Founded 2017 in Milpitas, CA MemMachine: open-source AI memory layer LoCoMo recall score 84.87% ~$43.5M raised: Intel · Cisco · NetApp · SK hynix · Lightspeed Micron demo: 77% higher GPU utilization ~50% lower token cost claim Works across OpenAI · Claude · Gemini · Llama · DeepSeek SC25: CXL KV-cache offload, >5x vs SSD caching Founded 2017 in Milpitas, CA MemMachine: open-source AI memory layer LoCoMo recall score 84.87% ~$43.5M raised: Intel · Cisco · NetApp · SK hynix · Lightspeed Micron demo: 77% higher GPU utilization ~50% lower token cost claim Works across OpenAI · Claude · Gemini · Llama · DeepSeek SC25: CXL KV-cache offload, >5x vs SSD caching
Company AI Infrastructure · Enterprise · Developer Tools

MemVerge

The company that spent years teaching data centers to pool their memory - and is now teaching AI agents to remember.

Founded2017
HQMilpitas, California
Team~66 people
StageSeries A · ~$43.5M
MemVerge company logo

The wordmark of a company that keeps betting on the same idea - memory belongs at the center of computing - through three hardware eras: DRAM, Optane, and now CXL.

2017
Year founded
$43.5M
Total raised
84.87%
LoCoMo recall
2x
GPU utilization
01

The Memory Company

Profile · MemVerge, Inc.
Milpitas, California

There is a certain kind of technology company whose whole existence is a bet on a single, slightly contrarian idea. MemVerge's idea is that memory - the fast, expensive, forgetful stuff that sits between the processor and the disk - is more important than everyone treats it, and that if you virtualize it properly, interesting things happen. This is not an obviously thrilling premise. It is, however, an idea that keeps turning out to be right at inconvenient times for everyone who bet against it.

The company was founded in 2017 in Milpitas by three people with unusually intertwined résumés. Charles Fan, the CEO, had earlier co-founded a startup called Rainfinity with Shuki Bruck, a Caltech professor who happened to be his PhD advisor; EMC bought it. Fan then went to VMware and started the storage business unit that built VSAN into a billion-dollar product. Bruck became MemVerge's chairman. Yue Li, the CTO, had been Bruck's post-doc. So the founding team is, essentially, an advisor, his student, and his student's student, which is either a charming academic lineage or a slightly recursive org chart, depending on your temperament.

The original product was called Memory Machine, and its trick was to virtualize different types of memory - ordinary DRAM and Intel's then-new Optane persistent memory - into a single pool that software could tier, replicate, snapshot and recover at memory speeds. MemVerge gave this a name, "Big Memory Computing," because naming a category is half the work of creating one.

When the hardware died

Here is the inconvenient part. MemVerge's early bet leaned heavily on Optane. Intel discontinued Optane. In the normal course of things, this is the sort of event that quietly ends a company - you have built a business on a chip, and the chip is now a museum piece.

MemVerge's response was to notice that its actual expertise was never really about Optane. It was about memory: tiering it, pooling it, moving it around without applications noticing. And it turned out there was a much larger customer arriving who cared about exactly that. That customer was generative AI.

"AI Memory: The Next Frontier."

— MemVerge company tagline

AI has two memory problems, and MemVerge decided to attack both. The first is hardware. GPUs are the most expensive idle asset in a modern data center; they frequently sit waiting on memory rather than computing, which is a bit like buying a fleet of Ferraris and parking them. Memory Machine's tiering, pooling and transparent checkpointing aim to keep those GPUs fed. A joint demo with Micron reported 77% higher GPU utilization and more than doubled the speed of OPT-66B batch inference. A financial-services customer reportedly doubled GPU utilization outright.

The second memory problem

The second problem is stranger and more human. Large language models have no memory. Every conversation begins from zero; the model that felt like it knew you yesterday has, today, never heard of you. MemVerge calls this "stateless amnesia," and in September 2025 it launched a product aimed squarely at curing it: MemMachine, an open-source AI memory layer.

MemMachine gives agents three kinds of memory, borrowed from how psychologists describe human memory: episodic (what happened), profile (who you are), and procedural (how to do things). It works across the major models - OpenAI, Claude, Gemini, Grok, Llama, DeepSeek, Qwen - and on any cloud or on-prem. The company reported a LoCoMo recall benchmark of 84.87% and claims roughly 50% lower token costs, on the logic that remembering something is cheaper than re-explaining it in every prompt.

MemMachine retains episodic, personal and procedural knowledge across sessions, models, agents and environments - ending the stateless amnesia of today's LLMs.

— MemVerge, on the MemMachine launch

The open-core wager

The MemMachine strategy is a familiar one in infrastructure: give the core away and sell the enterprise. The open-source project lives at memmachine.ai, with a free playground; the commercial side at memverge.ai adds scalability, security, compliance, orchestration, observability and a "White Glove" service tier reported around $2,500 a month. The open project is the front door. The business is what happens after developers walk through it.

Meanwhile the hardware thread continues. At SC25 in late 2025, MemVerge and XConn Technologies demonstrated a CXL - Compute Express Link - memory pool that offloads and shares the KV cache across GPUs and CPUs, claiming greater than 5x performance over SSD-based caching while cutting cost of ownership. This is the same company idea it started with, wearing new hardware: pool the memory, share it, and stop letting expensive silicon wait.

MemVerge is not a large company - roughly 66 people, about $43.5 million raised across a 2020 Series A backed by Intel Capital, Cisco, NetApp, SK hynix, Lightspeed and others. It competes in a crowded new category of AI-memory startups such as Mem0 and Zep, and against the vector-database stacks everyone already uses. Whether it wins is genuinely unknown. But there is something worth noticing in a company that has now been correct about the same unglamorous idea across three hardware generations, and simply kept waiting for the rest of the industry to arrive.

