Our vision

PIM solves data movement

Processing-in-memory (PIM) is the solution to reach the best computing  efficiency at system level for big data apps, because it solves the data movement dominant problem by offloading the computations to processing units sitting next to the data, that is the memory chips. Leveraging massively available memory production capacities and current server architecture have massive benefits in terms of costs and scalability. Programmability ensures lasting performance and flexibility adapted to the acceleration of applicative innovations.

 Our mission

Build a great PIM solution

Our mission is to develop the first processing-in-memory solution that is fully programmable,  scalable, and accelerate drastically data-intensive applications without requiring any hardware architecture changes. Not only our solution delivers game-changing acceleration for data intensive applications, but is also much more energy efficient as it drastically limits data movement in the processing board.

 Understand value and applications


« UPMEM PIM is a low-cost and general purpose computing solution that already demonstrated a 20x acceleration factor for multiple applications including genomics, analytics, search, database scan, AI…. Those benchmarks demonstrate that this drastic improvement can be done at very limited application porting costs. PIM will become a core component of next generation servers and computing systems.

Join our team

UPMEM team is at the core of our success and development. We are looking for talent for hardware design and implementation, PIM software tools, and software development on big data verticals (Analytics, AI, genomics, crypto…)     

Everybody talks about UPMEM and PIM