Etched, a company focused on building frontier inference clusters, has emerged from stealth with significant momentum. The firm announced it has secured over $1 billion in customer contracts and raised a total of $800 million in financing. This launch marks a major new entry into the competitive AI hardware market, promising to make inference more efficient.
A New Player in AI Infrastructure
Founded on the principle that AI's potential is constrained by inference costs, Etched aims to reshape the landscape. The company has developed a new class of inference systems, co-designed across the entire technology stack for optimal performance. This integrated approach is engineered to dramatically improve throughput and latency while bending the cost curve for AI deployment.
The company's financial backing underscores strong investor confidence in its mission and technology. Etched has raised $800 million across several rounds, with the most recent being a $500 million financing at a $5 billion valuation. Key investors include VentureTech Alliance, Peter Thiel, Jane Street, and numerous other prominent firms and AI luminaries.
From Silicon to Systems
In a remarkable display of execution, Etched achieved first-pass silicon success on TSMC's advanced N4P process in less than three years. The company is now actively validating its initial rack-scale product directly with its customers. This rapid progress from concept to working hardware signals a high level of technical and operational capability within the organization.
Etched's systems are designed for broad compatibility and future-proofing in the fast-evolving AI landscape. They are currently running popular models such as DeepSeek, Qwen, Mamba, and Llama. The architecture is built to support models of all shapes and can scale to handle arbitrarily large numbers of parameters.
Scaling for a Gigawatt Future
Co-founder and CEO Gavin Uberti emphasized the growing urgency for better AI infrastructure. "As AI rapidly embeds in every industry and application, the need for accelerated inference infrastructure has never been greater," he stated. This vision addresses the bottleneck that currently exists between developing powerful AI models and deploying them economically at scale.
The company is already preparing for massive deployment, with a focus on manufacturing and operations. Co-founder Rob Wachen noted their strategy is to "build for gigawatt-scale," establishing a Taiwan factory and a comprehensive San Jose headquarters. This operational focus is designed to translate advanced technology into systems that can be deployed reliably and at scale.
A key asset for Etched is its formidable team of more than 400 professionals. The company has attracted top talent from industry leaders like NVIDIA, Google's TPU program, Broadcom, and high-frequency trading firms. This concentration of expertise is crucial for building and scaling the complex systems required for frontier AI inference.
Etched's public debut establishes it as a serious contender in the AI infrastructure race, backed by substantial capital and customer demand. With its working silicon, a clear vision for scalable production, and a world-class team, the company is poised to challenge incumbents. Its mission to make inference abundant could unlock the next wave of AI innovation across the global economy.