Reports
AI-generated structured vendor updates
AMD Launches Next-Gen HPC/AI Supercomputing Solution
AMD introduces a supercomputing solution based on new compute architecture, integrating CPU and GPU acceleration technologies optimized for HPC and AI workloads. The solution improves energy efficiency and compute density, supporting exascale and hyperscale computing systems.
AMD Partners with TCS to Deploy Helios AI Rack Architecture in India
AMD partners with Tata Consultancy Services to introduce the Helios rack-scale AI architecture in India, built on Instinct MI300 accelerators for large-scale AI training and inference workloads. The solution is delivered as complete racks, scalable to thousands of nodes, optimized for generative AI and HPC. The collaboration leverages TCS's integration services in cloud, AI, and cybersecurity for end-to-end AI solutions.
AMD Launches CDNA 4-based MI430X Accelerator for AI Compute
AMD launches Instinct MI430X accelerator with CDNA 4 architecture, featuring enhanced matrix cores and FP8 precision support optimized for LLM training and inference. Utilizes HBM3e memory and Infinity Fabric interconnect for improved AI workload performance and efficiency.
Arm and Tensor Collaborate on AI-Defined Automotive Compute Architecture
Arm and Tensor form a multi-year strategic partnership to provide an Arm-based compute foundation for embodied AI robocars. The architecture integrates over 400 security-certified Arm cores with specialized domain optimization, supporting NVIDIA-accelerated AI processing.
Meta and AMD Form 6GW AI Infrastructure Strategic Partnership
Meta announced a multi-year strategic partnership with AMD to deploy up to 6GW of AMD Instinct GPU computing capacity. The collaboration involves multi-generational integration of AMD GPUs, EPYC CPUs, and jointly developed Helios rack architecture, supporting Meta's diversified computing strategy. First deployments are scheduled for late 2026.
Cisco Launches G300 Chip and Systems for AI Agent-Era Data Center Networking
Cisco introduces 102.4Tbps Silicon One G300 switching chip with liquid-cooled N9000/8000 systems delivering 70% energy efficiency, 1.6T optics support, and Nexus One unified management plane upgrade.
NVIDIA and SK hynix Co-Architect Next-Gen Memory for AI Factories, Locking HBM4 to Vera Rubin
NVIDIA and SK hynix announce a multi-year tech partnership to co-develop next-gen memory for Vera Rubin, RTX Spark, and Jetson Thor. Separately, SK Telecom deploys a gigawatt-scale AI cloud using the full DGX stack, targeting 2027. This elevates SK hynix from supplier to co-architect, strengthening NVIDIA's lock-in on HBM and the AI ecosystem.
Intel's 18A Xeon 6+ and Rack Scale AI: A CPU-Centric Challenge to NVIDIA's Inference Empire
At Computex 2026, Intel launched the 18A-node Xeon 6+ processor, the Rack Scale AI platform with SambaNova's SN-50 RDU, and a fully disaggregated inference service (Vector Core Compute). This CPU-centric hybrid architecture targets agentic AI inference workloads, directly challenging NVIDIA's Vera Rubin NVL72 and GPU-dominated ecosystem.
NVIDIA RTX Spark and Nemotron-3 Ultra: AI Control Shifts from Cloud to Personal Edge
NVIDIA launched RTX Spark personal AI supercomputer (co-developed with MediaTek) and Nemotron-3 Ultra open-source model at GTC Taipei 2026. The N1X chip delivers 1 PFLOPS local AI compute, bringing LLM inference to PCs. This marks NVIDIA's pivot from cloud GPU vendor to edge AI infrastructure monopolist, redefining the PC as an AI-native device.
NVIDIA Absorbs Groq LPU: Feynman GPU to Integrate SRAM Inference Tile, Hybrid Architecture by 2028
NVIDIA secures Groq's LPU inference technology via a non-exclusive license and key hires, planning to integrate large SRAM tiles into its 2028 Feynman GPU using TSMC SoIC hybrid bonding. This enables deterministic scheduling and 80TB/s on-chip bandwidth, shifting NVIDIA from a pure GPU vendor to a hybrid inference/training platform.
SGLang 0.5.13 Delivers 25x MoE Inference Speedup via Predictive Routing and Sparse KV Cache
SGLang 0.5.13 introduces two-stage MoE routing prediction and sparse KV cache, achieving a 25x inference speedup on NVIDIA GB300 NVL72. Benchmarks on A100 show 65% throughput gain, 40% latency reduction, and 62% lower routing overhead. This optimization directly attacks the core bottleneck of MoE inference, potentially reshaping AI inference economics.