Your GPUs are ready before your building is. L1240C45 removes the building: 1,240 kW of training capacity in one factory-tested container — GB300 NVL72 / B300 HGX density, connected to power, water, and network in days, earning in 75–90 days on repeat orders. 机柜就绪往往早于土建。L1240C45 用集装箱替代大楼： 单台工厂测试的集装箱提供 1,240 kW 训练容量 — 支持 GB300 NVL72 / B300 HGX 密度，数日接通电、水、网， 复购订单可在 75–90 天内创收。
You've seen the business case — modular blocks, BESS-first power, time-to-revenue. L1240C45 is the Computing Unit in that equation: a 45-ft High Cube container holding 8× 150 kW liquid-cooled racks plus one 40 kW air-cooled rack, with all heat carried to the outdoor side on a single 26 °C warm-water loop. Cold plates sit directly on the GPUs; a rear-door heat exchanger and ceiling units sweep up the rest. Everything is integrated, tested, and sealed at the factory — on site you connect power, water, and network. 商业逻辑你已经看过——模块化区块、BESS 优先供电、投产时间。 L1240C45 就是这个等式里的计算单元：一台 45 英尺高柜集装箱，内置 8 个 150 kW 液冷机柜加 1 个 40 kW 风冷机柜，全部热量经一条 26 °C 温水回路 送往室外侧。冷板直接贴装在 GPU 上；后门换热器与顶置空调处理其余热负荷。所有系统在工厂 完成集成、测试与封装——现场只需接入电、水、网。
L1240C45SUR150IT Load vs. Facility Load. 1240 kW is what the servers draw. Total facility load is higher once cooling and auxiliaries are added — the gap is set by PUE, which depends on your site's climate. *Cold, dry sites land at the low end. Always confirm which figure an "X MW" requirement refers to. IT 负荷与设施总负荷。1240 kW 指服务器实际用电。计入冷却与辅助系统后， 设施总负荷更高——差值由 PUE 决定，而 PUE 取决于站点气候。 *寒冷干燥的站点落在区间下限。遇到「X MW」的需求时，务必确认指的是哪一个口径。
A cold plate in direct contact with the GPU carries heat out through water — supporting up to 150 kW/rack, the regime NVLink platforms require and air cooling cannot reach. Junction temperature stays precisely controlled under sustained training load. 冷板与 GPU 直接接触，靠水把热量带走——支撑 最高 150 kW/机柜，这正是 NVLink 平台所需、而风冷无法达到的区间。 在持续训练负载下，结温始终被精确控制。
L1240C45 cools on a 26 °C loop, not legacy 7–18 °C chilled water. At that temperature, dry coolers reject most of the heat directly and mechanical refrigeration stays a trim function — which is what holds PUE near 1.15–1.20, lower in cold, dry climates. L1240C45 使用 26 °C 回路，而非传统 7–18 °C 冷冻水。在这一温度下， 干冷器可直接排走大部分热量，机械制冷只作为补冷环节——这正是 PUE 能维持在 1.15–1.20 的原因；寒冷干燥气候下还更低。
DLC adds real engineering — coolant loops, leak detection, redundant pumps. L1240C45 absorbs all of it into one tested, standardized module. You get the density without owning the integration risk. 液冷确实增加了工程量——冷却回路、漏液检测、冗余泵。L1240C45 把这些全部吸收进 一个经过测试的标准化模块。你拿到密度，却不必承担集成风险。
Orbit the cutaway — rack rows, CDU, warm-water piping, and the air-cooled rack — before you visit a factory.环绕剖视模型——机柜列、CDU、温水管路与风冷机柜——在到工厂前先看清布局。
Side-by-side twin containers with interconnect fire door — 2× 1,240 kW IT reference layout.并排双箱与贯通防火门 — 2× 1,240 kW IT 参考布置。
Heavy USD is explicit download only. Specs: USD output.大体积 USD 仅显式下载。规格见 USD 输出。
The full engineering model is live and interactive — every number computed from the same physics engine our design team uses, referenced to ASHRAE W50 and the NVIDIA GB300 design envelope. 完整工程模型可在线交互——每个数字都来自我们设计团队所用的同一套物理引擎， 参照 ASHRAE W50 与 NVIDIA GB300 设计包络。
Walk the rack rows, piping, and cooling units of the actual L1240C45 layout.在真实的 L1240C45 布局中查看机柜列、管路与冷却设备。
See how 1.24 MW of heat splits across cold plates, rear doors, and room air.看清 1.24 MW 热量如何在冷板、后门换热器与室内空气之间分配。
Flow rates, temperatures, and pump head for each configuration you choose.针对你选择的每种配置给出流量、温度与泵扬程。
Set your site's ambient conditions and watch efficiency respond.设定站点环境条件，实时观察效率变化。