Reports
AI-generated structured vendor updates
AMD's Experimental Topological Ghost Protocol Boosts MI300X Inference 10x
AMD introduces experimental Topological Ghost Protocol (TGP) on MI300X GPUs, achieving 431 tokens/sec with 100% success in high-concurrency inference, 10x improvement over standard vLLM. TGP uses KV-cache recycling and segmented state management, still experimental but potentially redefining AI inference benchmarks.
Huawei Unveils AI-Centric Network with Token Monetization, UCM Caching Breaks Long-Context Barriers
At MWC Shanghai 2026, Huawei unveiled an AI-native network architecture integrating service, network, and compute, shifting from traffic-centric to intelligence-centric operations. The Unified Cache Manager (UCM) extends KV cache to petabyte-scale external storage, achieving 372% token throughput gains on GLM-5.1 at 128K sequence lengths. Token monetization frameworks and agentic operations enable carriers to charge for AI inference capacity and personalize services.
Huawei and Hubei Mobile Validate AI Inference Acceleration: External KV Cache Boosts Throughput 372%
Huawei and Hubei Mobile completed the first operator AI inference acceleration trial, using OceanStor A800 storage and Ascend A3 supernode with UCM to externalize KV Cache to PB-level storage, achieving up to 372% TPS improvement for long-context inference on GLM-5.1 and MiniMax M2.5 models.
AMD MLPerf 6.0: MI350 GPUs Achieve 3.5x Leap with MXFP4, Debut Multi-Node Training
AMD submitted its most comprehensive MLPerf Training 6.0 results, including first multi-node training (FLUX.1 on 512 GPUs) and MXFP4 training recipe. MI355X delivers 3.5x generational leap over MI300X on Llama 2-70B, within 5% of NVIDIA B200. 10 ecosystem partners validated reproducibility.
AMD Acquires MEXT: AI-Predicted Flash Nears DRAM Performance to Cut AI Memory TCO
AMD acquires MEXT, an AI-driven memory optimization startup. MEXT's predictive technology makes NAND Flash behave like DRAM, expanding effective memory capacity for AI workloads and lowering TCO. The tech will be integrated across AMD's data center portfolio (EPYC, Instinct) to address memory bottlenecks in large models.
AMD Open-Sources AI Software Stack on Vultr, Taking on NVIDIA CUDA Ecosystem
AMD launches a suite of open-source, modular enterprise AI software components on Vultr Marketplace, including AMD Inference Microservices (AIMs), AI Workbench, Resource Manager, and Solution Blueprints. This aims to provide production-grade AI infrastructure without vendor lock-in, directly challenging NVIDIA's CUDA ecosystem.
NVIDIA's Desktop DGX Station with GB300 Shifts Control from Cloud to Local Hardware
ASUS launches ExpertCenter Pro ET900N G3, built on NVIDIA DGX Station GB300 architecture with GB300 Grace Blackwell Ultra chip, 748GB coherent memory, and 20 PFLOPS AI performance. This deskside AI supercomputer enables local LLM fine-tuning, inference, and agentic AI workflows via NVLink-C2C and the full NVIDIA AI software stack including NemoClaw.
NVIDIA GB300 NVL72 Delivers 20x Agentic Coding Efficiency, Setting New Inference Benchmark
NVIDIA's GB300 NVL72 achieves 20x more concurrent coding agents per megawatt than H200 on the new AA-AgentPerf benchmark, leveraging 72-GPU NVLink fabric, MXFP4 kernels, and MoE optimizations. This first standardized agentic inference benchmark redefines data center capacity planning for AI agents.
NVIDIA Optimizes Google's DiffusionGemma for 1,000 tok/s Parallel Text Generation
NVIDIA optimizes Google DeepMind's DiffusionGemma, a diffusion-based text model generating 256 tokens per step in parallel. On a single H100, it achieves 1,000 tok/s, with deployment via NIM and NeMo. This breaks the sequential token bottleneck, slashing serving costs and latency for real-time AI.
NVIDIA Locks Local AI Inference Control with DiffusionGemma Parallel Generation
NVIDIA optimizes Google DeepMind's DiffusionGemma open model, which generates 256 tokens in parallel for 4x speedup over autoregressive models. Achieves 1000 tokens/sec on H100, 150 tokens/sec on DGX Spark, running fully locally with no cloud cost. This reinforces NVIDIA GPU's centrality in compute-bound local AI inference.
