350 MW of Data Centre Cooling Shipped in a Single Month
XNRGY Climate Systems has reported that across its manufacturing network in October 2026 it completed and shipped nearly 115 XNChillers and 500 airside units, delivering over 350 megawatts of total cooling capacity, described as nearly 100,000 tons of refrigeration and more than 70 million cubic feet per minute of air-handling capacity, to customers across North America. The company operates facilities in Mesa, Arizona and Montreal, Quebec, and shipped its first unit in 2021. Its founder, chairman and chief executive Wais Jalali attributes the output to automotive-style lean manufacturing and to the workforce across sales, engineering, operations, logistics and field services. The release gives no efficiency rating, no named customer, no order book, no revenue, no headcount and no delivery lead time.
What 350 megawatts actually means
A capacity figure in megawatts is the right unit and an abstract one, so it is worth converting into what it builds. That is the useful way to read any data centre cooling announcement.
Cooling capacity of 350 megawatts corresponds broadly to the heat that 350 megawatts of IT load produces, so the October output supports something close to that much computing. A current high-density AI rack draws somewhere between 100 and 140 kilowatts, which puts 350 megawatts in the region of 2,500 to 3,500 racks. Expressed the other way, it is roughly three and a half 100-megawatt AI campuses, and XNRGY shipped that in thirty-one days.
The release’s own unit conversions check out, which is worth noting because it is not universal in manufacturing announcements. Nearly 100,000 tons of refrigeration converts to about 350 megawatts at 3.517 kilowatts to the ton, and 70 million cubic feet per minute across 500 airside units works out at roughly 140,000 CFM per unit. Both sit in the plausible range for equipment of this class. A reader can trust these numbers, and that is a meaningful thing to be able to say.
The five-year ramp is the underrated fact
Buried in the chief executive’s quotation is the detail that reframes the milestone. The company shipped its first unit in 2021.
Five years from a first shipment to 350 megawatts in a month is an extraordinary manufacturing ramp, and it is not how industrial HVAC normally moves. Equipment of this class involves pressure vessels, large-frame compressors, heat exchangers, control systems and refrigerant handling, and a new entrant typically spends years qualifying designs, building a supply chain and earning acceptance from consulting engineers before volume arrives. Reaching this output in five years describes an organisation that industrialised the process rather than scaled a workshop.
Jalali’s explanation is the mechanism: automotive-style lean manufacturing. That phrase means something specific in this context. It describes takt time, standardised work, line balancing, and a supply chain managed to a production rhythm rather than to project orders. Applying it to chiller assembly is harder than applying it to cars, because the units are large, the build is not fully repetitive and the configuration varies by project. A company that made it work has a production advantage that a competitor cannot copy by buying the same components.
Efficiency is the data centre cooling number that is missing
The release uses the word efficiency twice. It describes the company as a leader in sustainable thermal management, and it says the October shipments reflect demand for an advanced cooling architecture designed to support high-density AI data centres while significantly reducing energy consumption.
What it does not give is a single efficiency figure. For cooling equipment that is the specification that matters most, because it determines the operating cost over a life measured in decades. Engineers express a chiller’s efficiency in kilowatts per ton, or as a coefficient of performance, and the difference between a good and a mediocre machine compounds across twenty years of continuous operation at hundreds of megawatts.
The gap matters more here than it would in a general HVAC announcement, because the customer is an AI data centre operator whose constraint is power, and data centre cooling is where that power gets spent twice. A campus that has secured 100 megawatts from a utility has to decide how much of it reaches the GPUs and how much the cooling plant consumes. Every point of cooling efficiency is compute capacity returned to the business, and the operator evaluating XNRGY against a competitor needs the kilowatts-per-ton figure to make that comparison. A production milestone without it describes the factory and not the product.
Manufacturing has become the AI bottleneck
The most useful sentence in the release is not about XNRGY at all. It states that as AI data centre operators deploy increasingly powerful computing infrastructure, the ability to manufacture and deliver cooling equipment at scale has become a critical consideration in project planning and execution.
That is an accurate description of where the constraint has moved. The conversation about AI infrastructure spent two years on accelerators and is now about everything around them: transformers with multi-year lead times, gas turbines, switchgear, generators, water rights and cooling. A chip arriving does not open a data centre; the last of those items going in and commissioning does. A supplier that can deliver 350 megawatts of cooling in a month is solving a scheduling problem that has held up projects.
This is why a production milestone is worth covering at all. In most industries, a company announcing how much it shipped is a company talking to itself. In AI infrastructure, throughput at the cooling supplier is a leading indicator of how fast capacity can come online, and it is one of the few places where the physical constraint on AI buildout is visible as a number.
Airside and chilled water together
One compositional detail deserves attention because it says something about where the market is heading. The October output was 115 chillers and 500 airside units, and those are different products serving different cooling approaches.
Airside equipment, being air handling and computer room air handling units, moves cooled air through a hall. It suits traditional densities and remains the default for most enterprise facilities. Chillers produce chilled water, which feeds either air handlers or, at high density, direct-to-chip liquid cooling loops. As AI racks move past the point where air can remove the heat, the chilled water side becomes the growth business.
A supplier with volume in both has a position that survives the transition. A supplier with volume only in airside equipment faces a density shift it cannot serve, and one with volume only in chillers has no route into the facilities that are not yet dense enough to need them. XNRGY shipping both at scale in the same month is a stronger signal about its market position than the aggregate megawatt figure.
Lead time is the number a planner needs
There is a metric that would matter more to a project manager than any capacity figure, and the release does not give it. That metric is delivery lead time.
A data centre reaches commissioning when the last piece of equipment arrives and goes in. In the current buildout the long poles are transformers, gas turbines and switchgear, and those run to multi-year waits. Data centre cooling has historically been a shorter-lead category, and a supplier that can hold a short lead time while shipping at this volume has a commercial advantage that no megawatt figure conveys.
The distinction matters because capacity and availability are different promises. A factory capable of 350 megawatts a month that quotes twelve months to a new customer is a different proposition from one that quotes twelve weeks. The first describes a plant; the second describes a supply chain a project can plan against, and in a market where schedule slippage costs more than equipment, the second is what buyers actually pay for.
What would complete the picture
Five additions would let a reader assess the business rather than admire the output.
The efficiency rating is the first and the most important. Kilowatts per ton at rated and part-load conditions, for the chiller range, would turn a production story into a product story.
The order book is the second. A month of output describes capacity; a backlog describes demand, and the two together describe a business. Whether the October shipments came from contracted orders or from build-to-stock is the difference between a factory running hot and a factory running ahead.
Named customers are the third, and their absence is the largest gap after efficiency. One hyperscaler or colocation provider willing to go on the record would establish the tier of the market the equipment serves.
Delivery lead time is the fourth, because that is the number a project planner actually needs. If a customer ordering today receives equipment in twelve weeks, that is a competitive advantage worth more than any capacity figure.
And the fifth is what the ramp cost. Headcount, capital investment and how hard the Mesa and Montreal plants run would describe whether the five-year build was financed efficiently or whether the growth has outrun the balance sheet.

What to watch
Whether XNRGY publishes efficiency figures, because until it does the sustainable thermal management positioning rests on a claim rather than a measurement. Whether the company names a customer, since one hyperscaler reference would establish the tier this equipment competes in. And, whether the monthly run rate holds through the winter quarters, because a single record month and a sustained production rhythm are different achievements and only the second one matters to a project schedule. And whether the airside-to-chiller mix shifts toward chillers over the coming year, because that ratio is the clearest available indicator of how quickly AI densities are moving to liquid.

