ERRATUM, added 5 November 2026 The manifest of this run (manifest.json, fields "statement" and "workload.extra.run") was copied from the 8x H200 run captured the same evening and still reads: "VoltageGPU h200-xlarge Confidential VM (8x H200, NVIDIA Protected PCIe mode), proof run of 2026-11-04" "run": "ppcie-8x-h200-2026-11-04" That text is wrong. It cannot be corrected without breaking the evidence: the TDX quote carries SHA-512 of this exact manifest and the NRAS tokens carry SHA-256 of it. The machine of this run is the 8x H100 node (h100-xlarge): - node-state.txt lists eight "NVIDIA H100 80GB HBM3" GPUs with 82,542,720 KiB of protected memory each. The H200 run lists "NVIDIA H200" with 146,115,584 KiB. - Its challenge is c985e0e6b85c349f724e2fa735e0d821f1f55e6b8d26777f910a0363d130ed4e. The H200 run used 5fe630be5d8455ca3bca9f9bee88d66915c91df39f2136186fbcb2530ef6a206. - Its bundle, TDX quote and NRAS tokens are different files from the H200 run's. Nothing else in this folder changed: README.txt, SHA256SUMS and every evidence file are as published on 5 November 2026.