Olix Raises $312M to Push Optical Inference Appliances with DX-1 Chip
Olix
optical interconnect
LLM
KV cache
inference workflow
Series C funding
data center appliance
7
Deep Hardware Focus: Strategic Shift to Inference Efficiency
Media Hype
5/10
Real Impact
7/10
What is the Viqus Verdict?
We evaluate each news story based on its real impact versus its media hype to offer a clear and objective perspective.
AI Analysis:
The technical depth and significant funding underscore a genuine, high-impact shift in hardware architecture (Impact: 7), though the detailed technical nature keeps the immediate media hype from being exponential (Hype: 5).
Article Summary
Olix Computing secured $312 million in a Series C funding round, valuing the company at $3.3 billion. The company is focused on optimizing large language model (LLM) inference by developing the DX-1 chip, specifically tailored for decode workloads. The core innovation involves creating a data center appliance (X-1) where multiple chips are linked via optical interconnects transmitting data as light. Crucially, the DX-1 chip utilizes on-chip SRAM memory to store large KV caches, avoiding the need for slower HBM memory, while the optical linking employs a reliable 'slow and wide' NRZ encoding method, which minimizes complexity and cost compared to traditional PAM4 interfaces. Olix aims to start shipping the full system in the first half of 2027.Key Points
- Olix secured $312 million in Series C funding, backed by industry leaders including Arm and Netflix co-founder Reed Hastings.
- The DX-1 chip is designed to address LLM inference limitations by keeping large KV caches in fast on-chip SRAM, bypassing the slower HBM memory structure.
- The system uses slow and wide optical interconnects (NRZ encoding) to link multiple X-1 racks, allowing for processing of models up to 10 trillion parameters.
- The company plans to begin shipping its advanced data center appliance in the first half of 2027.