jain.com — data centers, connectivity and security, analyzed daily

Tag: liquid cooling

  • Frore’s 35% Tokens-per-Watt Claim Turns GPU Hotspots Into an AI Power Lever

    Frore’s 35% Tokens-per-Watt Claim Turns GPU Hotspots Into an AI Power Lever

    TL;DR · 30-second read

    The Short Version

    A California company called Frore Systems says it can get up to about a third more work out of the chips that run artificial intelligence, without using more electricity.

    The trick is cooling. These chips get extremely hot in a few tiny spots, and when they overheat they slow themselves down. Frore places small pieces of diamond, which moves heat better than metal, right over those spots.

    Why it matters: electricity is now the scarcest ingredient in building artificial intelligence. Squeezing more from the same power could mean fewer new buildings. Frore has not yet published independent test results or named customers.

    Frore Systems announced LiquidJet Diamond on October 8, 2026, in San Jose, California. The product is a liquid-cooling coldplate (the metal block that sits on a chip and carries heat away into circulating fluid) with diamond wafers built into its base, positioned directly over the hottest regions of a GPU. Frore says the design cools hotspots of up to 770 watts per square centimeter and delivers 30–35% more AI tokens per watt than a conventional skived coldplate.

    The company will demonstrate LiquidJet and LiquidJet Diamond cooling a 2,300-watt NVIDIA Rubin GPU, along with its LiquidJet Nexus integrated coldplate system for half-height compute trays, at the OCP Global Summit in San Jose, October 12–16, 2026.

    Executive Summary

    LiquidJet Diamond extends Frore’s existing LiquidJet coldplate, which uses what the company calls a 3D ultra short-loop multi-stage design customized to each GPU’s power map. Frore says the base product lowers die temperature by 12°C, worth 25% more tokens per watt, and the diamond spreader lowers hotspot temperature a further 10°C, lifting the gain to 35%. Against a skived coldplate, it claims over 150% higher maximum hotspot power density, 75% higher heat transfer efficiency, a GPU running 16–22°C cooler, and a drop-in upgrade path.

    The significance lies in the framing rather than the hardware alone. Frore is arguing that in AI data centers capped by grid connections and transformer capacity, chip temperature is a capacity variable: a cooler chip produces more useful output from the same fixed power budget. If the numbers hold up under customer testing, cooling moves from a facilities cost line to a revenue lever.

    The headline claim of up to 35% higher profitability, however, rests on a chain of assumptions about demand, pricing and component cost that Frore has not detailed, and the performance figures have not yet been independently validated.

    Why a Few Square Millimeters Can Set a Data Center’s Output

    A modern AI accelerator does not heat evenly. Some regions of the die, the slab of silicon doing the computation, run much hotter than others. A chip protects itself by slowing down when its hottest point crosses a temperature limit, a behavior known as thermal throttling, and hotter silicon also leaks more electrical current that does no useful work. Both effects mean the temperature of the worst spot on the chip, not the average temperature of the rack, can determine how much computing each watt buys.

    That is the mechanism behind Frore’s numbers. The base LiquidJet design cuts die temperature by 12°C, which Frore equates to 25% more tokens per watt. Tokens are the chunks of text an AI model reads or writes, so tokens per watt is a rough measure of useful AI output per unit of electricity. Adding diamond, which conducts heat far better than copper, directly over the hotspots cuts hotspot temperature a further 10°C and lifts the claimed gain to 35%. Frore says the diamond version handles hotspot power density more than 150% higher than a skived coldplate, a statement about peak heat concentrated in a small area rather than total heat removed.

    The reason that ratio matters is the constraint Frore itself names: utility interconnection (the queue to connect a site to the grid), transformer capacity and on-site power delivery cap what an AI facility can draw. In a building whose power is fixed, more tokens per watt is effectively added capacity without a new substation. Chief executive Seshu Madhavapeddy makes that case directly, saying that with 35% more tokens from the same resources, hyperscalers “would reach their growth goals faster with fewer AI Factories.” The operators most exposed are those deploying the newest, highest-power GPUs, where hotspot density climbs fastest.

    From Tokens per Watt to Profit Takes Several Assumptions

    The release headline promises up to 35% higher profitability; the body says tokens per watt and revenue rise 35%, “with the increased revenue flowing directly to profit.” Each step in that chain carries conditions. Extra tokens only become revenue if there are buyers for them at the same price. Extra throughput only materializes if the GPU is limited by heat rather than by memory bandwidth, networking or software. And the gain only flows straight to profit if the diamond coldplate costs no more over its life than the part it replaces, a figure Frore has not published.

    The temperature figures also mix measures. The 12°C reduction refers to die temperature and the additional 10°C to hotspot temperature; their sum, 22°C, matches the top of the 16–22°C range Frore quotes for a cooler GPU, and the 30–35% tokens-per-watt range suggests results vary by chip and workload. “Up to” marks 35% as the best case. The baseline matters too: “traditional skived coldplates” covers a wide range of designs, and the size of any improvement depends on which one is being compared.

