TCO Analysis →
Revenue in ≤ 6 Months · 75–90 Day Delivery · 0.4–200 MW ≤ 6 个月投产 · 75–90 天交付 · 0.4–200 MW

AI Token Factory,
Revenue-Ready.
AI Token 模块工厂,
即刻创收。

MDCX builds and operates modular data centers for AI compute — where hyperscalers won't. Zero to revenue with one partner, hardware to operating system. MDCX 为 AI 算力建造并运营模块化数据中心——覆盖超大规模厂商不愿触达的场景。从硬件到操作系统,一家伙伴,从零到创收。

≤ 6mo
Time to Revenue投产周期
0.4–200 MW
Scalable Range功率范围
$3M/MW
Starting CapEx起步 CapEx
Tier III-class
Availability可用性等级
Scroll下滑
Products产品

Liquid & Immersion Cooling. One OS. One market. 液冷与浸没液冷。一套 OS。一个市场。

Every MDCX data center is assembled from factory-built Liquid Cooling and Immersion Cooling blocks, run by CIOS — and its spare capacity will earn on DCM. 每一座 MDCX 数据中心由工厂预制的液冷与浸没液冷模块组装,由 CIOS 运行——富余算力将在 DCM 上变现。

How we choose these blocks — Why MDCX → · Specs — Docs · Standards · USD 模块选型逻辑 — 为何 MDCX → · 规格 — 文档 · 标准 · USD

Why MDCX为何 MDCX

We Build Where
Hyperscalers Won't.
建在超大规模
不愿到达的地方。

MDCX fills the gap hyperscalers leave: high-density AI capacity from 0.4 MW, deployed in months, built to hold its resale value. MDCX 补上超大规模厂商留下的空白:从 0.4 MW 起的高密度 AI 容量,数月交付,模块可再部署、可转售。

MDCX
Power Scale功率规模0.4 – 200 MW
Time to Revenue投产周期≤ 6 months≤ 6 个月
Asset Liquidity资产流动性High — redeployable, resellable blocks高 — 可再部署、可转售模块
Revenue创收≤ 6 months≤ 6 个月
Hyperscaler超大规模
Power Scale功率规模≥ 10 MW minimum viable≥ 10 MW 起步才经济
Time to Revenue投产周期18 – 36 months18 – 36 个月
Asset Liquidity资产流动性Low — fixed site, stranded on exit低 — 固定场站,退出易成搁浅资产
Revenue创收18 – 36 months18 – 36 个月
Infrastructure Stack基础设施栈

From Infrastructure to Services 从基础设施到服务

Power, infrastructure, and compute — run by one operating system, sold as one revenue-ready service. 电力、基础设施与算力 — 由同一操作系统运行,作为可创收服务交付。

Power infrastructure
Power电力
Grid, off-grid, renewable, BESS — MDCX handles power, permits, and compliance.并网、离网、可再生与 BESS — 电力、许可与合规由 MDCX 统筹。
380V – 800V HVDC
Modular infrastructure
Infrastructure基础设施
High-efficiency thermal systems. Factory-built, delivered as a single plug-and-deploy unit.高效热系统。工厂预制,整机即插即部署。
Liquid · Immersion Cooling液冷 · 浸没液冷
AI compute racks
Compute算力
GPU and ASIC — NVIDIA GB300/B300 to RTX Pro 6000. Configured for your workload from day one.GPU 与 ASIC — 从 NVIDIA GB300/B300 到 RTX Pro 6000,首日按负载配置。
GPU · ASIC
CIOS operations
CIOS · Operating SystemCIOS · 操作系统
CIOS · Compute Infrastructure Operating System — telemetry, operations, and metering for every block below.CIOS · 算力基础设施操作系统 — 为下层每一模块提供遥测、运维与计量。
See · Control · Monetize看见 · 管控 · 变现
Revenue-ready services
Services服务
Revenue-ready compute service — standardized, scalable, bankable.可创收的算力服务 — 标准化、可扩展、可融资。
Full Lifecycle全生命周期

Zero to Revenue — One Partner. 从零到创收 — 一家伙伴。

Each service layer moves infrastructure closer to revenue. 每一层服务都把基础设施推向可创收。

Lifecycle service
01
Consulting Services咨询服务
Feasibility, site, and investment assessment.可行性、选址与投资评估。
Lifecycle service
02
Design Services设计服务
Deployment-ready system architecture.可部署的系统架构设计。
Lifecycle service
03
Deployment Services部署服务
Factory-built, on-site deployment.工厂预制,现场部署。
Lifecycle service
04
MDCX Care+
Lifecycle support & optimization, powered by CIOS.全生命周期支持与优化,由 CIOS 驱动。
Lifecycle service
05
AI Model OptimizationAI 模型优化
Performance & efficiency tuning.性能与能效调优。
Lifecycle service
06
Aggregation & Brokerage聚合与撮合
We match your idle capacity to buyers.将闲置算力匹配给买方。
Trust可信赖

Tier II Investment. Tier III-Class Availability. Tier II 投资结构。Tier III 级可用性。

Differentiated redundancy — 2N where failure hurts, N+1 where it doesn't. Availability where it matters, without blanket-2N cost. 差异化冗余 — 关键路径 2N,其余 N+1。在需要处保证可用性,避免全盘 2N 成本。

MDCX designs, builds, and operates modular data centers for AI compute — turnkey, from strategic planning through equipment supply to CIOS-driven operations. Factory-built liquid-cooled blocks cut deployment from years to months and are engineered for measured ROI: model yours in the TCO wizard. MDCX 为 AI 算力设计、建造并运营模块化数据中心 — 交钥匙交付,从战略规划、设备供应到 CIOS 驱动运维。工厂预制液冷模块将部署从数年压缩到数月,并以可测算 ROI 为目标:可在 TCO 向导中建模。

Headquartered in Sunnyvale, California, with our APAC hub in Singapore, we work with global partners to serve clients in over 40 countries. 总部设于加州 Sunnyvale,亚太枢纽在新加坡,与全球伙伴协作,服务 40 余个国家的客户。

2021
Founded成立
200 MW
Scale envelope规模上限
40+
Countries国家/地区
~1.06
PUE (annualized)PUE(年化)
420 GWh
Electricity Saved节电量
440 N Wolfe Rd, Sunnyvale, CA 94085, United States
Warranty质保
  • OEM-backed immersion warrantyOEM 背书的浸没质保
  • Intel DataCenter CertifiedIntel 数据中心认证
Scale on Demand按需扩展
  • Minimal unit: 400 kW最小单元:400 kW
  • Blocks hold resale value — no stranded-asset writedown模块保有转售价值 — 避免搁浅资产减记
Power Density功率密度
  • Liquid Cooling: up to 150 kW/rack (L1240C45)液冷:最高 150 kW/机柜(L1240C45)
  • Immersion: up to 50 kW/tank (I400C45 / I400C40)浸没:最高 50 kW/罐(I400C45 / I400C40)
800 V HVDC — Roadmap800 V HVDC — 路线图
  • 380/400/415/480 V AC today当前支持 380/400/415/480 V AC
  • 800 V HVDC (Roadmap · Q3 2026)800 V HVDC(路线图 · 2026 Q3)
Investment Analysis投资分析

Model Your Return
Before You Invest.
投资前
先测算回报。

Our interactive TCO wizard walks through CapEx, power costs, utilization, and projected IRR — tailored to your site parameters. 交互式 TCO 向导覆盖 CapEx、电价、利用率与预期 IRR — 按站点参数定制。