Semiconductors are employed in almost every industry and have become a valuable asset. The development and production of these chips is a complex task. Advanced deep learning features and needs are frequently overlooked by AI chip builders, who feel that software stacks between their silicon and algorithms may provide adequate abstract layers. However, the more native hardware support enabled, the better the performance.
Rebellions Inc, an AI-semiconductor startup, is addressing this issue by developing AI accelerators that bi-directionally bridge the gap between underlying silicon architectures and deep learning algorithms. They are pushing algorithm boundaries to use silicon budgets better and re-architecting AI processors to include complex deep learning capabilities via silicon-dedicated DL kernels.
Rebellions’ first chip, ION, was released in November 2021. The company claims that it reduces latencies and improves trading speeds. It is considered to be twice as fast as Intel Habana Labs’ AI Chip Goya in terms of execution. Its main target customers are global investment banks. Taiwan Semiconductor Manufacturing Company (TSMC) has been contracted to manufacture the ION chips by next year.
The company aims to create a domain-specific AI processor and optimal software for it. Customers will be able to arm themselves with the most potent yet energy-efficient AI hardware as well as seamless software integration because of this technique and a unique decentralized programming model.
The South Korea-based startup has raised $50 million (62 billion KRW) in a Series A funding round. Temasek’s Pavilion Capital backed the round. Korean Development Bank, SV Investment, Mirae Asset Capital, Mirae Asset Ventures, IMM Investment, KB Investment, KT Investment, and existing investors Kakao Ventures, GU Equity Partners, and Seoul Techno Holdings also participated in the round. This round brings the total funding raised by Rebellions to about $80 million (100 billion KRW) at an estimated valuation of $283 million (325 billion KRW).
The company plans to use these funds to double its employees and set up an office in the U.S. They’ll also focus on the mass production of their second AI chip prototype, ATOM, which will target companies in the cloud sector and data centers.