Targon + Intel whitepaper
Targon Compute co‑authored a whitepaper with Intel engineers arguing confidential compute on untrusted Intel/NVIDIA hardware is a viable, cheaper alternative to traditional cloud infra — the pitch doubles down on decentralized, lower‑cost stacks. That framing challenges cloud lock‑in and highlights new procurement vectors. (x.com)
Intel’s community blog published “Decentralized Compute on Untrusted Hardware Using Intel® TDX and Encrypted CVMs” on March 17, 2026 and lists Manifold Labs authors Venish Patidar, Dhruv Bindra, Ahmed Darwich and Josh Brown alongside Intel engineers Haidong Xia and Sathi Nair. (community.intel.com) The paper specifies an architecture built on Intel Trust Domain Extensions (TDX), Intel Trust Authority (ITA) and NVIDIA Confidential Computing, with Confidential Virtual Machines delivered as uniquely encrypted Ubuntu 24.04 images. (community.intel.com) Intel’s post explicitly calls out existing GPU cluster examples — “8x H100s, 8x H200s, 8x B200s” — as financial barriers that the proposed design aims to address for smaller organizations. (community.intel.com) Manifold’s Targon Virtual Machine (TVM) documentation shows integration with NVIDIA’s nvTrust SDK for GPU attestation and details host‑verification checks including TPM 2.0, Secure Boot, BIOS configuration and kernel‑level protections. (manifold.inc) Targon’s public site advertises an inventory of “1000+ GPUs & CPUs” and states support for Intel TDX and AMD SEV to enable encrypted, hardware‑backed confidential VMs. (targon.com) Targon announced acceptance into NVIDIA’s Inception program on March 18, 2026, which the project said would provide technical training, developer tools and preferred pricing from NVIDIA and its ecosystem partners. (tao.media) The technical materials describe four runtime roles — decentralized hardware providers, decentralized validators, users submitting confidential workloads, and the Targon coordinator — and document continuous re‑attestation behavior (example cadence: every 72 minutes) for ongoing verification. (community.intel.com)
