Graphon Labs, a San Francisco-based applied artificial intelligence research company developing persistent memory infrastructure for AI agents, has raised $35 million in Series A funding led by RRE Ventures. The financing included participation from HP, Energy Impact Partners, Hitachi Ventures, and Gaia VC and will support the company’s expansion of its Persistent Intelligence technology across enterprise systems, developer platforms, and consumer devices.
The announcement comes less than six months after Graphon Labs emerged from stealth with $8.3 million in seed funding, bringing its disclosed financing to $43.3 million.
Graphon Labs is developing an AI infrastructure architecture designed to help autonomous agents retain information, understand relationships, and build upon knowledge accumulated through previous interactions.
The company believes persistent memory represents a critical requirement for the next generation of AI agents, particularly systems expected to operate continuously and perform complex tasks over extended periods.
Traditional AI models generally process information within limited context windows, creating challenges when applications require access to extensive histories, changing information, or relationships spanning multiple sources.
Current retrieval technologies, including retrieval-augmented generation and graph-based retrieval systems, can help models access external information.
However, Graphon Labs argues that these approaches can lose important context, relationships, and historical information as datasets expand.
Its Persistent Intelligence architecture is designed to address these limitations through a unified approach to storage, search, retrieval, reasoning, and memory.
Developing A New Architecture For AI Memory
Graphon Labs’ technology is built on graphons, a mathematical framework used to analyze large graphs.
The company’s founders developed expertise in this mathematical approach during their doctoral research.
Graphon Labs applies the framework to AI memory infrastructure, organizing information in ways that preserve relationships between data points.
Rather than treating information as isolated documents or fragments, the architecture is designed to connect knowledge across multiple sources and time periods.
This allows AI agents to retain information about previous actions, changes in their operating environment, and the outcomes of earlier decisions.
The platform also preserves information about sources and timing, which can be important for understanding how knowledge evolves.
Graphon Labs believes this architecture can improve the efficiency of AI reasoning while reducing the computing resources needed to process extensive information histories.
The company is targeting applications where autonomous systems must continuously access and update large volumes of information.
These include enterprise AI agents, industrial systems, and connected devices.
Benchmark Results Demonstrate Memory Performance
Graphon Labs reported strong results across three benchmarks designed to evaluate long-term context and memory capabilities.
The company achieved the highest accuracy among the systems it evaluated on LoCoMo, ConvoMem, and ATM-Bench.
On the LoCoMo benchmark, Graphon achieved 93.9% accuracy while using 58% fewer tokens than an approach that loads the complete conversation history into a model.
On ConvoMem, the system achieved 90.2% accuracy while using 95% fewer tokens.
The company also reported 81.7% accuracy on ATM-Bench, a benchmark developed by University of Cambridge researchers to evaluate question answering across video, audio, images, and text.
That result was 5 percentage points above the next system on the public leaderboard at the time of the announcement.
The LoCoMo and ConvoMem comparisons used baseline results published by Databricks.
These results suggest that Graphon’s approach may improve the ability of AI systems to retrieve relevant historical information without requiring entire datasets to be repeatedly processed.
However, benchmark performance does not necessarily establish equivalent results across every production environment or application.
Graphon Labs intends to continue developing its infrastructure as customers deploy persistent memory capabilities across more complex systems.
Enterprise Applications And Commercial Interest
The company’s technology is being evaluated for applications involving large, distributed information environments.
The San Francisco 49ers are exploring how Graphon Labs’ technology could support business development and operational activities.
Potential use cases include identifying sponsorship opportunities and helping internal teams access useful information.
Hitachi Rail is also evaluating the technology for applications involving complex industrial information.
The organization is interested in improving knowledge discovery, connecting information sources, and strengthening AI inference across operational systems.
These examples demonstrate the potential relevance of persistent memory technology beyond conventional enterprise software applications.
Organizations managing large volumes of operational data often need AI systems to interpret historical information, identify relationships, and retrieve relevant details from different sources.
Graphon Labs aims to provide infrastructure supporting these requirements while maintaining access controls and deployment flexibility.
Three Products Supporting Enterprise, Developers, And Devices
Graphon Labs is developing three products based on its Persistent Intelligence architecture.
Graphon Enterprise provides governed AI memory capabilities across different enterprise data types, including text, images, audio, and other information formats.
The platform supports hosted deployments, Graphon-managed virtual private clouds, customer-managed virtual private clouds, and on-premises infrastructure.
These deployment options are designed to accommodate organizations with different security, governance, and infrastructure requirements.
Graphon Developer provides tools for developers building AI applications that require persistent context and memory across multiple data sources.
The offering is intended to reduce the need for developers to construct their own memory and retrieval infrastructure.
Graphon On-Device Memory is designed to bring private, persistent multimodal intelligence to smartphones, laptops, cameras, wearable devices, and other connected products.
The company believes local memory capabilities could support more personalized AI experiences while allowing information to remain on individual devices.
Using New Capital To Accelerate Development
Graphon Labs plans to use the Series A financing to expand enterprise deployments, make its infrastructure available to more developers, and accelerate research into persistent intelligence.
The company also intends to advance its on-device memory technology.
Its founding team includes researchers and engineers with experience at Amazon, Meta, Google, Apple, NVIDIA, and other technology and research organizations.
Investors believe persistent memory could become an important infrastructure layer as AI agents move beyond individual interactions toward continuous operation.
For businesses, the ability to preserve context may improve the usefulness of AI agents that manage ongoing workflows and interact with changing datasets.
Graphon Labs sees these capabilities as part of a longer-term effort to advance artificial general intelligence.
With $35 million in new financing, the company is expanding its research and commercial development while pursuing a broader architecture designed to help AI systems remember, reason, and accumulate knowledge over time.
KEY QUOTES
“Models can reason, and agents can act, but neither gets better if they cannot build off of prior knowledge. Graphon Labs lets intelligence accumulate which can have profound impacts upon efficacy, compute time, complexity of queries, and much much more. We see it becoming the source of truth beneath every agent, and that’s why we led this round.”
Raju Rishi, General Partner At RRE Ventures
“The world is heading toward agents and physical systems that create and depend on more information than anyone has ever had to manage. In a world where we are moving to AI using computers, we need to fundamentally rethink how computers access data to pre-fill their context windows. The old way, using vector search, is like searching through the pages of a torn-up book. We take those pages and turn them into a connected map, unlocking seamless memory search.”
“This funding takes us from early experimentation to scaled enterprise deployments, opens our infrastructure to developers, accelerates our research, and brings Persistent Intelligence on-device. Those are steps toward a bigger goal. When intelligence becomes lossless and continuous, it compounds, it improves faster, and eventually superintelligence becomes available to everyone. We’ll measure our success by whether we build the infrastructure that gets intelligence to its next stage and make it available to all.”
Arbaaz Khan, Founder And CEO Of Graphon Labs
“At the 49ers, we’re always looking for innovative ways to strengthen our business and drive growth. From uncovering new sponsorship opportunities to empowering our teams, we’re excited to explore how Graphon Labs’ technology can help us gain a competitive edge.”
Costa Kladianos, Executive Vice President Of Technology For The San Francisco 49ers
“At Hitachi Rail, we seek to unlock greater value from complex and distributed information landscapes, often with mission-critical operational outcomes. Our evaluation of Graphon Labs’ innovative technologies shows promise to facilitate the discovery of context knowledge, automate the connection to data sources and improve AI inferencing results in our specific industrial context.”
Amaury Jourdan, Deputy Manager And CTO For Digital And AI Activities At Hitachi Rail
