← All companies & organizations

Chips / Company profile

EnCharge AI

What EnCharge AI does

EnCharge AI is a private semiconductor company developing analog in-memory computing (AIMC) hardware and a software stack for efficient AI inference, targeting deployments across edge-to-cloud rather than only cloud data centers. The company positions its approach around performing computation “in-memory” using analog charge-domain operations, aiming to reduce the energy and performance penalties associated with moving weights and intermediate data between separate memory and compute units. EnCharge AI’s product strategy (as described publicly) is to integrate its technology into multiple system form factors—ranging from chiplets to PCIe cards and other partner configurations—so that AI workloads can run with lower power, cost, and CO2 footprint while still meeting latency and deployability needs for constrained environments.

News

Company record

Source map · 2 recurring channels · 43 references
EnCharge AIOfficial Site · monitor↗MTS Frontier / Control RoomCurated Feed · monitor↗Official company sourceOfficial Reference · reference↗sec.govRegulatory · reference↗aibmag.comReporting · reference↗anzupartners.comReporting · reference↗app.daily.devReporting · reference↗arxiv.orgReporting · reference↗assets.ctfassets.netReporting · reference↗axios.comReporting · reference↗boozallen.comReporting · reference↗cbinsights.comReporting · reference↗crn.comReporting · reference↗crunchbase.comReporting · reference↗eetimes.comReporting · reference↗en.wikipedia.orgReporting · reference↗encharge.ioReporting · reference↗energizerpower.comReporting · reference↗enphase.comReporting · reference↗enpower-energy.comReporting · reference↗enpower-greentech.cnReporting · reference↗gomactech.netReporting · reference↗hkexnews.hkReporting · reference↗img.electronicdesign.comReporting · reference↗invertersupply.comReporting · reference↗irp.cdn-website.comReporting · reference↗linkedin.comReporting · reference↗morgancreekcap.comReporting · reference↗netwealth.com.auReporting · reference↗parolaanalytics.comReporting · reference↗polyn.aiReporting · reference↗privatemarketview.comReporting · reference↗publish.illinois.eduReporting · reference↗qualcomm.comReporting · reference↗reddit.comReporting · reference↗siliconangle.comReporting · reference↗spectrum.ieee.orgReporting · reference↗streetinsider.comReporting · reference↗talk-about.aiReporting · reference↗techcrunch.comReporting · reference↗techtimes.comReporting · reference↗telehilladvisors.comReporting · reference↗trendhunter.comReporting · reference↗websets.exa.aiReporting · reference↗williamblair.comReporting · reference↗

Still resolving: Newsroom · Official X

Funding

Latest disclosed valuationNot disclosed
Tracked capital$144.3M3 sourced rounds

What it builds

EN100

EN100 is EnCharge AI’s first commercial AI accelerator, described as an analog in-memory computing chip designed for on-device AI inference on laptops, workstations, and edge devices.

source ↗

Milestones & partnerships

May 29, 2025
EN100 announced as first commercial accelerator for on-device AI

EnCharge AI publicly announced its EN100 accelerator for laptop/workstation/edge on-device computing use cases, describing 200+ TOPS within client constraints.

↗
Feb 13, 2025
Series B announced (commercialization support)

EnCharge AI announced an oversubscribed Series B led by Tiger Global, describing commercialization of first client computing-focused AI accelerator products and progress along a product roadmap.

↗
Dec 14, 2022
Series A announced

EnCharge AI announced raising $21.7M in Series A financing led by Anzu Partners to further develop and commercialize its semiconductor hardware and software stack.

↗
Date not disclosed
Technology characterization: analog in-memory computing using capacitor-based charge-domain operations

EnCharge AI describes its approach as analog in-memory computing based on intrinsically-precise metal capacitors, positioned to overcome signal-to-noise ratio limitations associated with analog processing.

↗
Date not disclosed
Deployment form-factor roadmap across edge-to-cloud

The company states its core technology integrates into diverse form factors, describing deployment options spanning chiplets, ASICs, and standard form-factor PCIe cards, with hardware and software orchestration between on-device and cloud deployments.

↗
Date not disclosed
“Virtualized In-Memory Compute™” positioning for reducing data movement in AI inference

EnCharge AI describes a concept called Virtualized In-Memory Compute™ (VIMC) as an abstraction that enables compute-in-memory across the memory stack, with the goal of minimizing data movement relative to conventional architectures.

↗