Engram Raises $98 Million at $600 Million Valuation to Tackle AI Memory Costs

U.S. startup Engram, co-founded by Israeli researcher Dr. Dan Biederman, has raised $98 million at a $600 million valuation. The round was led by General Catalyst, Kleiner Perkins and Sequoia, with participation from prominent Israeli tech investors including Assaf Rappaport, Merav Bahat, Tomer Schwartz, Ofir Arlich, Roy Tiger and Ilan Twigg.

Engram is focused on one of generative AI’s biggest technical problems, online continual learning, letting models learn and update in real time without catastrophic forgetting and without expensive retraining from scratch. Its architecture separates a model’s reasoning layer from its memory layer, allowing enterprise and personal systems to adapt to user preferences, conversation history and new streaming data within seconds or hours.

Biederman said the company was founded in October last year directly out of Stanford University’s AI lab, where he worked with Professor Chris Re, and that it is advised by senior AI figures including Alan Carthy, who moved to Anthropic. He said Engram sits between a model company and an infrastructure company, and aims to solve the memory problem of AI systems. He added that current systems must save information externally and retrieve it later, which becomes costly in complex software projects because agents repeatedly search through folders and consume large numbers of tokens.

According to Biederman, the company already has contracts with major customers and is on track to generate significant revenue. He said the rapid rise in token usage this year, especially after newer models such as Claude increased consumption, turned cost control into a critical issue for organizations. Engram has also formed partnerships with Microsoft, Notion and others to reduce AI operating expenses, with Biederman saying some customers can cut costs by 10 to 100 times. Despite the new valuation, the company has only 13 employees, including his wife Natali, its first hire.