The escalating race for artificial intelligence supremacy is increasingly shifting focus from model development to the underlying hardware infrastructure. Current general-purpose chips are struggling to efficiently handle the complex, multi-stage process of AI inference, where models reason and generate output. Olix aims to disrupt this by building a "token factory" – a production line approach with specialized machines for each stage of token production.

Olix, a UK-based AI chip startup founded two years ago, announced a substantial $312 million Series B financing round, pushing its valuation to $3.3 billion. The funding round saw participation from prominent investors including Fundomo, Arm, and Hudson River Trading, alongside angel investors such as Netflix co-founder Reed Hastings. Existing investors also significantly increased their commitments, reflecting strong confidence in Olix's vision.

The company's core innovation lies in its X-1 platform, which features the DX-1 decode accelerator. This specialized chip is engineered specifically for the output generation stage of AI models. Olix claims that for 100-billion parameter models, the DX-1 can achieve over 10,000 tokens per second per user, delivering superior output token throughput per watt compared to general-purpose chips running large batch sizes. This approach aims to make AI inference more affordable and abundant.

Further bolstering its strategic direction, Olix has appointed Professor Nick McKeown to its board of directors. McKeown is a renowned figure in networking, known for co-inventing software-defined networking, OpenFlow, and P4, and is a Professor Emeritus at Stanford University. Additionally, Matt Briers, former CFO of Wise, has joined as Chief Financial Officer. These appointments bring deep industry expertise and financial acumen, crucial for scaling a hardware startup in a competitive market.