Rebellions raises $124 million to develop new AI Rebel chip with Samsung

Image credits: Jason Mars/Getty Images

Rebellions, a South Korean fabless AI chip startup, announced today that it has completed a $124 million (165 billion won) Series B funding round to develop its third AI chip, called Rebel. The company also plans to use the new funding, which exceeded its initial goal of $90 million, to expand production of its Atom data center chip and hire employees.

Rebellions CFO Sunggye Shin said in an exclusive interview with TechCrunch that the three-year-old Series B startup is valued at about $658 million (880 billion won) post-money. . This latest capital injection brings Rebellions’ total funding since its launch in 2020 to approximately $210 million.

South Korean telecom giant KT led this latest round as a strategic investor. In addition to previous backers Temasek’s Pavilion Capital and Korea Industrial Bank, new investors including Corellya Capital and DG Daiwa Ventures also participated.

Rebellions’ funding comes at a critical time for the chip industry, particularly regarding the development and use of AI chips.

Nvidia is a leader in the AI ​​chip market, and its name is synonymous with the AI ​​boom currently sweeping the technology world. Many have observed that how Nvidia has flourished is also due to the moat it has built around its hardware and software ecosystem. But for the rest of the field, it’s not game over yet. Data processing and the high costs associated with it continue to be major challenges when it comes to AI applications, and the race for innovative breakthroughs to improve them continues.

Development is coming from multiple directions. Big tech companies like Google, Amazon, Apple, and Microsoft have developed or own their own chips to integrate AI into their products and services. OpenAI CEO Sam Altman visited South Korea last week and reportedly met with the country’s chip industry leaders Samsung and SK Hynix. Additionally, Open AI is said to be raising billions of dollars to set up a chip manufacturing factory and manufacture its own AI chips. In addition to Rebellions, there are many other startups that are bringing new concepts to speed up processing while increasing efficiency.

The funding, which has been rumored for months, comes on the heels of other moves by the startup. Last October, Rebellions announced it was partnering with Samsung Electronics to develop the latest Rebel chips, building on a relationship originally built around Atoms chips. Singh said the two companies aim to complete development of Rebel by the end of this year and begin mass production in 2025, with the next-generation AI chip running large-scale language models (LLM) and hyperscalers. It added that it would target the generative AI market.

Shin told TechCrunch that Rebel uses Samsung Electronics’ 4-nanometer manufacturing process and that the company’s AI chips are designed to handle the high-bandwidth memory used to build and manipulate large language models. Samsung said its advanced memory chip technology will be introduced in HBM3E. Rebellions’ unique selling point is his claim that its technology and products are more versatile than customized AI chips and can support a variety of generative AI models that require AI accelerators.

The company’s CFO emphasized that Rebellions will work with Samsung from joint development and chip design to mass production of Rebel. Samsung’s efforts here have his second motivation. Apart from its chip efforts, South Korea’s largest memory chip maker is working on its own generative AI model, Samsung Gauss.

We have also worked with customers using previous generation chips. In May 2023, Rebellions’ strategic investor, KT, installed Atom, his AI chip targeted at Rebellions’ data centers, on its cloud-based Neural Processing Unit (NPU) infrastructure. Rebellions expects to generate revenue from Atom in the second half of this year and says it will continue to produce its chip models through Samsung’s 5-nanometer manufacturing process. Shih pointed out that Atom is designed for data centers and language models with up to 7 billion parameters, while Rebel is targeted at larger, large-scale language models.

Meanwhile, the company’s first AI chip, “Ion,” released in November 2021, is undergoing certification testing in the United States, but has not yet signed a contract with a commercial customer. Ion is designed for edge computing, and one of the primary use cases is financial services applications, where large institutions that build their own hardware use the chip to run stock predictions and trading applications. The company believes that it can strengthen the

Rebellions CEO Sunghyun Park is a former quantitative developer at Morgan Stanley in New York, and his four co-founders founded the AI ​​chip startup in 2020.