EnCharge AI

EnCharge AI designs AI inference hardware built on analog in memory computing. Its silicon performs computation in the charge domain using metal capacitors, an approach the company says overcomes the signal to noise limits that held back earlier analog designs while removing most of the data movement that dominates power consumption in GPUs and digital accelerators. The output is sold as components others build with: chiplets, ASICs and standard form factor PCIe cards, together with software for orchestrating work between on device and cloud deployment. It is aimed at device makers, system builders and businesses that want to run current models where power, space and cost rule out a GPU, from mobile devices and local servers through to the data center. The company publishes measured silicon comparisons against leading products claiming 20 times higher efficiency in operations per watt and 9 times higher compute density per square millimeter, along with 10 times lower total cost of ownership per inference and 100 times lower CO2 emissions than cloud inference. It states that the scalability and maturity of the technology have been demonstrated across five design generations and multiple process nodes, and that the platform supports several model types and operators and fits existing workflows. No pricing is published. The site has no pricing page and no product catalog with amounts, and evaluation begins through a contact form. EnCharge AI was founded in 2022 and works from Santa Clara, California and Bengaluru, India. Its core technology came out of research at Princeton University, where chief executive Naveen Verma has been a professor since 2009 and spent six years leading the work before co founding the company. Chief technology officer Kailash Gopalakrishnan previously led worldwide AI hardware efforts as a Fellow at IBM.