San Francisco-based Graphon Labs has raised $35 million in Series A funding led by RRE Ventures, with participation from HP, Energy Impact Partners, Hitachi Ventures, and Gaia VC. The AI startup is developing Persistent Intelligence, an infrastructure designed to help artificial intelligence systems continuously access, understand, remember, and build upon information from past interactions and experiences.
The latest funding round comes less than six months after Graphon Labs emerged from stealth with $8.3 million in seed funding, taking its total funding to $43.3 million.
The company aims to address one of the fundamental limitations of today’s AI systems: their inability to retain and effectively use knowledge over extended periods. Through its Persistent Intelligence architecture, Graphon Labs wants to provide the foundational infrastructure for the next generation of AI agents and accelerate progress toward Artificial General Intelligence (AGI).
The company has already deployed its technology with major organizations, including the San Francisco 49ers and Hitachi.
Graphon Labs Is Building Persistent Intelligence for the Next Generation of AI Agents
As artificial intelligence evolves from systems that primarily answer questions to autonomous agents capable of performing tasks continuously, long-term memory and contextual understanding are becoming increasingly important.
While modern AI models can reason and AI agents can take action, they often struggle to retain information from previous interactions and build upon past experiences.
Graphon Labs is addressing this challenge through what it calls Persistent Intelligence. The technology enables AI systems to continuously access, understand, remember, and build upon previously accumulated information.
The company’s vision is to create an infrastructure where intelligence can accumulate over time rather than repeatedly starting from limited or fragmented context.
This capability is particularly important for AI agents expected to operate around the clock, where remembering previous actions, understanding changes, and learning from past outcomes are essential.
Raju Rishi, General Partner at RRE Ventures, highlighted the importance of this approach.
“Models can reason, and agents can act, but neither gets better if they cannot build off of prior knowledge,” Rishi said.
He added that Graphon Labs enables intelligence to accumulate, potentially improving effectiveness, reducing compute time and complexity, and becoming the underlying source of truth for AI agents.
Why Graphon Labs Is Taking a Different Approach to AI Memory
Major AI research laboratories have invested billions of dollars in developing increasingly powerful models. However, even advanced AI models and agents continue to face limitations when it comes to retaining information and using historical context effectively.
Existing approaches, including Retrieval-Augmented Generation (RAG), GraphRAG, context windows, and traditional retrieval architectures, help AI systems retrieve relevant information.
However, according to Graphon Labs, these approaches can compress, fragment, or lose important contextual details, relationships, historical information, and temporal understanding as data grows and changes.
Instead of simply improving existing retrieval systems, Graphon Labs started with a different question: What infrastructure does intelligence need to continuously learn from everything that came before?
Its answer is based on graphons, a mathematical framework for analyzing very large graphs that the company’s founders developed during their PhD research.
The technology rethinks the entire AI information infrastructure stack, covering storage, search, retrieval, reasoning, and memory.
Rather than treating information as disconnected pieces of data, Graphon Labs organizes it into connected relationships while preserving what an AI agent knows, what actions it has taken, what has changed, and what happened as a result.
The architecture also preserves information sources and timing, enabling AI systems to maintain a more complete understanding of their accumulated knowledge.
According to the company, this approach can also reduce the computing time required to represent and reason across enormous volumes of information.
Graphon Labs Achieves Leading Results on Three AI Memory Benchmarks
Graphon Labs has reported the highest accuracy among evaluated systems across three prominent benchmarks designed to assess long-term context and memory capabilities: LoCoMo, ConvoMem, and ATM-Bench.
The company highlighted the following results:
- LoCoMo: Achieved 93.9% accuracy while using 58% fewer tokens compared with loading the full conversation into the AI model.
- ConvoMem: Recorded 90.2% accuracy while using 95% fewer tokens.
- ATM-Bench: Achieved 81.7% accuracy, outperforming the next system on the public leaderboard by 5 percentage points.
ATM-Bench, developed by researchers at the University of Cambridge, evaluates question-answering capabilities across multiple data formats, including video, audio, images, and text.
The LoCoMo and ConvoMem comparisons are based on baseline results published by Databricks.
These benchmark results demonstrate Graphon Labs’ focus on improving how AI systems retain, retrieve, and reason over information without requiring entire historical conversations or datasets to be loaded into model context windows.
San Francisco 49ers and Hitachi Explore Graphon Labs Technology
Beyond benchmark performance, Graphon Labs is positioning its Persistent Intelligence technology for practical enterprise applications across industries.
The company has already been deployed with the San Francisco 49ers, Hitachi, and other large enterprises.
For the San Francisco 49ers, the technology presents opportunities to improve business operations, identify potential sponsorship opportunities, and empower internal teams with better access to information.
