EnCharge AI bets analog in-memory chips run 200 TOPS AI on 8W laptops

Designs charge-based analog in-memory AI accelerators (EN100) that run inference on laptops, workstations, and edge devices at a fraction of a GPU's power draw.

Analog in-memory computing can make on-device inference ~20x more efficient than digital chips, so laptops and robots adopt a specialist accelerator over GPUs.

$21.7M (Dec 2022); $18.6M DARPA OPTIMA grant (2024); >$100M Series B led by Tiger Global (Feb 2025) - total ~$144M.

Verma's Princeton lab researched charge-based analog in-memory computing for about a decade with Department of Defense funding; he spun out EnCharge AI in 2022 with ex-IBM CTO Kailash Gopalakrishnan and ex-Macom COO Echere Iroaga, raising $21.7M to build a complete hardware-plus-software stack.

Feb 2025: oversubscribed >$100M Series B led by Tiger Global, with Samsung Ventures, Foxconn-linked HH-CTBC, RTX Ventures, and In-Q-Tel; first product EN100 (M.2 module: 200+ TOPS at 8.25W, 32GB LPDDR; 4-chip PCIe card: ~1 petaOPS at 40W) unveiled, with trade-press coverage into 2026; targeting laptops, workstations, robotics, and industrial automation.

EN100 launched and in commercialization; the company remains private, with no revenue or customer names disclosed.

Analog in-memory computing attacks the biggest cost in AI inference, moving data between memory and compute, rather than just scaling digital chips.

A decade of Princeton research and DoD funding de-risked the technology before venture money changed the company's agenda.

The investor list spanning Samsung, Foxconn, RTX, and In-Q-Tel signals demand for efficient on-device AI across consumer electronics, manufacturing, and defense.

The bet is contrarian: if efficiency wins, edge devices displace cloud GPUs for everyday inference; if digital roadmaps keep improving, the window closes.

De-risking fundamental tech for a decade before venture money gave EnCharge patents and a contrarian story, but analog inference still must win OEM sockets against Nvidia's roadmap.

As of April 2026, EnCharge has launched its first product, EN100, and is pursuing OEM adoption in laptops, workstations, and edge/industrial systems; it remains a private company of roughly 90 people in Santa Clara. No revenue or named customers have been disclosed, and the claim that analog inference will displace digital accelerators is still unproven at volume.