AWS and NVIDIA to Deploy 2 Million AI GPUs by 2028

Amazon Web Services (AWS) and NVIDIA announced an expanded AI infrastructure partnership that will add 2 million NVIDIA GPUs across AWS data centers in 2027 and 2028. The agreement also covers CPUs, networking, open AI models, data processing, and robotics, signaling a massive new wave of demand for AI-optimized electronics manufacturing.

The 2 million GPU deployment will be staged across 2027 and 2028, adding to AWS’s existing NVIDIA-based AI cloud capacity. Each AI accelerator module demands dense, high-speed interconnects, ultra-low-loss PCB materials, and advanced thermal management. For PCB and PCBA suppliers, this translates into growing orders for high layer count PCBs (often 20+ layers), HDI boards, and substrate-like PCBs that route thousands of signals between GPUs, CPUs, and HBM stacks. The partnership also extends to NVIDIA’s Grace CPUs and networking silicon, increasing complexity of the server boards and backplanes that OEMs and ODMs must source.

Beyond the board-level impact, the deal underscores the tight coupling between AI chip supply and advanced packaging. NVIDIA’s latest GPUs rely on CoWoS and other 2.5D/3D packaging technologies, driving demand for IC substrates and substrate-like PCBs with finer line/space and higher layer counts. AWS’s custom silicon efforts, such as Trainium and Inferentia, further diversify the PCB and substrate mix. Meanwhile, power delivery for AI racks is pushing heavy copper PCB and IMS (insulated metal substrate) designs to handle higher currents, while rigid-flex and ceramic boards appear in high-frequency and high-temperature zones. Supply chain managers should expect extended lead times for high-end laminate and via-fill processes.

For electronics buyers sourcing AI server boards, LT CIRCUIT supplies high-layer-count, HDI, substrate-like, and heavy copper PCBs, along with rigid-flex and ceramic options for high-frequency and thermal zones. Tight signal integrity and advanced via-fill capability align with the requirements of large-scale GPU deployments announced by AWS and NVIDIA.