Etched Emerges from Stealth with $800M Raised for AI Inference Systems
Etched emerged from stealth mode, announcing a working AI inference chip, more than $1 billion in signed customer contracts, and $800 million raised ...
In Focus: Etched
🏢 Etched AI Transformer-Specific AI Inference Silicon Updated: June 30, 2026 | |
| Etched AI is a private semiconductor startup building vertically integrated AI inference systems. Its rack-scale platform combines custom transformer-specific silicon, networking, memory, cooling and software to accelerate large language model inference with higher throughput, lower latency and improved power efficiency than conventional GPU-based systems. | |
| Why It Matters | Rather than selling standalone AI accelerators, Etched is designing complete inference clusters optimized for frontier AI models. The company’s vertically integrated approach reflects a broader industry shift toward co-designing chips, interconnects, memory, cooling and software as AI infrastructure scales toward gigawatt-class deployments. |
| Founded | 2022 |
| Headquarters | San Jose, California, USA |
| Leadership | Gavin Uberti, Co-founder & CEO Rob Wachen, Co-founder |
| Company Type | Private semiconductor startup |
| Funding | Reported total funding of approximately $800 million, including a $500 million financing completed in December 2025 at a reported $5 billion post-money valuation. |
| Commercial Traction | Etched reports more than $1 billion in signed customer contracts, with rack-scale inference systems currently undergoing customer validation ahead of production shipments. |
| Technology Status | First-pass (A0) silicon on TSMC’s N4P process is operational. Etched says its rack-scale systems are validating with customers and running DeepSeek, Qwen, Mamba and Llama models ahead of production shipments planned for Summer 2026. |
| Key Technology | Sohu Transformer ASIC • Low Voltage Inference (LVI) • Cluster Scale Memory (CSM) • Rack-Scale AI Inference • Custom Interconnect • Transformer-Specific Architecture |
| Roadmap | First production racks ship in Summer 2026. The company is targeting gigawatt-scale AI inference deployments beginning in 2027 while expanding manufacturing and validation capabilities. |
| Team | More than 400 engineers with experience from NVIDIA, Google TPU, Broadcom, SK hynix, TSMC, quantitative trading firms and other AI infrastructure organizations. |
| Editorial Coverage | Converge Digest tracks Etched across AI infrastructure, custom silicon, inference acceleration, rack-scale AI systems, semiconductor startups, transformer architectures and next-generation AI data center platforms. |
| Explore More | AI Infrastructure • Semiconductors • Inference • Custom Silicon • Data Centers |
Etched.ai targets highest performance ASIC for AI transformers
Etched.ai, a start-up based in Cupertino, California, has introduced its Sohu ASIC, a specialized chip designed specifically for transformer architecture-based ...
A private dossier for networking and telecoms