Databricks Co-Founder Ion Stoica Joins Liquid AI's Advisory Council
Ion Stoica, the Berkeley professor behind Databricks, Anyscale, and Chatbot Arena, joins Liquid AI's advisory council to help scale its hardware-aware, on-device AI platform.

- Ion Stoica joins Liquid AI's Advisory Council to guide the company's growth as it builds hardware-aware, on-device AI infrastructure.
- Stoica's track record is unmatched: co-founder of Databricks (valued at $100B+), Anyscale ($1B), and Chatbot Arena/LMArena ($600M).
- Liquid AI raised $250M at a $2.35B valuation in a Series A led by AMD Ventures, with ~121 employees as of mid-2026.
- Liquid Foundation Models (LFMs) use a hybrid architecture (short convolutions + attention) to deliver 2x faster CPU inference than Qwen3 at comparable sizes.
- Major enterprise partnerships already in place with Mercedes-Benz and Shopify, targeting in-car AI and sub-20ms e-commerce inference.
- The strategic fit: Stoica's expertise in vertically integrated AI systems aligns directly with Liquid's hardware-aware, full-stack deployment approach.
Liquid AI, the MIT-spinout building a non-transformer AI architecture for edge and physical deployment, has added one of the most consequential infrastructure builders in modern computing to its advisory bench. Ion Stoica, UC Berkeley professor and co-founder of Databricks, Anyscale, and Chatbot Arena (now LMArena), has joined the company as a strategic member of its Advisory Council.
The man who built the AI infrastructure stack
If you have ever run a Spark job, scaled a distributed training workload with Ray, or checked a model's Elo score on Chatbot Arena, you have used something Ion Stoica helped build. He co-founded three companies: Anyscale in 2019, Databricks in 2013, and Conviva in 2006. He transformed the open-source projects Spark and Ray into successful AI infrastructure companies. More recently, Chatbot Arena emerged from UC Berkeley's Sky Computing Lab, led by a group of graduate students and faculty including professors Joseph Gonzalez and Ion Stoica.
After its latest financing round, the valuation of AI data platform Databricks exceeded $100 billion, making it expected to become the fourth most highly valued AI unicorn globally. In addition to Databricks, Stoica co-founded the AI managed computing platform Anyscale with a valuation of $1 billion, LMArena with a valuation of $600 million, and the video-stream analysis technology company Conviva with a valuation of $300 million. That is a remarkable track record for someone who has never left his Berkeley professorship.
His research fingerprints are all over the modern AI stack. His area of research sits at the intersection between AI and systems, cloud computing, and distributed systems. His founding team at Inferact includes Stoica, who is currently the director of UC Berkeley's Sky Computing Lab, which developed the original version of vLLM in 2023. vLLM is now one of the most widely used inference engines in production deployments.
What Liquid AI is actually building
Liquid AI, a startup founded by MIT computer scientists in 2023, introduced its Liquid Foundation Models series (LFM) with a pitch to deliver the fastest on-device foundation models using a new "liquid" architecture with training and inference efficiency. The goal is to make small models a serious alternative to cloud-only LLMs. In 2024, Liquid introduced its first generation of LFMs and argued that the dominant transformer architecture was not the final answer for every AI workload. In 2025, it released LFM2, a new generation built around a hybrid architecture that combines short convolution blocks with attention.
The practical result is a model family that punches well above its weight class on constrained hardware. LFM2 delivers twice as fast decode and prefill performance as Qwen3 on CPU and significantly outperforms models in each size class, making them ideal for powering efficient AI agents. The latest iteration, LFM2.5-1.2B, is the company's most capable release yet for edge AI deployment, building on the LFM2 device-optimized architecture.
The company's ambitions span well beyond chatbots. In a recent partnership, Liquid AI and Mercedes-Benz announced a collaboration to scale embedded in-car intelligence, with Mercedes using Liquid's models to bring more AI capability directly into vehicles, with an emphasis on local performance, efficiency, responsiveness, and privacy. In a multi-year partnership, Liquid said Shopify would license LFMs for search and co-develop a generative recommendation system designed for sub-20-millisecond inference.
Why this hire makes sense right now
Liquid AI is not just a model company. It is building a full deployment stack for hardware-aware AI, meaning models that are designed around the specific constraints of the chip they run on, not just ported to it after the fact. That is exactly the kind of systems-level thinking Stoica has spent his career on. Stoica has warned that tomorrow's workloads will demand vertically integrated, multi-accelerator systems. Liquid is betting its entire architecture on that thesis.
The timing also reflects where Liquid is in its growth arc. Liquid AI secured $250 million in Series A funding at a $2.35 billion valuation, with AMD Ventures leading the round alongside existing investors including Automattic and OSS Capital. The company was founded in 2023 by Daniela Rus, Ramin Hasani, and Alexander Amini. With capital secured and partnerships with Mercedes and Shopify in hand, the next challenge is scaling the platform and the go-to-market motion. That is where an advisor who has done it three times becomes genuinely useful.
AMD Ventures' lead investment reflects strategic alignment between Liquid AI's efficiency-focused architecture and AMD's hardware capabilities. Traditional transformer models require expensive NVIDIA GPUs with high memory bandwidth. Liquid neural networks' reduced computational requirements enable efficient operation on AMD processors and GPUs, potentially disrupting NVIDIA's dominance in AI inference workloads.
What it means for the broader landscape
Stoica's involvement signals something beyond a resume boost. He has a specific, demonstrated pattern: take a research-lab system that solves a real infrastructure problem, commercialize it, and let the open-source community do the distribution work. Two companies that Stoica founded, Databricks and Anyscale, use open-source software to support large swaths of today's AI infrastructure. Liquid's open-weight model releases on Hugging Face and its LEAP developer platform follow a similar playbook.
The physical AI angle is also worth watching closely. The AI industry has spent years optimizing for cloud-scale inference. The next wave, from in-car assistants to robotics to wearables, requires models that run locally, offline, and on cheap silicon. Shifting large generative models from distant clouds to lean, on-device LLMs unlocks millisecond latency, offline resilience, and data-sovereign privacy, capabilities essential for phones, laptops, cars, robots, wearables, satellites, and other endpoints that must reason in real time.
For practitioners, the near-term implication is straightforward: Liquid's models are worth evaluating for any workload where latency, privacy, or hardware cost is a constraint. The Stoica advisory role does not change the models, but it does suggest the company has the strategic credibility to become a serious platform, not just a research curiosity.