02

The Founders

An advisor, his student,
and his student's post-doc
CF

Charles Fan

Co-founder & CEO

Co-founded Rainfinity (acquired by EMC), started VMware's storage unit behind VSAN, and was CTO of Cheetah Mobile. 25+ years in data-center infrastructure.

YL

Yue Li

Co-founder & CTO

Former senior post-doctoral scholar in memory systems at Caltech, working under Shuki Bruck before co-founding the company.

SB

Shuki Bruck

Co-founder & Chairman

Gordon & Betty Moore Professor of Computation & Neural Systems and Electrical Engineering at Caltech; Fan's former PhD advisor.

03

What They Build

Two problems: idle GPUs,
and forgetful models
2025 Open Source

MemMachine

The open-source AI memory layer. Gives LLMs and agents persistent episodic, profile and procedural memory across sessions, models and clouds. Works with OpenAI, Claude, Gemini, Llama, DeepSeek and more.

2025 Enterprise

MemMachine Enterprise

Commercial edition adding enterprise-class scalability, security, compliance, orchestration, observability and support - plus a White Glove service tier.

2023 GPU Infrastructure

Memory Machine AI

Maximizes GPU utilization through memory tiering, pooling and transparent checkpointing for AI training and inference workloads.

2023 CXL

Memory Machine X

Harnesses Compute Express Link (CXL) memory to expand and share memory pools in big-memory data centers.

2022 Cloud

Memory Machine Batch

Runs compute-heavy batch workloads on AWS using checkpointing and spot instances to cut cloud costs - researchers report 50%+ savings.

"Big Memory software virtualizes DRAM and persistent memory so it can be replicated, tiered and recovered at memory speeds - the same insight now reshaped for AI."

By the numbers

Micron collaboration, 2024

+77%

higher GPU utilization, with 2x faster OPT-66B batch inference in the joint demo.

04

Timeline

Three hardware eras,
one idea
2017

MemVerge founded

Charles Fan, Yue Li and Shuki Bruck start the company in Milpitas to make memory-centric computing real.

2019

Big Memory takes shape

The company details software that virtualizes DRAM and Optane persistent memory into a single clustered pool.

2020

Launch & Series A

Big Memory Computing launches publicly, backed by a Series A from Intel Capital, Cisco, NetApp, SK hynix and others.

2022

Into the cloud

Memory Machine extends to cloud and batch workloads, using checkpointing and spot instances to cut costs.

2024

Micron milestone

A joint demo shows 77% higher GPU utilization and 2x faster OPT-66B inference.

2025

MemMachine launches

The open-source AI memory layer ships alongside enterprise editions; XConn partnership demos CXL KV-cache offload at SC25.

2026

Ecosystem expansion

MemVerge expands the MemMachine ecosystem and AI memory system capabilities.

05

Funding

~$43.5M total
Strategic-heavy cap table
RoundDateAmountSelected Investors
Seed / Early2019part of $43.5MGaorong Capital, LDV Partners, Northern Light VC
Series AMay 2020$24.5M (reported)Intel Capital, Cisco Investments, NetApp, SK hynix, Lightspeed, Jerusalem Venture Partners, Glory Ventures

Figures aggregated from public sources (Crunchbase, Tracxn, company disclosures). Round-level amounts are approximate.

06

Worth Knowing

The amusing details
Lineage

A recursive org chart

CTO Yue Li was chairman Shuki Bruck's post-doc; Bruck was earlier CEO Charles Fan's PhD advisor. An advisor, his student, and his student's student.

Second act

Not their first company

Fan and Bruck previously co-founded Rainfinity, which EMC acquired. MemVerge is the band getting back together.

Pivot

Optane outlived, not outbet

MemVerge's early bet leaned on Intel Optane. Intel killed Optane - so MemVerge redirected the same memory expertise toward AI and CXL.

Two domains

.ai and .ai

The company splits its identity: memverge.ai for the commercial side, memmachine.ai for the open-source project.

07

FAQ

Quick answers
What does MemVerge do?
MemVerge builds AI infrastructure software that virtualizes and optimizes memory - maximizing GPU utilization for AI workloads and, with MemMachine, giving LLMs and agents persistent memory across sessions and models.
What is MemMachine?
MemMachine is MemVerge's AI memory layer, released in 2025 as an open-source project with enterprise editions. It stores episodic, profile and procedural memory so AI agents can recall context across sessions, models and clouds.
Who founded MemVerge and when?
MemVerge was founded in 2017 by CEO Charles Fan, CTO Yue Li and chairman Shuki Bruck, with roots in memory-systems research at Caltech.
How much funding has MemVerge raised?
Roughly $43.5M total, including a 2020 Series A, from investors such as Intel Capital, Cisco Investments, NetApp, SK hynix and Lightspeed Venture Partners.
Who competes with MemVerge?
On AI memory, competitors include Mem0, Zep and RAG/vector-database stacks; on infrastructure, GPU-orchestration tools like Run:ai and CXL memory-pooling efforts from hardware vendors and hyperscalers.
08

Watch & Demos

Interviews & product
MemVerge YouTube Channel
MemMachine Product Demos
Charles Fan Interviews
09

Find MemVerge

Official channels & press
Websitememverge.ai Open Sourcememmachine.ai LinkedIn/company/memverge Twitter / X@MemVerge FacebookMemVerge CommunityDiscord Docsdocs.memverge.com PressPress Releases

Share this profile

MemVerge — teaching AI to remember