Graphon Labs, which develops memory infrastructure for artificial intelligence agents, has closed a Series A funding round worth $35m led by RRE Ventures. Additional backers include HP, Energy Impact Partners, Hitachi Ventures and Gaia VC, the company announced on Tuesday. This investment brings Graphon's cumulative funding to $43.3m, arriving less than six months after the startup emerged from stealth mode with an $8.3m seed round.
The 20-person team plans to allocate the capital primarily toward hiring engineers and researchers, with some resources directed toward sales operations. Chief executive Arbaaz Khan outlined these priorities in remarks to Business Insider, which first reported the funding announcement.
Khan uses an analogy to describe the limitations of conventional AI search mechanisms. "You've basically lost all that information because each piece is living on its own," he told Business Insider, comparing the approach to tearing pages from a book and searching through scattered fragments.
Graph-based architecture replaces fragmented approaches
Artificial intelligence agents can execute code and perform complex tasks, yet they frequently struggle to maintain continuity across past interactions and evolving organizational information. Graphon contends that standard techniques, including retrieval-augmented generation (RAG), which supplies models with text excerpts from storage, fail to preserve context, interconnections and temporal information as datasets expand.
The company's technology relies on graphons, a mathematical framework designed for extremely large-scale graphs. During their doctoral research, Khan and co-founder Clark Zhang created graph-based memory systems for robotic applications, Khan explained in a blog post. Graphon's software constructs a map of connections across an organization's documents, videos and images, capturing information provenance and modification history.
"You can basically point your Claude or GPT or whatever to Graphon," Khan stated to Business Insider, describing the system's compatibility with leading language models.
Graphon demonstrated strong performance across multiple evaluation frameworks. On LoCoMo, which measures memory retention during extended conversations, the system achieved 93.9% accuracy while consuming 58% fewer tokens than approaches that load entire conversation histories into models. On ATM-Bench, a University of Cambridge benchmark spanning video, audio, images and text, Graphon scored 81.7%, positioning it 5 points ahead of the next-ranked system on the public leaderboard.
Enterprise adoption and product roadmap
The San Francisco 49ers, Hitachi and other major enterprises have deployed Graphon's platform. HP also uses the technology, according to Business Insider. Costa Kladianos, the 49ers' executive vice president of technology, indicated the organization is investigating applications for the system, with sponsorship opportunity identification among the potential use cases. Hitachi Rail reported that its assessment of the technology "shows promise" in industrial contexts.
Graphon currently offers three distinct products: Graphon Enterprise, a developer platform, and on-device memory solutions for smartphones, laptops, cameras and wearable devices. Khan indicated to Business Insider that the company will maintain its focus on enterprise customers before expanding into device and consumer-facing markets.
Major artificial intelligence firms are simultaneously integrating memory capabilities into their own assistants. Anthropic unified Claude's memory across chat and Cowork in August. Apple introduced memory functionality for Siri in iOS 27.2 during October.
Source: The Next Web
