Liquid AI company profile: valuation, funding & revenue

This Liquid AI company profile covers the company's valuation, revenue, funding rounds, investors and history. Liquid AI works in artificial intelligence, with a focus on efficient foundation models. It was founded in 2023 and is headquartered in Cambridge, Massachusetts, United States. Its latest valuation on record is More than US$2 billion (reported) (December 2024). Every figure links to a public source, and anything the company hasn't disclosed is marked as such.

Revenue

Funding history

Liquid AI timeline

Key dates from the sources on this page, oldest first. Metric snapshots are listed after the funding events.

• 2023 — Liquid AI is founded in Cambridge by Ramin Hasani (CEO), Mathias Lechner, Alexander Amini and Daniela Rus.
• Dec 2024 — Series A: US$250M, valuing the company at More than US$2B (reported; some outlets cited US$2.3B). Investors: AMD Ventures (lead); OSS Capital, Duke Capital Partners, PagsGroup.

Investors

Overview

Liquid AI develops foundation models based on a different architecture from the transformers behind most large language models. Its liquid foundation models (LFMs) are inspired by research on liquid neural networks, which drew on the simple nervous system of a microscopic worm, Caenorhabditis elegans. The company says its models can match larger models on many tasks while using much less memory and computing power, making them suited to running on devices such as phones, laptops, cars and robots, and to cheaper deployment in data centres. In December 2024 it raised a US$250 million Series A led by AMD Ventures, which TechCrunch, citing Bloomberg, reported valued it at more than US$2 billion.

Liquid AI was founded in 2023 as a spin-out from MIT's Computer Science and Artificial Intelligence Laboratory (CSAIL). Its co-founders are Ramin Hasani, the CEO, Mathias Lechner and Alexander Amini, researchers who developed liquid neural networks, and Daniela Rus, the director of CSAIL and a leading robotics researcher.

Products

• Liquid foundation models: efficient language and multimodal models in a range of sizes.
• On-device AI: models designed to run locally on phones, PCs, vehicles and edge devices.
• Enterprise solutions: customised models for industries such as financial services, consumer electronics and biotech.
• Developer tools: platforms for testing and deploying LFMs.

Funding

Liquid AI raised a seed round in 2023. In December 2024 it raised a US$250 million Series A led by AMD Ventures, with OSS Capital, Duke Capital Partners and PagsGroup participating. TechCrunch, citing Bloomberg, reported a valuation of more than US$2 billion; other outlets reported about US$2.3 billion. Because the company didn't disclose the figure, we label it as reported. As part of the deal, Liquid AI agreed to optimise its models for AMD's GPUs, CPUs and AI accelerators.

Business model

Liquid AI licenses its models and provides customised solutions to enterprises and device makers, with revenue from licences, deployments and services. Efficient models could reduce costs for customers running AI at scale or enable new applications on devices without cloud connections, where privacy and latency matter.

Competition

Liquid AI competes with developers of small, efficient models, such as Meta's Llama, Google's Gemma, Microsoft's Phi, Mistral AI and Cohere; with companies building alternative architectures such as state space models; and with device makers developing their own on-device AI. Its differentiation is its novel architecture and academic pedigree.

Liquid AI at a glance

Liquid AI is a Cambridge, Massachusetts, AI company founded in 2023 as an MIT CSAIL spin-out by Ramin Hasani, Mathias Lechner, Alexander Amini and Daniela Rus. It builds efficient liquid foundation models for devices and enterprises. It raised a US$250 million Series A led by AMD Ventures in December 2024, at a valuation reported at more than US$2 billion.

Timeline

• 2020–2022: liquid neural networks research at MIT.
• 2023: company founded.
• 2024: releases first liquid foundation models; raises US$250 million.
• 2025–2026: releases further models and on-device deployments.

Why efficient AI matters

Large AI models require expensive GPUs and lots of electricity, both in training and every time they answer a query. Running models on devices, rather than in the cloud, can improve privacy, reduce latency and lower costs, but devices have limited memory and battery. Efficient architectures that deliver good performance with fewer resources could enable AI features in billions of devices and reduce data centre costs. That's why chip makers such as AMD and device manufacturers are interested in companies like Liquid AI.

Who funds Liquid AI

Liquid AI's investors include AMD Ventures, which led its Series A, along with OSS Capital, Duke Capital Partners and PagsGroup.

Who uses Liquid AI

Liquid AI's customers include enterprises and device makers in consumer electronics, financial services, automotive and biotech.

Valuation context

A reported valuation above US$2 billion for a young research-led company reflects investor interest in alternatives to transformer models. Our guide to the most valuable AI startups compares it with peers.

Risks

• Competition: major labs release small, efficient models freely.
• Architecture risk: transformers continue to improve.
• Commercialisation: turning research into revenue at scale.

Questions for buyers

• How do LFMs perform on your tasks compared with alternatives?
• What hardware do they run on, and with what memory needs?
• What licensing terms apply?

How liquid neural networks differ

Transformer models, used by most large language models, compare every piece of input with every other piece, which becomes very expensive for long inputs. Liquid neural networks, developed at MIT, are inspired by how simple biological neurons adapt their behaviour over time, and are designed to process sequences efficiently with far fewer parameters. Liquid AI has built on this research to create foundation models that use less memory, especially for long contexts, and can run on smaller hardware. The approach is one of several alternatives to transformers being explored, including state space models. Whether these architectures can match the best transformer models at the largest scales remains an open question, but their efficiency advantages are attractive for devices and cost-sensitive applications.

Liquid AI's partnership with AMD gives it access to hardware expertise and a route to optimising models for a wide range of chips, which matters for on-device deployment.

Device makers and on-device AI

Phone, PC and car makers are racing to add AI features that work without an internet connection, protect privacy and respond instantly. They need models small enough to fit in limited memory and efficient enough to preserve battery life. Liquid AI targets these customers with models optimised for specific chips, competing with models from large AI labs and device makers' own research teams.

Independent benchmarks on real devices will be key to proving its efficiency claims.

If its models run well on mainstream phones and laptops, it could reach very large numbers of users through device partnerships.

Latest developments (October 2026)

Our October 2026 review found no new priced round for Liquid AI after its December 2024 Series A of US$250 million led by AMD Ventures, reported at about US$2.35 billion. The MIT spin-out continues to release efficient foundation models designed to run on phones, laptops and other devices rather than only in data centres.

What to watch

Watch model releases, device partnerships, independent benchmarks and new funding.

Takeaways

Device makers may run capable AI locally. Founders can see how academic research becomes a startup. Investors should note that its valuation is reported.

Sources

Every figure on this page comes from the public sources below. Where sources disagreed, we used the company's own announcement or a primary news report. Private-company data changes; check the date beside each figure.

1. TechCrunch (citing Bloomberg)
2. Maginative