01Data + AI
Databricks
Data and AI platform quadrupling its London HQ.
FitzroviaGrowth
Sector · United Kingdom
AI Infrastructure & Chips startups in the UK
The picks-and-shovels layer of AI: GPU cloud, data platforms, ML infrastructure and AI chips. British infrastructure companies are building the compute the rest of the ecosystem runs on.
19
Companies
318
UK & remote roles open
02AI Chips
Graphcore
Designs AI processors and compute systems, now owned by SoftBank.
BristolSubsidiary
03ML Platform
Hugging Face
Open-source machine learning model and dataset hub.
LondonSeries D
04AI Infrastructure
Nscale
AI hyperscaler providing GPU cloud and data-centre infrastructure.
LondonSeries C
05ML Infra
RunLocal AI
AI agent that automates optimisation and deployment of model inference on edge hardware (YC S24).
LondonSeed
06AI Infra
HelixDB
Open-source graph-vector database giving AI agents low-latency memory and relationship-aware retrieval for RAG (YC X25).
LondonSeed
07AI Infra
Parsewise
API for agentic multi-document processing giving risk teams validated, source-traced data and AI oversight (YC X25).
LondonSeed
08ML Infra
Plexe
Open-source agents that build, evaluate and deploy production-grade predictive ML models from a prompt (YC X25).
LondonSeed
09ML Infra
mlop
Open-source, performant experiment tracking for ML training, a Weights & Biases alternative with API compatibility (YC X25).
LondonSeed
10ML Infra
Fuzzy Labs
Open-source MLOps consultancy building bespoke machine-learning infrastructure for teams that want to avoid proprietary vendor lock-in.
ManchesterBootstrapped
11AI Infrastructure
Callosum
Founded by two Cambridge neuroscience PhDs; systems software that breaks an AI workload into tasks and routes each to the model and chip best suited to it, so heterogeneous silicon can stand in for banks of identical GPUs. First equity investment from the UK's Sovereign AI Fund.
LondonSeed
12AI Chips
Fractile
Inference silicon that puts compute and memory on the same die, skipping the DRAM bottleneck; claims LLMs run far faster at a fraction of the running cost. London HQ plus a Bristol hardware engineering site, with first datacentre chips due 2027.
LondonSeries B
13AI Infrastructure
Fireworks AI
San Mateo inference platform for open models, serving 40tn+ tokens a day; valued at $17.5bn in July 2026 and now standing up an EMEA team in London.
LondonSeries D
14AI Chips
Singular Photonics
University of Edinburgh spin-out building single-photon (SPAD) image sensors with compute on the chip, for machine vision, physical AI and medical imaging; its founding professor designed the first SPAD sensor on silicon CMOS.
EdinburghSeed
15ML Infra
Flower
The open-source framework for federated learning — training models on data that never leaves the device or organisation holding it. Founded by three researchers who met at Cambridge, including a Cambridge ML professor who ran Samsung's AI lab there; research labs at MIT, Stanford, Harvard, Berkeley and Oxford build on it. $20m Series A led by Felicis.
CambridgeSeries A
16Data + AI
Orchestra
An agentic control plane putting data pipelines and governed AI agents in one environment, so lean data teams can run both across an existing stack. Customers include Experian and the San Francisco Giants; $4.6m raised, seed led by Differential Ventures.
LondonSeed
17AI Infrastructure
Modal
New York serverless cloud for AI workloads — custom model inference, fine-tuning, RL and agentic sandboxes — scaling from zero to hundreds of GPUs on containers defined in code. Raised $355m at a $4.65bn valuation in May 2026 and is now opening a London office.
LondonSeries C
18AI Chips
OLIX
Photonic inference hardware that moves data between chips with light rather than copper, so large models can be served without high-bandwidth memory. Its first chip, the DX-1 decode accelerator, ships in systems from 2H 2027. Founded in 2024 as Flux Computing; $312m Series B at a $3.3bn valuation, with Arm and the UK's Sovereign AI fund.
LondonSeries B
19AI Infrastructure
Verda
Helsinki full-stack AI cloud, formerly DataCrunch: GPU clusters, managed inference and training run from its own data centres in Finland and Iceland on renewable power, with a London office in Euston. $165m annualised revenue; $189m Series B led by Emergence Capital at a $1bn+ valuation.
Knowledge QuarterSeries B