Definition
- 2026-07-30: Programmable networking silicon round — xsight-labs E1 DPU / X2 switch (2026-08-01-xsight-labs-official-300m)
AI hardware refers to specialized processors and accelerators designed for artificial intelligence workloads, including training and inference operations for machine learning models.
Recent Developments
-
2026-07-23: etched doubles valuation to $10.3B in ~7 months (2026-07-23-etched-300m-series-c-10-3b-valuation)
-
2026-07-22: Anthropic commits up to 2GW AMD mi450 Helios capacity (amd-anthropic-2gw-mi450-partnership, gpu-vendor-diversification)
-
2026-06-24: openai + broadcom jalapeno-chip — frontier lab enters custom inference silicon; 9-month ASIC cycle with AI-assisted design (2026-06-24-openai-broadcom-jalapeno-inference-chip)
Key Categories
- Training Chips: GPUs and custom ASICs optimized for model training (e.g., NVIDIA H100, H200)
- Inference Chips: Specialized processors for deploying trained models (e.g., jalapeno-chip, NVIDIA Blackwell, cerebras, etched)
- Edge AI Hardware: Low-power processors for on-device inference
- Neural Processing Units (NPUs): Specialized AI accelerators in consumer devices
Related
-
amd-helios Concepts