Lenovo unveiled the ThinkCentre X Ultra at IFA 2026, a compact 1.6-liter desktop computer designed for high-performance computing and AI tasks. It will be available in November 2026, starting at $3,699, and can be configured with up to 128GB of unified memory. The system supports clustering up to four units together, allowing users to combine their memory capacities for a total of 512GB, which could be useful for running complex AI models locally without relying on cloud resources. The ThinkCentre X Ultra is powered by AMD's Ryzen AI Max+ PRO 495 APU, part of AMD's Gorgon Halo refresh. This chip combines 16 Zen 5 CPU cores, a 40-compute-unit Radeon 8065S integrated GPU, and an NPU (neural processing unit) capable of about 55 TOPS (trillion operations per second). A key feature is its shared memory architecture, where the CPU and GPU access the same LPDDR5X memory, allowing the system to handle larger AI models than traditional setups with discrete GPUs limited to 24GB or 32GB of VRAM. Lenovo has equipped the ThinkCentre X Ultra with several enterprise-focused features, including ThinkShield security, AMD's PRO technologies, DASH out-of-band manageability, and a self-healing BIOS. It also integrates the AMD Ryzen AI Developer Center, providing users with preconfigured tools and workflows for both Windows and Linux environments. These features are aimed at businesses and developers looking for secure, manageable, and flexible computing solutions. Alongside the ThinkCentre X Ultra, Lenovo also introduced the Yoga laptops, which also support up to 128GB of unified memory. However, unlike the ThinkCentre, the Yoga laptops cannot be clustered together for additional memory or used as part of a localized AI server setup. The high cost of the ThinkCentre X Ultra could be a barrier for users wanting to build a 512GB memory cluster, which would require nearly $15,000 for four units, not including connectivity costs. While the ThinkCentre X Ultra is equipped with dual Thunderbolt 4 ports capable of 40Gb/s each, Lenovo has not yet provided details on how these ports will handle data transfer when multiple units are connected. The performance benefits of the larger memory pool are also unclear, as the integrated GPU, though capable, is significantly slower than high-end discrete GPUs like the Nvidia H100, which can be rented for a fraction of the cost. Whether the ThinkCentre X Ultra will remain competitive as more Ryzen AI PRO 400 APUs become available is still uncertain.