    None of this makes the direction implausible. Cooler silicon generally runs faster and wastes less power, and hotspot-targeted cooling is a logical response to rising chip power. The open question is magnitude under real data center conditions, which is what customer qualification testing exists to establish.

    The Harder Claim Is Manufacturing at Hyperscale

    Frore positions LiquidJet between two incumbent approaches. Skiving, in which thin fins are carved from a copper block, is mature and cheap but, Frore argues, cannot deliver the precision today’s GPUs demand. Electrochemical additive manufacturing (ECAM), a form of 3D metal printing, offers design freedom but, in Frore’s telling, cannot meet hyperscaler quality, reliability, volume or timelines. Frore says it combines semiconductor-fabrication precision with high-volume manufacturing, with production operations in Taiwan.

    Size is part of the pitch. Frore cites die areas ranging from 1,488 mm² for NVIDIA Blackwell Ultra to 46,225 mm² for the Cerebras Wafer-Scale Engine, roughly 30 times larger, and claims its process is the only one able to support the power, performance and size requirements of such chips. Its LiquidJet Nexus system goes further, integrating multiple coldplates and eliminating hoses, connectors and manifolds to enable half-height (½U) compute trays, which would mean fewer potential leak points and denser servers.

    If those claims survive qualification, the pressure falls on suppliers of skived coldplates and on 3D-printing approaches competing for the same high-end sockets. The “seamless drop-in upgrade” claim is commercially important because it lowers switching costs, but hyperscale buyers typically qualify cooling hardware over long cycles, and a component that sits directly on a multi-thousand-dollar GPU faces a high bar for reliability evidence.

    Background

    Frore Systems built its reputation on AirJet, a solid-state active air-cooling chip used in consumer, industrial and Internet-of-Things devices, including a concept ultra-slim notebook unveiled with Lenovo. The company, which describes itself as an AI industry unicorn, has since moved into data center liquid cooling with LiquidJet and LiquidJet Nexus, and earlier published a report, “Chip to Chiller,” arguing that heat costs AI facilities a large share of their compute capacity.

    The broader market context is a shift from air cooling to direct-to-chip liquid cooling as AI accelerators have climbed into the thousands of watts per chip. With grid interconnection and transformer capacity now limiting how fast AI data centers can grow, suppliers across the thermal stack are increasingly pitching their products on output per watt rather than on cooling alone. The OCP Global Summit, run by the Open Compute Project industry consortium, is a key venue where hyperscalers evaluate such hardware.

    Sources

    Source: LiquidJet Diamond Coldplates can increase AI Factory profitability by up to 35% — Frore Systems’ announcement of its diamond-spreader liquid-cooling coldplate for AI GPUs, distributed via PR Newswire.

  • Vertiv’s Second Fluid-Services Deal Since 2025 Bets AI Coolant Loops Need Lifelong Care

    Vertiv’s Second Fluid-Services Deal Since 2025 Bets AI Coolant Loops Need Lifelong Care

    TL;DR · 30-second read

    The Short Version

    The newest AI computers run so hot that many are cooled with liquid piped right to the chips, a bit like a car’s radiator. That liquid has to stay clean, free of air bubbles and chemically balanced. If it doesn’t, the cooling quietly gets worse and parts wear out.

    Vertiv, a big maker of power and cooling equipment for data centers, is buying King Environmental Services, an Irish company that specializes in exactly that upkeep. It is Vertiv’s second such purchase in under a year. Looking after the coolant is becoming an ongoing job, not a one-time setup.

    Vertiv Holdings (NYSE: VRT), which makes power and cooling equipment for data centers, has agreed to acquire King Environmental Services Ltd. (KES). KES is headquartered near Dublin, Ireland, and provides fluid management, commissioning and load-testing services for high-density, liquid-cooled data centers across Europe, the Middle East and Africa (EMEA). Zawya carried the announcement on October 5, 2026.

    Financial terms were not disclosed, and Vertiv said they are not expected to be material to its results. The deal is expected to close in the fourth quarter of 2026. It follows Vertiv’s December 2025 acquisition of PurgeRite, which the company describes as the foundation of its fluid management business in North America.

    Executive Summary

    Vertiv is buying a specialist field-services firm rather than a product line. KES flushes, filters, treats and validates the liquid that circulates through cooling systems. It also runs thermal load tests on liquid cooling infrastructure and electrical load tests on critical power systems before a facility goes live. Vertiv says the deal extends its “thermal management services capabilities” into EMEA and supports its strategy of serving customers “across the thermal chain.”

    The significance is in the pattern and the scope. This is the second fluid-management acquisition Vertiv has named in roughly ten months, after PurgeRite in North America. The scope Vertiv describes runs well past installation day, covering “monitoring, maintenance and lifecycle optimization.” That is a statement about where Vertiv expects work and revenue to sit in liquid-cooled AI data centers: in the ongoing care of the coolant loop, not only in the hardware that carries it.

