EnCharge, an AI chip startup, claims its novel analog accelerators can rival desktop GPUs while consuming a fraction of the power. The company’s in-memory compute architecture is said to offer a remarkable 20x performance-per-watt advantage over traditional digital accelerators. Developed from research at Princeton, the AI inference chips promise 150 TOPS at 8-bit precision on just one watt. With $100 million in Series B funding, EnCharge plans to manufacture chips for various applications like mobile and PC, aiming to significantly improve AI compute efficiency and energy consumption in real-world applications.
EnCharge's analog AI chips aim to achieve a 20x performance-per-watt advantage over traditional GPUs, potentially revolutionizing AI inferencing efficiency.
With a focus on in-memory compute architecture, EnCharge claims their chips deliver 150 TOPS of AI compute at only 1 watt, conserving power significantly.
CEO Naveen Verma highlights their innovative technology origins, stating it addresses fundamental challenges in AI compute, aspiring to reshape the desktop GPU landscape.
The startup has secured $100 million in funding to manufacture production chips, indicating confidence in their technology and market potential.
Collection
[
|
...
]