Olix Raises $312M at $3.3B Valuation To Scale Specialized AI Inference Silicon
As AI has exploded over the last few years, there have been different types of AI chips designed to solve different tasks. Olix raises $312 million to create one chip The post Olix Raises $312M at $3.3B Valuation To Scale Specialized AI Inference Silicon appeared first on Ventureburn .
Olix, a New York-based AI chip startup, has successfully raised an impressive $312 million to create a single chip that handles every aspect of an AI model's work. The investment round was led by New York firm Fundomo, with additional contributions from Arm, Hudson River Trading, and Reed Hastings, co-founder of Netflix. Existing backers such as Hummingbird Ventures, Crane, Plural, Creandum, Phoenix Court, and Transition also participated in the funding.
This infusion of capital has catapulted Olix's valuation to an astounding $3.3 billion, almost tripling its $1 billion mark just six months after its $220 million raise.
The brainchild of James Dacombe, Olix was founded in London in 2024, following his earlier venture in brain monitoring technology. Olix's chief argument to investors is that Nvidia's general-purpose chips are facing a roadblock. Dacombe posits that AI inference data centers operate like factories, producing tokens, and each token requires hundreds of different operations. Most companies are still relying on a single chip to handle all these tasks, prompting Olix to step in.
Olix's proposed solution is the X-1 chip, a system that splits a model's workload across multiple chips instead of relying on one processor. These chips interact using light instead of copper wiring, reducing latency and power consumption. The company's first chip, DX-1, deals with the decode stage, where a model generates its output.
Olix claims that its DX-1 can generate over 10,000 tokens per second per user using less power than a standard chip would at the same batch size. Moreover, the X-1 architecture is designed to scale to models with 10 trillion parameters or more.
One notable aspect of Olix's approach is its avoidance of advanced packaging and high-bandwidth memory. These components are among the most supply-constrained elements in the chip industry, and Olix chose to sidestep these issues to avoid being susceptible to supply queues shared by other companies chasing Nvidia. With the infusion of fresh capital, Olix has bolstered its team, hiring Matt Briers and Nick McKeown to its board.
Briers brings experience as the former CFO of Wise, while McKeown is recognized for co-inventing software-defined networking and OpenFlow.
Despite the impressive performance claims, the real test will come when Olix's DX-1 reaches customers in the second half of 2027. With over 140 employees spread across London, Bristol, Toronto, Austin, and San Francisco, Olix is well-positioned to capitalize on the growing market for AI chips. As the AI chip industry is already estimated to be worth $84 billion and projected to more than triple by 2030, Olix's innovative approach and substantial backing present a compelling case for investors and industry observers alike.
Written by urgent.news from Ventureburn's reporting — not their text. Machine-written — it may contain errors, so check the original before relying on it.
