server supply chain and on published comments regarding AWS's Trainium program and customer commitments.