“Many top models lock memory in. We’re building memory as infrastructure: one layer that works across every model, every framework, every platform,” said Taranjeet Singh, co-founder and CEO of Mem0, to AIM Media House.
Mem0 has raised $24 million to expand what it calls the memory layer for artificial intelligence. The company wants to give AI systems the ability not just to respond intelligently, but to remember, maintaining context and continuity across time, applications, and users.
Read: Mem0’s Commitment to AI Agents with Improved Memory
A memory layer for the agent era
The round combines earlier seed funding with new capital from investors including Basis Set Ventures, Peak XV Partners, GitHub Fund, and others. Mem0 said the funds will accelerate development of its infrastructure stack, which lets any AI agent or model recall, update, and refine long-term context.
The problem is what Singh calls digital amnesia. “Users paste the same context into ChatGPT over and over. Coding assistants suggest the same rejected patterns dozens of times. Our most sophisticated intelligence is trapped in this cycle of forgetting,” he said.
Mem0’s approach is to externalize memory into a shared layer that remains independent of the model or framework being used. “Big labs are building memory as lock-in,” Singh explained. “They use memory in their consumer products as a retention mechanism. Developers don’t want that constraint. We’re building memory as infrastructure.”
That model-agnostic stance is central to the company’s positioning. The system handles complexity such as deciding which memories to keep, which to forget, and how to reconcile conflicting information. “Behind our three-line API, we handle the complexity: extracting memories, managing what’s relevant and what’s outdated, updating intelligently when facts conflict,” Singh said.
Mem0’s usage metrics illustrate momentum. API calls grew from 35 million in Q1 to 186 million in Q3 2025, and the company’s technology is now integrated with CrewAI, Flowise, and Langflow. Amazon Web Services selected Mem0 as the exclusive memory provider for its Strands Agent SDK, allowing AI agents built on AWS to retain long-term context across sessions.
According to Singh, the new funding supports the company’s three-part roadmap: Make it Work, Make it Neutral, Make it Portable. In practice, that means making memory reliable at scale, ensuring it functions across different models, and building toward portability so user context can move seamlessly between systems.
Proof of memory at work
Mem0’s memory layer is already deployed across applications ranging from education to recovery support and agent automation. Platforms such as Sunflower Sober, OpenNote, RevisionDojo, and BrowserUse report faster integration, improved personalization, and cost reductions of up to 40 percent, showing how persistent memory can reduce redundant prompts and extend context across sessions. “What we’re seeing across customers is developers changing how they think,” Singh said. “It used to be ‘How do I learn about this user?’ Now it’s ‘How do I use what I already know?’ That’s the shift we’re enabling.”
The AWS partnership gives Mem0 access to enterprise-grade infrastructure while keeping its model-agnostic stance intact. Through integrations with AWS Bedrock, OpenSearch, and Neptune, developers can now deploy agents with persistent memory inside secure, production-scale environments. The collaboration also aligns Mem0 with enterprise teams looking for governance and reliability without vendor lock-in.
The market for AI memory is quickly expanding, with players such as Zep, Letta, and Supermemory pursuing their own architectures. Mem0 differentiates itself by presenting memory as foundational infrastructure: a universal layer that underpins, rather than competes with, large models and frameworks. Its funding signals growing investor belief that this layer will become central to the agentic AI stack.
A foundation for the next wave of AI
For Singh, the mission remains straightforward: “Intelligence needs memory.” The new funding enables Mem0 to pursue that mission at global scale, strengthening its neutral architecture and preparing for portability across systems and platforms.
If successful, the company’s work could redefine how AI applications are built, moving from reactive models to systems that learn, adapt, and remember across time.
