
Apoha
London deep-tech company building a layer of empirically measured molecular-behaviour data for AI systems that need to reason about chemistry and materials.
Last refreshed: 7 June 2026 · Appears in 1 active topic
What is 'liquid state intelligence' and why can't AI just learn it from the internet?
Timeline for Apoha
raised £26.7m to build proprietary empirical molecular data for physical-world AI
UK Startups and Innovation: Apoha bets £26.7m on lab data- What is Apoha's liquid state intelligence and what problem does it solve?
- Apoha's liquid state intelligence is a layer of empirical molecular-behaviour data collected from laboratory experiments. It addresses a gap in physical-world AI: current models trained on internet text lack accurate molecular-scale data needed for materials Science, drug discovery and robotics.Source: Apoha funding announcement, June 2026
- Why can't AI models just learn molecular behaviour from existing data on the internet?
- Molecular behaviour at the scale Apoha is capturing has not been digitised in a form accessible on the internet. It must be measured empirically in laboratories; there is no equivalent of a web scrape for fine-grained physical-world molecular data.Source: Apoha funding announcement, June 2026
- Who invested in Apoha and how much has the company raised?
- Apoha raised £26.7m ($36m) in June 2026, led by Singular with Seedcamp, Draper Associates and Redalpine, alongside an Innovate UK government grant.Source: Funding press release, June 2026
Background
Apoha is a London deep-tech company that raised £26.7m ($36m) on 3 June 2026, led by Singular with Seedcamp, Draper Associates and Redalpine following, alongside an Innovate UK grant. The company is building what it calls "liquid state intelligence": a layer of empirical molecular-behaviour data collected from laboratory experiments that cannot be reproduced by scraping internet text. This data is intended to underpin AI systems that need to reason about the physical world accurately, rather than rely on statistical patterns from digitised human knowledge.
The core thesis is that current large language models and physical-world AI systems (including robotics, materials Science and drug discovery) are limited by the absence of reliable, empirically derived molecular data. Apoha's approach is to instrument laboratory environments and capture measured behaviour of matter at the molecular scale, building a proprietary corpus that is structurally different from anything a web scrape can produce. The Innovate UK grant alongside the equity round signals that the UK Government recognises the strategic importance of physical-world AI data infrastructure.
Apoha occupies an unusual position in the AI landscape: not a model company, not an application company, but a data-infrastructure layer for physical-world AI. If the thesis holds, the value capture model resembles that of proprietary scientific databases, with high switching costs once researchers and model trainers depend on the data. The involvement of Draper Associates, a US-originated firm with a long track record in deep Science, alongside European-first investors Seedcamp and Redalpine, gives the round a transatlantic character that reflects the cross-border appeal of the company's unusual data layer.