Costa Kladianos, Executive Vice President of Technology for the San Francisco 49ers, said the organization is always looking for innovative ways to strengthen its business and drive growth.
He expressed enthusiasm about exploring how Graphon Labs’ technology could help the organization identify new opportunities and gain a competitive advantage.
Hitachi Rail is also evaluating the technology for its potential to improve access to complex and distributed information.
Amaury Jourdan, Deputy Manager and CTO for Digital and AI Activities at Hitachi Rail, explained that the organization operates in environments where information is distributed across complex systems and operational outcomes can be mission-critical.
He noted that Graphon Labs’ technology shows promise in facilitating contextual knowledge discovery, automating connections to data sources, and improving AI inference results in specific industrial environments.
Graphon Labs Plans Three Products for Enterprise, Developers, and Devices
With its latest $35 million Series A funding, Graphon Labs plans to expand its Persistent Intelligence infrastructure beyond existing enterprise deployments.
The company is building a unified intelligence architecture designed to scale from large enterprises to individual users and connected devices.
Its vision is to create an underlying intelligence layer that allows AI systems to know, reason, act, remember, and learn continuously.
This architecture will power three primary products.
1. Graphon Enterprise
Graphon Enterprise provides secure and governed memory across enterprise data, including different data modalities.
It enables AI agents and enterprise teams to search and reason over organizational information while preserving context and relationships.
The solution supports multiple deployment options, including hosted infrastructure, Graphon-managed Virtual Private Cloud (VPC), customer-managed VPC, and on-premises environments.
2. Graphon Developer
Graphon Developer is designed for developers building AI-powered applications and agents.
It allows developers to build over different types of data and modalities without having to recreate the underlying context and memory infrastructure themselves.
The goal is to simplify the development of AI applications that require persistent memory and contextual understanding.
3. Graphon On-Device Memory
Graphon On-Device Memory aims to bring private, persistent, multimodal intelligence directly to personal and connected devices.
The product is being developed for devices such as smartphones, laptops, cameras, wearables, and other hardware.
This approach could extend Persistent Intelligence beyond centralized enterprise systems and make it available closer to individual users and their everyday devices.
How Graphon Labs Plans to Use Its $35 Million Series A Funding
The latest funding will accelerate Graphon Labs’ research and infrastructure development while supporting the expansion of its technology across different markets.
The company plans to focus on three broad priorities:
- Scaling Enterprise Deployments: Expanding its infrastructure and supporting broader adoption among large organizations.
- Opening Its Infrastructure to Developers: Making Persistent Intelligence accessible to developers who want to build the next generation of AI agents and applications.
- Accelerating Research and On-Device AI: Advancing its research while bringing persistent, multimodal intelligence to personal and connected devices.
The funding represents an important step in the company’s ambition to make Persistent Intelligence available beyond enterprise customers and eventually accessible to everyone.
Commenting on the company’s long-term vision, Arbaaz Khan, Founder and CEO of Graphon Labs, said the latest investment would help the company move from early experimentation to scaled enterprise deployments, expand access for developers, accelerate research, and bring Persistent Intelligence to devices.
He emphasized that the broader objective is to build infrastructure that allows intelligence to accumulate knowledge continuously, improve faster, and eventually make superintelligence more widely accessible.
Graphon Labs’ Vision: Building the Infrastructure for Artificial General Intelligence
The development of increasingly capable AI models has brought artificial intelligence closer to performing complex tasks autonomously. However, the ability to remember previous experiences and continuously learn from them remains an important challenge.
Graphon Labs believes that persistent memory is a critical missing layer in the evolution of AI agents.
Its approach focuses on creating an architecture that preserves knowledge, relationships, information sources, and temporal context across different types of data.
As AI agents become more autonomous and operate continuously, the ability to retain historical information and learn from previous actions could become essential.
With its $43.3 million in total funding, deployments involving organizations such as the San Francisco 49ers and Hitachi, and its three-product strategy, Graphon Labs is positioning itself in the emerging market for persistent AI memory infrastructure.
The company’s long-term ambition is to make Persistent Intelligence a foundational layer beneath AI agents, supporting the next stage of artificial intelligence development and progress toward AGI.
About Graphon Labs
Graphon Labs is an applied AI research and technology company building Persistent Intelligence to power the next generation of AI agents and accelerate progress toward Artificial General Intelligence.
Its technology enables AI systems to continuously access, understand, remember, and build upon previously accumulated information.
Built on graphons, the company’s architecture is designed to preserve knowledge, relationships, information sources, and temporal context across different data types.
Graphon Labs offers three primary products: Graphon Enterprise, Graphon Developer, and Graphon On-Device Memory.
The company was founded by senior engineers and researchers from organizations including Amazon, Meta, Google, and Apple, as well as NVIDIA and other leading research institutions.
For more information, visit graphonlabs.com.