Qualcomm Inc Snapdragon Summit 2026 Day 2

Good morning, everyone. Welcome to day 2 of Snapdragon Summit. I am going to start day 2 with a very interesting conversation. We talked to you yesterday about this transition to the AI smartphone. Hope everybody understood how that is going to play out, and this is going to be a great example of what is happening. I am very pleased to welcome, to have this conversation with me on stage, Ramin Hasani, co-founder and CEO of Liquid AI. Please come in, Ramin. Thank you so much for being here.

Nice to be here. Thank you.

Great to be here. This is going to be a conversation that is going to probably give you yet another example of how this thing is really happening right now.

We are going to get started. Ramin is a great guy. Every guy with a mustache is a great guy. Why do not we start? I think everybody wants to know about yourself and about Liquid AI.

Thank you. Yes, I am Ramin Hasani, co-founder and CEO of Liquid AI. I started this company three and a half years ago, spun out of MIT CSAIL. We have been decade long researching how to maximize the amount of intelligence capabilities you can get into the smallest unit of compute. This is devices. We wanted to enable physical AI. We wanted to enable AI agents inside the physical AI. Flash forward, Liquid AI, we are a foundation model company. We are building very efficient models that can be the core of agentic behavior. At the technology of our company, at the core, is a hardware in the loop approach to really design foundation models from scratch. We do not do post-hoc optimization. You can do those post-hoc optimization, but we care about what we do beforehand. We design the models to be inherently designed for the device world.

The technology that comes out of it is called Liquid Foundation Models, LFMs. These LFMs are very popular. We open weight them. There are about 1.4 million downloads on a weekly basis. On Hugging Face, you can actually see that. The models are ranging from 200 million parameters to 24 billion parameters. This is a fabric of layers that can enable all sort of devices, from wearables to cars, or maybe even PCs, bigger compute. We have built two sets of products at this company. Around the models, we built something that we call Model plus X, model plus a harness that is ready to get deployed on top of device. This is product number one.

Product number two is a model development stack that allows us to take these opportunities and give OEMs, in a B2B fashion, give OEMs the opportunity to really customize this model plus harness for their applications and downstream use cases that they want to do. The goal is very ambitious. We want to be deployed. We want to have LFMs deployed on 50% of world's annually produced devices across wearables, handsets, automotive, PCs, robots, and industrial devices.

That is very good. We like ambitious goals. Look, in building on the conversation we had yesterday, from your perspective, where do we think we are right now in the AI journey, and what tells you that we are now entering a new phase?

We are seeing that AIs are becoming very capable. Capability is increasing exponentially. We are seeing fundamental math innovations. We are seeing better reasoning models. We see longer time and more complex workflows are getting done. There is always a disconnect. You always see that individual product, if you look at the metrics, individual productivity based on these AI models that we have today, is way ahead of business transformation. The gap is always between how are you bringing these capabilities and matching them to value, to actual value? That gap has been the problem. I feel like on the AI journey, we have been on this. The fundamental question that we have to address is production-grade AI. How can we make AI to be useful? How can we increase the amount of use of AI? That is something that is extremely difficult to get.

On this journey as well, what does production AI actually mean? For me, it means reliable actions, agents that can take reliable actions, quality of agents that are really high at the level of frontier that we are seeing. Verifiable outputs. The outputs of the systems are completely verifiable. At the same time, you are satisfying the deployment constraints that you have. Use cases, they might need privacy, they need connectivity, they need many different aspects, security, and sensitivity. Then there is use cases, and then there is deployment environments. You have the physical world deployments. You have the cloud deployments. You got to be able to satisfy those things also with the technology.

I think the next wave of this technology, basically hardware and kernel and model and agents, all basically co-design of these things together, is something that I am looking forward, and I think we are on the right path to get there.

Very good. What must change now for agents to become the next computing platform across billions of devices? All those devices you talk about.

I think five things has to change. The first one is quality of. We are talking about device intelligence. We need to reduce the gap between frontier-level performance on the cloud and devices. That is something that is absolutely important. One of the reasons why the device AI and local AI revolution hasn't happened yet, for me, is just that quality bar. Getting to that quality bar is number one. Number two is context. When we are thinking about devices, we are talking about a fabric of devices. Users want to use AIs on one device and then be able to transfer their experiences and preferences to other devices as well. How do you do that? You need to have the context layer. You need to have something that allows you a memory layer that connects these devices together. Third thing I would say is a harness.

Harness is all the toolings that goes around the foundation model. When we talk about agentic AI, we are talking about a model plus a harness. That harness, the adaptation of the harness, and the adaptation of a model should happen at the same time. The fourth thing I would say is understanding the limitations of the physical world. For example, we are talking about the real world has constraints, devices have constraints, understanding the constraints of processors, and then feeding them back to the model design process. That becomes the co-design process. The last thing I would say is always developers, developers. We need basically an army of developers. We need to open source. We need to give developers access to the technologies to build to really bring the AIs to billions of devices. You're talking about local AI.