In the quiet arithmetic of the machines that now think for us, a single company has offered a sealed box in which heat and power are tamed together. At a conference in Singapore, ZTE unveiled its ZXEDC M8000, a prefabricated module meant to hold the densest racks of artificial intelligence hardware and to cool them with air, liquid, or both. The claims it makes about efficiency and safety remain its own, awaiting the judgment of independent eyes, but the larger story is familiar: the cost of thinking is increasingly measured in heat.
ZTE Unveils M8000 Micro-Module Data Center for AI Computing, Blending Air and Liquid Cooling
One architecture, three generations of hardware, according to ZTE
Mark
So what did ZTE actually launch?
Mimi
A prefabricated data center module, the ZXEDC M8000. It bundles racks, cooling, power, monitoring and fire systems into one unit built for AI servers.
Mark
Why does the cooling piece matter so much?
Mimi
Rack power is climbing. ZTE lists 20 to 60kW per rack, peaking at 250kW, and it says the same unit can switch between air and liquid cooling, even retrofitted on site later.
Luke
Before we go further, those numbers are ZTE's. The 250kW peak, the PUE of 1.15 and the 98% power efficiency all come from the press release, not from independent testing.
Mark
What about the UL certification claim?
Mimi
ZTE says it's the first system-level leak-proof certification of its kind, covering the CDU, racks, servers and switches.
Luke
It's a notable claim, but the release gives no certificate details and no verification of the word first. I'd treat it as attributed, not confirmed.
Mark
Is there any real-world evidence?
Mimi
One case. A 2025 bid for 666 modules for a Chinese internet company in the Yangtze River Delta, with 345 sets delivered and one month from order to site.
Luke
And the customer isn't named, and we don't know what PUE it achieves in operation. The delivery count is a claim about shipments, not performance.
Mark
So what should a reader watch for?
Mimi
Independent benchmarks, and other named customers beyond that one project.
Il Polso
- Rack power climbs from 20 to 60 kilowatts, with peaks near 250, and the old air-cooled rooms are running out of room to breathe.
- ZTE pitches one prefabricated module that can be deployed once, then switched between air and liquid cooling as demand shifts.
- Vendor claims of a PUE as low as 1.15, 98% power efficiency and UL leak-proof certification have not yet been independently verified.
- The only cited reference is a Yangtze River Delta project for a major Chinese internet firm, with 345 units reported delivered so far.
In the quiet arithmetic of the machines that now think for us, a single company has offered a sealed box in which heat and power are tamed together. At a conference in Singapore, ZTE unveiled its ZXEDC M8000, a prefabricated module meant to hold the densest racks of artificial intelligence hardware and to cool them with air, liquid, or both. The claims it makes about efficiency and safety remain its own, awaiting the judgment of independent eyes, but the larger story is familiar: the cost of thinking is increasingly measured in heat.
At Data Center World Asia in Singapore, ZTE Corporation introduced the ZXEDC M8000 Intelligent Computing Micro-Module on 22 October 2026. The company, which is listed in Hong Kong and Shenzhen, frames the product as an answer to four pressures facing data center operators: the boom in AI computing, rapidly rising power per rack, uncertain changes in business needs, and the move from air cooling to liquid cooling. Its promise is a single foundation that can be deployed once and evolve over time.
The M8000 packs cooling, power, racks, monitoring, lighting and fire linkage into one prefabricated system. Racks draw 20 to 60 kilowatts, peaking at 250 kilowatts, and can be retrofitted from air to liquid cooling on site later. ZTE describes shared air and liquid cooling sources that can be rebalanced dynamically, and claims a PUE as low as 1.15, up to 98% power supply efficiency, and the industry's first UL system-level leak-proof certification.
Every one of these figures comes from ZTE itself. The only reference case is a 2025 contract for 666 IT modules at a Yangtze River Delta computing center for a leading Chinese internet enterprise, with 345 sets delivered so far and a reported one month from order to site. What observers should watch next is whether independent benchmarks appear, and whether deployments beyond that single project are named.
At Data Center World Asia in Singapore, ZTE Corporation announced on 22 October 2026 that it had launched a new integrated IT POD product, the ZXEDC M8000 Intelligent Computing Micro-Module. ZTE is listed in Hong Kong and Shenzhen. The announcement is a vendor press release, and the issuer is solely responsible for its content.
ZTE says the product is aimed at four pressures on data center operators: the AI computing boom, rapidly rising power density per rack, uncertain business iteration, and the shift from air cooling to liquid cooling. Its pitch is a single foundation that can be deployed once and evolve over time.
The M8000 bundles a long list of components into one prefabricated system. These are an enclosed cold aisle, liquid-cooled racks, network racks, a coldplate liquid cooling CDU, secondary-side piping, high-voltage DC power, intelligent busbars, a thin-sheet air wall, monitoring, lighting, cabling and fire linkage. ZTE says it supports dense deployment of AI servers, flexible power distribution, switchable air or liquid cooling, scalable capacity, and mixed deployments of general-purpose and AI computing.
The density figures are the headline. Rack power runs from 20 to 60kW, with a peak of 250kW per rack. The racks are decoupled from the frame, so full racks and special-form-factor racks can both be used, and air-to-liquid cooling retrofits can be done on site later. ZTE says standardized, productized components allow fast assembly and shorter delivery times.
On cooling and power, ZTE describes an air-and-liquid shared-source architecture. The CDU works with thin-sheet air walls and two cooling sources, so the ratio of air to liquid cooling can be adjusted dynamically. The company says the design maximizes free cooling and reaches a PUE as low as 1.15 across all regions. It also pairs 800V HVDC with intelligent busbars for in-row power supply to racks, claiming up to 98% power supply efficiency.
ZTE also says the M8000 is the industry's first solution to obtain UL system-level leak-proof certification. According to the company, leak protection covers the CDU, liquid-cooled racks, liquid-cooled servers and switches, with leak detection and automatic liquid replenishment. For operations, it describes unified control of air and liquid cooling, AI-driven optimization that cuts cooling CLF energy consumption by more than 10%, and visibility into power and cooling through its iDCIM management platform.
ZTE lists three customer benefits. One architecture is said to support three generations of IT hardware upgrades, avoiding large retrofits as capacity grows. Full-chain liquid cooling protection is meant to give high-power AI clusters stable operating conditions. And free cooling plus AI optimization is meant to lower energy use and operating expense.
The only reference case is a 2025 bid. ZTE says it won a contract for 666 IT modules for a computing center in the Yangtze River Delta, for a leading Chinese internet enterprise. So far 345 sets of M8000 IT PODs have been delivered. The design uses a single row of racks and a single row of thin-sheet air walls, with factory prefabrication and skid-mounted transport for on-site hoisting. ZTE says only one month passed between purchase order and arrival on site.
Every performance figure here comes from ZTE itself. The thing to watch is whether independent benchmarks appear, and whether deployments beyond the Yangtze River Delta project are named.
Citazioni salienti
One architecture supports three generations of IT hardware upgrades, avoiding large-scale retrofits during capacity expansion.— ZTE press release (paraphrased)
Vuoi la storia completa?
Leggi l'originale su The Manila Times ↗
Technology