After enterprises大规模 deploy generative AI, model invocation costs are rising rapidly. Engram, founded just eight months ago, has announced a $98 million funding round to reduce token consumption and improve response efficiency through a suite of enterprise-focused "memory layer" products.
Financing and Client List
This round of funding was led by General Catalyst, Kleiner Perkins, and Sequoia, with OpenAI co-founder Andrej Karpathy also among the investors. Engram stated that the funds will be primarily allocated toward computing power investment and talent acquisition.
The company currently has only 13 employees but has already secured clients such as Microsoft, Notion, and the legal AI startup Harvey. For a relatively new startup, this client list indicates that its product has entered the radar of enterprise trials and procurement.
Focus on reducing token expenses
Engram positions itself as the "learning memory" of AI. Its core idea is to enable models to retain internal workflows, context, and historical information, allowing faster access to relevant content in subsequent queries and reducing costs associated with repeated reasoning and lengthy contexts.
The company states that, for certain tasks, its model can achieve performance equal to or better than state-of-the-art laboratory models using up to 100 times fewer tokens. Tokens are the fundamental billing units in AI query processing—the more tokens used, the higher the typical cost for businesses.
Address the model's memory limitations
Co-founder and CEO Dan Biderman believes that many current models' issues lie not in insufficient understanding, but in the lack of stable memory. The longer the context, the more models are slowed down by additional information, and enterprises must pay higher costs for increased retrieval, reading, and reasoning.
He stated that Engram does not claim to outperform OpenAI or Anthropic’s models across all capabilities, but rather emphasizes its efficiency advantages in specialized scenarios. This funding also demonstrates that competition among AI startups is shifting from merely pursuing stronger models to focusing on cost control and enterprise deployment capabilities.