NVIDIA Nemotron 3 Ultra: A MoE-Based Control Plane for Cost-Efficient AI Agent Orchestration
NVIDIA launches Nemotron 3 Ultra, a 550B-parameter MoE model (55B active) purpose-built for AI agent orchestration. Featuring Multi-Teacher On-Policy Distillation (MOPD) and a Hybrid Mamba-Transformer architecture, it achieves 5x throughput and 30% cost savings on tasks like SWE-bench, signaling a shift of reasoning control to a layered agent system.
NVIDIA DGX Spark Update: One-Click Local AI Agents, Multi-Node Cluster for 400B Models
At Computex 2026, NVIDIA updates DGX Spark with NemoClaw for one-click local AI agent setup, 2.6x throughput boost for Qwen3.6-35B via vLLM optimizations, and Sync cluster assistant to connect 2-4 nodes over ConnectX-7 200Gbps RoCE, enabling local deployment of large models and multi-agent pipelines.
NVIDIA Cosmos 3: Open-Source Physical AI Model with MoT for Ecosystem Lock-in
NVIDIA releases Cosmos 3, a unified physical AI foundation model with Mixture-of-Transformers architecture combining reasoning, world generation, and action generation. Open-sourced with training scripts and six synthetic datasets, but deployment optimized for NVIDIA NIM and GPUs, signaling an ecosystem lock-in strategy.
NVIDIA RTX Spark: SoC Seizes PC Control, AI Compute Revolution with Ecosystem Lock-in
NVIDIA launches RTX Spark SoC, integrating Blackwell GPU with 20-core Grace CPU (MediaTek co-designed), NVLink-C2C at 600GB/s, up to 128GB unified memory, 1 petaflop FP4 AI, and local 120B-parameter LLM support. This marks a shift from GPU vendor to platform provider, directly challenging Apple M, Qualcomm, and x86 incumbents.
BadHost CVE-2026-48710: Starlette Auth Bypass Exposes AI Agent Infrastructure to HTTP Smuggling
BadHost (CVE-2026-48710) exploits Starlette's inconsistent URL reconstruction via Host header injection, bypassing path-based auth. Affecting 400K+ repos including FastAPI, vLLM, and MCP Server, it exposes AI Agent infrastructure to data theft and potential RCE, forcing a security paradigm shift in HTTP parsing.
AMD Ryzen AI Halo & Max PRO 400: Local 300B Parameter Inference, but Hidden Lock-in and Thermal Limits
AMD launches Ryzen AI Halo developer platform (128GB unified memory, 200B parameter models) and Ryzen AI Max PRO 400 series (first x86 client to run 300B parameter models locally). Unified memory, ROCm optimization, and OEM partnerships aim to shift agentic AI from cloud to local, but shared memory bandwidth and thermal constraints limit real-world throughput.
AMD Showcases Heterogeneous Computing Strategy for Enterprise AI with Dell
At Dell Technologies World, AMD highlighted its heterogeneous computing portfolio, aiming to match the right compute engine to specific enterprise AI workloads, while emphasizing hardware-based security and manageability. This signals a shift in AI infrastructure from generic solutions to fine-tuned, scenario-specific deployments.
AMD Highlights AI PC as Critical Infrastructure for Enterprise Agentic AI in IDC White Paper
AMD released an IDC white paper indicating that over 80% of enterprises are planning, piloting, or deploying AI PCs to support scaled Agentic AI. The report highlights high-performance NPUs and on-device AI processing as critical for enabling real-time, secure workflows, signaling a shift in enterprise AI infrastructure from cloud to endpoint.
NVIDIA Internalizes GPT-5.5 Powered AI Agents at Scale, Defining New Enterprise AI Infrastructure Paradigm
NVIDIA announced that over 10,000 employees have scaled the use of GPT-5.5 via the Codex app, running on NVIDIA GB200 NVL72 infrastructure. This demonstrates the technical feasibility of 'transformative' productivity gains from frontier model inference in enterprise workflows. It also provides a reference architecture for deploying AI agents with auditable, isolated security via dedicated cloud VMs.
Google Cloud Next '26: Agent Gateway Seizes Control Plane, TPU 8i Locks Inference
Google Cloud Next '26 announces 8th-gen TPUs (8t for training, 8i for inference), Agent Platform with Agent Gateway, Agent Identity, Agent-to-Agent Orchestration, Agentic Data Cloud, and Agentic Defense integrating Wiz. The move shifts control from infrastructure to agent orchestration, locking enterprises into a vertically integrated stack.