Lumai

What Lumai does

Lumai is an Oxford University spinout founded in January 2022 that builds fully optical computing hardware for AI inference workloads. Its core technology performs matrix multiplication using light (photons) rather than electrons, enabling computation in a three-dimensional volume with massive spatial parallelism rather than the two-dimensional constraints of silicon chips. The product line is the Iris Server Family: Iris Nova, available now for evaluation, demonstrated running Llama 8B and 70B in real time; Iris Aura, targeting 2028; and Iris Tetra, planned for 2029. Nova supports matrices up to 2,048 by 2,048 and uses a three-tier memory architecture optimized for AI KV Cache access alongside hardware-aware quantization schemes. The company claims a 50x performance improvement over transistor-only accelerators, a 90% reduction in power consumption, and 10% of the total cost of ownership compared with silicon-based servers, while emphasizing production-readiness on industry-standard processes without requiring new materials.

Lumai generates revenue through hardware sales of its optical computing servers, targeting hyperscalers, neo-clouds, large enterprises, and research institutes as its primary customer segments. The go-to-market motion is a direct, field-led enterprise sales process in which evaluation units of Iris Nova are deployed to prospects ahead of commercial purchase; pricing is not publicly disclosed. The company is based in Oxford, United Kingdom, and is actively expanding into the United States following its April 2025 Series A round.

Lumai has raised approximately £1.75M in seed funding (IP Group plc and Runa Capital, January 2022), a £1.1M Innovate UK Smart Grant (February 2023), participated in Intel Ignite (September 2023), and closed a $10M+ round led by Constructor Capital (April 2025) with backing from State Farm Ventures, TIS Japan, PhotonVentures, Qubits Ventures, Journey Ventures, LIFTT, and IP Group. It also receives support from the UK government's Advanced Research and Invention Agency (ARIA). Leadership comprises CEO and co-founder Xianxin Guo, co-founder James Spall, and Chairman René Bonvanie.

Name

Lumai

Legal name

Lumai

Website

https://lumai.co.uk

Company type

Private

Founded year

2022

Operating status

Operating

Headcount range

11–50 employees

Short description

Lumai is an Oxford University spinout that builds optical computing servers (Iris Nova, Aura, Tetra) for AI inference, targeting hyperscalers, neo-clouds, large enterprises, and research institutes with photonic matrix-multiplication hardware that it claims delivers 50x performance and 90% power reduction versus silicon-based servers.

Ownership category

akta.pro rank

Product category

Optical AI Accelerators

NAICS

Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (5182)

SIC

Electronic Computers (3571)

akta.pro primary industry

AI Compute Virtualization & Scheduling (GPU virtualization, cluster schedulers) (HDAAAAAG)

akta.pro secondary industry

GPU-Accelerated & AI Training/Inference Servers (HDACABAG)

Keywords

• Optical computing systems
• AI inference accelerators
• Photonic computing hardware
• Optical neural networks
• Data center AI servers

Headquarters

HQ city

Headington

HQ country

United Kingdom

HQ region

Europe

Offices1 record

Markets served

GTM type

B2B

Offering type

Hardware or Manufacturing

Cost components

Technology or R&D, Personnel, Operations, Marketing or Sales, Others

Revenue model

1. Optical Computing Hardware Sales: Sale of optical computing servers (Iris family) for AI inference workloads to hyperscalers, neo-clouds, enterprises, and research institutions. Revenue from hardware products targeting data center deployments.

Go-to-market motion2 records

Distribution channels2 records

Marketing channels3 records

Core offering

Lumai builds and sells the Iris family of fully optical computing servers for AI inference workloads, using photons to perform matrix multiplication in a 3D volume rather than relying on 2D silicon chips. The first commercially available unit, Iris Nova, runs billion-parameter large language models such as Llama 8B and 70B in real time while consuming up to 90% less power than GPU/TPU-based servers.

Product overview

Lumai offers the Iris family of optical computing servers for AI inference in the era of large language models. The product line includes Iris Nova (available now for evaluation), Iris Aura (targeting 2028), and Iris Tetra (planned for 2029). These systems use photonic technology to perform AI inference by processing data as light rather than electrons, enabling 50x performance improvements and 90% power reduction compared to traditional silicon-based servers. The technology processes matrices up to 2,048 x 2,048 and supports billion-parameter LLMs in real-time, demonstrated with Llama 8B and 70B models.

Differentiator

Problem solved

Functional benefit

Products and services

• Iris Nova The first optical computing server in Lumai's Iris family and the world's first optical computing system capable of running billion-parameter large language models (demonstrated with Llama 8B and 70B) in real time. It uses photons rather than electrons to perform matrix multiplication in a 3D volume, handles matrices up to 2,048 x 2,048, and is offered to hyperscalers, neo-clouds, enterprises, and research institutes for evaluation and deployment.

Quantifiable outcome

• 50x AI performance improvement compared to transistor-only accelerators
• +4 more outcomes

Segments4 records

Ideal customer profiles4 records

Technology focussed Yes

API detail

Has API

No

API docs

API detail

Core technology

AI maturity

App detail

AI capability2 records

Feature6 records

Scale indicators7 records

Recent moves7 records

Expansion highlights5 records

Direct peers

• Lightmatter: Lightmatter develops photonic AI processors using light-based computing for AI inference and training, making it the most direct optical AI accelerator competitor to Lumai with overlapping customer targets (hyperscalers, data centers).
• Celestial AI: Celestial AI builds photonic interconnect and compute fabric for AI workloads, addressing similar memory bandwidth and energy efficiency bottlenecks that Lumai targets through its optical computing approach.
• Lightelligence: Lightelligence develops optical computing chips for AI acceleration, directly competing with Lumai in the photonic AI inference market with similar 3D optical matrix multiplication technology.

Emerging players

• Luminous Computing: Luminous Computing is an early-stage optical AI compute startup focused on photonic processors for AI workloads, with overlapping technology approach and target customer base as Lumai.
• Xanadu: Xanadu develops photonic quantum computing hardware (Borealis) and has expertise in optical/photonic systems that overlaps with Lumai's optical compute approach, though serving different computational paradigms.

Broad incumbents

• Nvidia: Nvidia dominates AI inference with its GPU stack (H100, B200) and CUDA software ecosystem. While not optical, it represents the incumbent competitor Lumai must displace in hyperscaler data centers.
• Cerebras Systems: Cerebras builds wafer-scale AI accelerators (CS-3) targeting large language model training and inference at hyperscale, competing for the same enterprise and hyperscaler AI compute budget as Lumai.
• Groq: Groq develops LPU (Language Processing Unit) inference accelerators targeting real-time LLM inference at scale, directly competing with Lumai's Iris Nova for low-latency, high-throughput AI inference deployments.
• Tenstorrent: Tenstorrent designs AI accelerators and RISC-V based compute systems for training and inference, competing for the same enterprise and hyperscaler AI infrastructure spend as Lumai.
• Intel (Habana Labs): Intel's Habana Labs produces Gaudi accelerators for AI training and inference, representing an established commercial alternative for hyperscaler and enterprise AI workloads targeted by Lumai.

Market position

Strengths4 records

Weaknesses4 records

Competitive moat4 records

Key risks6 records

Key highlights6 records

Customer concentration

• LinkedInlinkedin.com/company/lumaitech

Revenue estimate

Valuation estimate

Number of profiles

Profiles3 records

Funding overview

Funding rounds4 records

Investors11 records

Funding detail is available on the Subscription and Enterprise plan.Contact sales →

M&A

Investments

M&A and investment is available on the Subscription and Enterprise plan.Contact sales →

Frequently asked questions about Lumai