Apoha Exits Stealth With $36M to Build 'Liquid Brain' AI for Materials

Apoha, a UK-based deep tech startup, emerged from stealth Wednesday with $36 million in Series A funding and a bet that the next big AI breakthrough won't come from text or images. It will come from a new kind of data nobody has at scale yet: how materials vibrate when suspended in liquid and physically stressed.

The Series A was led by Singular, the European venture firm, with participation from Draper Associates and existing seed investors Redalpine, Seedcamp, Wilbe, and Nucleus. The company also holds a grant from Innovate UK.

Why this matters

Most AI breakthroughs in the last decade have ridden a single pattern: a new modality of data gets digitized at scale, then a transformer-style model learns the patterns. Text on the web powered LLMs. Images on the web powered diffusion models. Protein structures from the PDB powered AlphaFold.

Apoha's bet is that the next breakthrough modality is biophysical — the wave forms materials produce when probed. Today, that data doesn't exist at scale because no one has built the hardware to collect it. Apoha is building both the hardware and the AI stack on top.

Whether or not the specific approach works, the broader theme is important. Frontier AI is starting to run out of pre-existing internet data. The labs that find new, proprietary data sources are positioning for the next decade of advantage.

What was announced

Apoha closed $36 million in Series A funding led by Singular. Draper Associates joined. Existing investors Redalpine, Seedcamp, Wilbe, and Nucleus all participated. Innovate UK contributed a non-dilutive grant.

The company is led by CEO Shamit Shrivastava, a mechanical engineer who developed the underlying methods during postdoctoral work at the University of Oxford. Co-founder Anshika Srivastava, a former Goldman Sachs banker, joined to launch the company in 2021.

Apoha calls its approach "liquid intelligence." The core hardware is a laboratory device that takes a sample of material — Apoha says small enough to fit on the head of a pin — suspends it in a liquid, and then applies a controlled series of mechanical stresses. The device records the wave patterns that ripple back through the liquid in response. Those wave forms become the training data for AI models that can predict how new combinations of materials will behave.

The targeted applications include drug discovery, food product design, paints and coatings, and synthetic biology. Apoha's positioning is that wave-form data captures emergent properties of materials — how they actually behave under physical conditions — in ways that molecular-structure data alone cannot.

Industry impact

For pharma and biotech, the pitch is concrete: faster screening of drug candidates by capturing how molecules behave dynamically rather than just how they're structured statically. AlphaFold and its successors solved protein structure prediction. They did not solve how proteins behave in real biological environments. If Apoha's hardware delivers reliable behavioral data, it slots into a real gap in the discovery stack.

For food science and consumer products, the use case is older and arguably more immediate. Reformulating products without animal-derived ingredients, predicting shelf stability, designing new textures — all are problems where existing computational tools fall short and where physical testing is slow and expensive. A hardware-plus-AI shortcut would matter to companies like Unilever, Nestlé, and the broader CPG industry.

For the AI industry, Apoha is a marker of where deep tech funding is going. Investors are increasingly looking past pure software AI to companies that pair AI with proprietary data acquisition systems — hardware that generates training data nobody else has. The thesis is that the next durable AI moats won't come from compute or model architecture but from data access.

Expert perspectives

Fortune's exclusive on the round framed Apoha as a contrast to the noisy generative AI startup market: a deep tech company building its own data infrastructure rather than racing to fine-tune off-the-shelf LLMs. The TheNextWeb coverage emphasized the "matter behavior" angle — how the science fills a gap between molecular simulation and physical testing.

Materials science researchers contacted by trade publications were cautiously interested but flagged a key open question: whether wave-form measurements are predictive enough across material classes to support a general-purpose model, or whether each application domain will need its own specialized version. The answer determines whether Apoha can be a platform or has to commit to one vertical.

Investors at Redalpine and Seedcamp, who backed the seed and re-upped at Series A, pointed to the founders' track record of bridging hardware, physics, and machine learning as the reason they wrote larger checks this time. Deep tech companies usually struggle to scale because the hardware-software interface is hard. Apoha has been pre-revenue building both halves for four years, which is unusual patience for a venture-backed company.

What's next

Apoha plans to scale up production of its measurement hardware, expand its in-house dataset, and place devices with research partners across pharma, food, and synthetic biology. The company is also hiring across AI and materials science.

Watch for the first major industry partnership. A deal with a top-tier pharma company or a CPG giant in the next 12 months would validate the platform thesis. Without one, Apoha will need to demonstrate its own end-to-end product — a discovered material or compound that shipped because of the platform.

For the broader market, Apoha is worth watching as a test case for "proprietary data" AI startups. If the company can show that wave-form data trains models that beat existing computational tools on real materials problems, expect a wave of imitators trying similar plays in adjacent modalities: vibration data, electromagnetic data, thermal response data. The hunt for new training data is on.

Bottom line

Apoha just raised $36 million to digitize a kind of materials data that has not existed at scale. The pitch is bold: a new modality for AI, built on custom hardware, aimed at drug and materials discovery. Whether it works depends on whether wave-form data is as predictive as the company believes. Either way, it's a clear signal of where deep tech AI investment is heading.