Etched Inc. raises $800M

Etched Inc., a newcomer in the artificial intelligence inference chip sector, has launched with a substantial $800 million fundraise. The financing round, culminating in December, values the company at $5 billion and highlights Etched's ambitions in the competitive landscape of AI hardware. VentureTech Alliance, a fund connected to Taiwan Semiconductor Manufacturing Co. (TSMC), spearheaded the investment with participation from over a dozen investors, including AI luminaries like Geoffrey Hinton, Fei-Fei Li, and Andrej Karpathy.

Etched aims to leverage TSMC’s advanced N4P process to manufacture its inference chips, achieving a performance edge with the 11% improvement over TSMC's original five-nanometer node. Prototype chips have already been produced via this technology, indicating significant progress toward bringing their product to market. Focused solely on AI inference — which removes the need for training circuits — the company claims its chips will enhance processing speed and reduce power consumption, compared to multi-purpose graphics processing units like Nvidia's Rubin.

The company has introduced proprietary innovations, such as the LVI technology intended to minimize thermal throttling, a common issue in high-performance chips. Etched asserts that its technology can achieve peak floating-point operations per second (FLOPs) without requiring clock rate reductions, potentially offering a performance that exceeds current market-leading AI processors. This technological advancement positions Etched’s chips as a formidable competitor in AI inference, prioritizing power efficiency and speed.

Etched is preparing to market its chips as part of a comprehensive rack-scale inference appliance, equipped with a mix of SRAM and HBM memories, custom interconnects, and sophisticated cooling systems. The infrastructure aims to enhance the memory processing capabilities and reduce latency in data handling, essential for efficient large-scale AI applications. Already, the company has received orders surpassing $1 billion, reflecting robust demand for its upcoming product line, with initial shipments expected this summer.

The funding and rapid progress come amidst a highly competitive AI hardware market, dominated by major players focused on both AI inference and training. Etched's focus on inference-specific chips could challenge existing giants by offering tailored solutions that promise increased efficiency and performance. The success of the venture will depend on its ability to meet production timelines and convince industry players of its superior performance metrics. As the company ramps up chip production, the sector will closely watch whether Etched can capitalize on its initial momentum and disrupt the existing market dynamics.