AI Memory Startup Engram Raises $98M to Cut Token Costs
Engram enters the AI infrastructure market with $98M in funding, targeting the surging cost of token usage in large language models.
A new entrant in the artificial intelligence infrastructure space, Engram, has secured $98 million in funding to tackle one of the industry's most pressing and underreported challenges: the escalating cost of token processing. As AI models grow more capable and more expensive to run, the expense of feeding vast amounts of context into those models has become a genuine bottleneck for enterprises looking to scale intelligent applications.
Token costs sit at the heart of how large language models are priced and consumed. Every word, code snippet, or document passed to a model consumes tokens, and as organizations push AI deeper into their workflows, those costs compound rapidly. Engram's focus on memory — essentially, smarter ways to store, retrieve, and manage context so that models don't have to reprocess the same information repeatedly — represents a bet that efficiency infrastructure will be as valuable as the models themselves.
Read more Qualcomm Acquires AI Startup Modular to Strengthen Data Center Push →
The timing is notable. The AI industry is experiencing a paradox: models are becoming more powerful, but that power comes at a steeper computational price. Providers have raised or restructured pricing on frontier models, and enterprise buyers are increasingly scrutinizing their AI spending. Startups that can demonstrably reduce the token overhead of running AI at scale are positioned to capture significant budget that would otherwise flow directly to model providers.
Engram's raise signals that investors see memory and context management as a distinct and fundable layer in the AI stack — not merely a feature to be absorbed by larger platforms, but a specialized discipline worthy of dedicated infrastructure. If the analogy to traditional computing holds, just as caching and database optimization became critical disciplines as software scaled, AI memory optimization may follow a similar trajectory toward indispensability.
Continue reading at US Top News and Analysis
