Graphon Labs Raises $35M Series A to Scale Persistent AI Memory

Graphon Labs closed a $35 million Series A on November 10, 2026, bringing its total funding to $43.3 million, according to the company's release. RRE Ventures led the round, with HP, Energy Impact Partners, Hitachi Ventures and Gaia VC participating. The money funds one bet: persistent-memory infrastructure for AI agents across enterprise, developer and on-device products.

Chief executive Arbaaz Khan said the capital moves Graphon from early experimentation to scaled enterprise deployments, opens its infrastructure to developers, accelerates research and supports on-device memory.

What does Graphon's memory layer actually change?

The company calls its approach Persistent Intelligence — AI that can continuously access, understand, remember and build on earlier information. Its argument against the current stack is specific. Retrieval-augmented generation, GraphRAG, context windows and conventional retrieval can compress or fragment relationships, history and temporal understanding as information grows and changes, Graphon says.

Instead, the platform ingests video, images, audio and documents, automatically building an ontology — a structured representation of the data and its connections — while retaining links back to sources.

Khan splits the architecture into two pieces: a detailed graph representing information and relationships, and a compressed graphon representation that guides retrieval through that larger structure. Graphons are a mathematical framework for very large graphs. The founders developed the underlying research during their PhDs at the University of Pennsylvania's GRASP Lab.

The engineering goals are concrete: incorporate new information, trace evidence back to its origin, and operate at practical cost and speed.

Why developers don't need the math

Here's the practical hook for builders. Developers work with S3-style storage and search concepts — buckets, objects, files, queries — without touching the mathematics underneath. The interface supports keyword, semantic and image searches, plus agent-assisted questions across multiple sources.

That means a simple lookup and a deeper investigation draw on the same platform. You don't assemble a separate retrieval stack for each kind of question.

The benchmark claim

Graphon reports 81.7% accuracy on ATM-Bench, a University of Cambridge benchmark for question answering across video, audio, images and text. That result sits 5.0 percentage points ahead of the next system on the public leaderboard, according to the company.

Three products, three deployment models

The architecture supports:

• Graphon Enterprise — governed memory across enterprise data; hosted by Graphon, deployed in a Graphon-managed or customer-managed virtual private cloud, or installed on-premises.
• Graphon Developer — for builders working across data types without recreating the context stack.
• Graphon On-Device Memory — targeting phones, laptops, cameras, wearables and other devices.

Who's actually using it?

Adoption among named customers varies in depth. The release broadly describes deployments with the San Francisco 49ers, Hitachi and other large enterprises.

Costa Kladianos, a technology executive with the 49ers, said the organization is exploring applications such as uncovering sponsorship opportunities and supporting internal teams. Hitachi Rail's Amaury Jourdan described an evaluation showing promise for contextual-knowledge discovery, automated data-source connections and improved AI inference in its industrial setting.

The seed announcement from May 14, 2026 — an $8.3 million round led by Arvind Gupta of Novera Ventures, when Graphon emerged from stealth — named South Korean conglomerate GS as an early enterprise customer. A GS representative cited applications including analysis of customer movement in convenience stores and CCTV-based construction-safety work.

The quote that frames the round

RRE general partner Raju Rishi said the firm sees Graphon becoming a source of truth beneath agents. His explanation: "Models can reason, and agents can act, but neither gets better if they cannot build off of prior knowledge."

What happens next

The company profile lists July 2025 as its founding date, with Khan, chief technology officer Clark Zhang and chief operating officer Deepak Mishra as founders. The San Francisco company is just over a year old.

The Series A funds a commercialization test as much as a research program. Graphon must turn exploratory enterprise work into scaled deployments, persuade outside developers to build on its memory layer, and adapt the same architecture for private on-device use. The $35 million buys time for all three.