Skip to main content

Olix

Category: AI Chips / Semiconductors

A semiconductor startup developing next-generation photonic AI hardware designed to eliminate the need for High Bandwidth Memory (HBM) in large-scale AI inference. Olix was founded in 2024. The company is led by James Dacombe. Based in London, United Kingdom. Team size: 70-200. Total funding raised: $250.0M. Latest round: Series A (Feb 2025). Key investors include ["Plural","Hummingbird Ventures","Vertex Ventures","LocalGlobe","Entrepreneur First"].

Founded
2024
Headquarters
London, United Kingdom
Team size
70-200
Total funding
$250.0M

Value proposition

Bypasses the 'memory wall' and HBM supply bottlenecks by using light-based computing to deliver significantly higher throughput and energy efficiency for AI inference at a lower cost than traditional silicon GPUs.

Products and solutions

["Optical Tensor Processing Unit (OTPU)","Photonic-Electronic Hybrid Architecture","Proprietary Photonic Compute Software Stack","Next-generation AI Inference Accelerators"]

Unique value

Olix is re-architecting the fundamental compute pipeline to forgo HBM entirely, which is currently the most expensive and supply-constrained component in AI hardware (like Nvidia's H100/B200).

Target customer

Hyperscalers, cloud service providers, frontier AI labs (e.g., OpenAI, Anthropic), and enterprises running massive LLM inference workloads.

Industries served

["Semiconductors","Artificial Intelligence (Inference)","Cloud Infrastructure","Data Centers"]

Technology advantage

Leverages photonics to perform matrix multiplications at the speed of light with near-zero heat generation, solving the power-density and memory-bandwidth limitations of traditional CMOS-based electronics.

How they differentiate

Olix differentiates by re-architecting the compute pipeline to eliminate the need for High Bandwidth Memory (HBM) entirely. While competitors like Lightmatter focus on photonic interconnects or hybrid compute, Olix uses an Optical Tensor Processing Unit (OTPU) to bypass the 'memory wall' and supply chain bottlenecks associated with HBM.

Main competitors

["Lightmatter","Celestial AI","Etched"]

Key partnerships

["Plural (Lead Investor)","Hummingbird Ventures","Vertex Ventures","LocalGlobe","Entrepreneur First"]

Notable customers

["Hyperscale Cloud Providers (Stealth)","Frontier AI Labs"]

Major milestones

["Founded in 2024 as Flux Computing","Developed proprietary Optical Tensor Processing Unit (OTPU) architecture","Raised $220M Series A at a $1B+ valuation in February 2025","Successfully exited stealth mode to compete in the AI inference hardware market"]

Growth metrics

Achieved unicorn status ($1B+ valuation) within its first year of operation; scaling team from 70 to 200+ employees.

Market positioning

High-performance AI hardware challenger targeting the LLM inference market dominated by Nvidia.

Geographic focus

Headquartered in London, UK, with a primary market focus on North American hyperscalers and global frontier AI labs.

Patents and IP

Proprietary IP portfolio covering Optical Tensor Processing and photonic-CMOS integration (specific public filings under 'Flux Computing' or 'Olix' are currently in transition/processing).

About James Dacombe

James Dacombe is a serial entrepreneur and Thiel Fellow who dropped out of high school at age 16 to pursue software engineering. At 17, he founded CoMind, a neurotechnology company that raised over $100M to build non-invasive brain-computer interfaces. In 2024, he launched Olix (formerly Flux Computing) to develop next-generation photonic AI hardware, successfully raising $220M in Series A funding at a $1B+ valuation within its first year.

Official website: