Rain AI

Rain AI is a San Francisco-based startup developing neuromorphic processing units (NPUs) for artificial intelligence applications. The company's proprietary Digital In-Memory Computing (D-IMC) cores are designed to mimic the structure and function of the human brain, offering unprecedented levels of performance and efficiency for AI workloads. Rain AI's technology utilizes a block brain floating point scheme that ensures no accuracy loss compared to FP32, while leveraging optimized 4-bit and 8-bit matrix multiplication for immense performance gains. The company's NPUs are built on the RISC-V open source architecture, allowing AI developers flexibility to implement any operator and compile any model. Rain AI claims its chips could potentially offer 100 times more processing power and 10,000 times greater energy efficiency for AI training compared to current GPUs. The company's initial chips are targeted at edge devices such as phones, drones, cars, and robots, capable of both training machine learning algorithms and inference. Founded in 2017, Rain AI has approximately 40 employees as of 2023, including experts in AI algorithm development and traditional chip design. The company expected to tape out its first test chip for fabrication in December 2023 and potentially deliver its first hardware to customers in October 2024.

Key customers and partnerships

Rain AI has held advanced talks to sell its systems to major technology companies including Google, Oracle, Meta, Microsoft, and Amazon. In 2019, OpenAI signed a non-binding letter of intent to purchase USD 51 million worth of Rain AI's chips once they become commercially available. This agreement underscores the potential significance of Rain AI's technology in addressing the growing demand for specialized AI hardware.