Lumai Iris Nova: Optical AI Inference Server
Lumai's optical AI inference server uses light-based computing to run billion-parameter LLMs with up to 90% less power than GPUs.
Hardware
Skymizer HTX301: 700B LLMs at 240W on PCIe
Skymizer's HTX301 uses six 28nm chips and 384 GB LPDDR5 to run 700B-parameter LLMs on a single PCIe card at just 240W.
Meta MTIA 450: 18.4 TB/s Inference Accelerator
Meta's MTIA 450 doubles HBM bandwidth to 18.4 TB/s and adds FlashAttention hardware acceleration for GenAI inference in 2027.
AMD Helios: 72-GPU Rack for AI at Scale
AMD Helios packs 72 Instinct MI455X GPUs and 31 TB HBM4 into a single rack delivering 2.9 FP4 ExaFLOPS for AI workloads.
Meta MTIA 400: GenAI Inference at Scale
Meta's second-gen ASIC delivers 6 PFLOPS FP8 and 288 GB HBM for GenAI and recommendation inference inside Meta's data centers.
Cerebras WSE-3 - The Wafer-Scale AI Engine
The Cerebras WSE-3 is the largest chip ever built - a TSMC 5nm wafer with 900,000 AI cores, 44GB SRAM, and 21 PB/s bandwidth. Now powering a $20B OpenAI deal and Amazon Bedrock deployments.
Google TPU 8i - Low-Latency Inference for Agent Era
Google's TPU 8i is a purpose-built inference chip with 10.1 FP4 PFLOPs, 288GB HBM3e at 8,601 GB/s, and a Boardfly topology that cuts collective latency 5x for agentic AI workloads.
Google TPU 8t - AI Training at ExaFLOP Scale
Google's TPU 8t packs 12.6 FP4 PFLOPs and 216GB HBM3e per chip, scaling to 9,600-chip superpods with 121 ExaFLOPS and 2 petabytes of shared HBM for massive model training.
Qualcomm AI250 - Near-Memory Computing for Inference
The Qualcomm AI250 applies near-memory computing to the same 768GB LPDDR5X design as the AI200, promising 10x higher effective memory bandwidth and lower power for LLM inference at rack scale.
Rebellions RebelRack - 64 FP8 PFLOPs at 5 Kilowatts
The Rebellions RebelRack packs 32 Rebel100 chiplet NPUs with 4.5TB HBM3E and 153.6 TB/s aggregate bandwidth into a rack drawing just 5kW - roughly 4x the compute-per-watt of an H100 DGX.
AMD Instinct MI430X - Dual-Precision CDNA 5 Accelerator
The AMD Instinct MI430X is AMD's CDNA 5 HPC accelerator with 432GB HBM4, full FP64 support, and 19.6 TB/s bandwidth - designed for sovereign AI and scientific supercomputing alongside the MI455X AI GPU.
NVIDIA Groq 3 LPU - SRAM-Based Inference Engine
The NVIDIA Groq 3 LPU is a pure-SRAM inference chip delivering 150 TB/s memory bandwidth and 1.2 PFLOPS FP8 per chip, designed to pair with Vera Rubin GPUs for trillion-parameter model serving.