    The Coolant Has Become Equipment That Needs Servicing

    Air-cooled data centers move heat with fans and air handlers. Liquid-cooled AI halls instead pump a water-based coolant through cold plates mounted on the chips, then through coolant distribution units (pumps and heat exchangers) that pass the heat to the building’s water system. Vertiv’s announcement lists what goes wrong in that loop: contamination, trapped air, corrosion, improper chemistry and flow imbalances. Each reduces heat-transfer efficiency, degrades equipment and creates risk both at startup and in operation. The fluid’s condition changes over time, which makes it a component to be maintained, not plumbing to be installed and forgotten.

    That mechanism explains why Vertiv describes KES’s scope in lifecycle terms: “system preparation, performance validation, monitoring, maintenance and lifecycle optimization.” It also says outright that “beyond initial commissioning, customers need lifecycle fluid management services.” CEO Gio Albertazzi framed it this way: “the performance of cooling and power infrastructure will depend as much on execution in the field as on engineering and system design.” Every liquid-cooled rack adds piping and fluid volume. That fluid must be flushed and validated before go-live, then sampled, treated and balanced for as long as the facility runs. Work of that kind recurs; it does not end at handover.

    The people affected are the operators of AI data centers, whose uptime depends on fluid quality, and the tenants leasing that capacity. Also affected are the mechanical contractors and independent specialists who have historically done flushing and commissioning as project work. One caveat applies: Vertiv has not disclosed how much of this business is recurring maintenance versus one-time commissioning. The recurring-service thesis is clearly stated in the company’s strategy, but not yet shown in numbers.

    Two Deals, Two Regions, One Template

    PurgeRite gave Vertiv a fluid management base in North America; KES is meant to do the same for EMEA. Two acquisitions by one company do not make an industry trend. They do show a deliberate geographic build-out of the same capability, by a supplier whose equipment already sits inside the cooling loops these teams service. KES’s Dublin-area base puts it in one of Europe’s established data center markets.

    KES also brings load testing. Thermal load testing uses heat-generating equipment to simulate a full room of running servers, so operators can confirm the cooling performs before costly AI hardware arrives. Electrical load testing does the equivalent for backup and critical power systems. Finding a flow imbalance or a power fault before the servers are installed is far cheaper than finding it afterward. Vertiv’s pitch is that one “single accountable partner” can handle commissioning, testing, fluid management and ongoing operations, plus compliance with regional rules on fluid handling, treatment and disposal.

    A Small Price, a Clear Signal and a Neutrality Question

    Vertiv says the terms are not material to its financial results. That suggests KES is small next to Vertiv’s business, and that this is a purchase of capability and footprint rather than revenue. Experienced field teams, local regulatory know-how and a safety record take years to build, and buying them is the faster route into a region. Vertiv’s own risk disclosures name the obvious exposure: keeping KES’s management and key employees, and maintaining its customer and supplier relationships, after the deal closes.

    There is a fair question for buyers too. Commissioning is often valued as an independent check on what a vendor delivered. When the company supplying cooling equipment also validates and maintains it, customers may reasonably ask how independence is preserved. Vertiv’s single-partner model offers clear accountability. Whether customers weigh that above independent verification is something the market, not the announcement, will settle.

    Background

    Vertiv, headquartered in Westerville, Ohio, and listed on the New York Stock Exchange, supplies power, cooling and IT infrastructure, along with software, analytics and services, to data centers, communication networks and commercial and industrial facilities. It does business in more than 130 countries.

    The shift behind this deal is the rise of liquid cooling. As AI and high-performance computing workloads push more heat into each rack, air alone struggles to remove it, and operators increasingly pipe coolant directly to the hardware. That makes the condition of the fluid, along with the testing and upkeep of the systems that carry it, a factor in uptime. Field services such as commissioning and fluid maintenance are becoming a larger part of how cooling suppliers compete.

    Sources

    Source: Vertiv announces agreement to acquire King Environmental Services Ltd., expanding global fluid management services: Zawya’s publication of Vertiv’s October 5, 2026 announcement of its agreement to acquire Ireland-based KES.

  • Castrol CORE’s 150-Country Network Recasts AI Coolant as a Lifecycle Service

    Castrol CORE’s 150-Country Network Recasts AI Coolant as a Lifecycle Service

    TL;DR · 30-second read

    The Short Version

    Castrol, best known for motor oil, is now selling a full package to keep the computers behind artificial intelligence from overheating.

    Those chips run so hot that many buildings now cool them with liquid instead of fans. Castrol says it will help design those cooling systems, supply the fluid, test it, and look after it for years, backed by service teams in more than 150 countries.

    This matters because a cooling failure can knock very expensive computers offline. Castrol is wagering that customers would rather pay one company to own that risk.

    Castrol announced on September 29, 2026, from Singapore, the launch of Castrol CORE, an end-to-end thermal management offer for data centers, in a release distributed through PR Newswire. The package combines Castrol’s cooling fluid technology with testing, new design capabilities, and support through deployment, commissioning, operation and maintenance. It is backed by Castrol labs in the UK, US, Germany and China, joint labs across Asia, and a service-provider network covering more than 150 countries.

    Castrol says its cooling technology has been tested and validated by leading chip and hardware makers including NVIDIA and Intel. It has completed a proof-of-concept program with an unnamed major data center operator in China and is working with other global operators on similar projects.

    Executive Summary

    Castrol CORE is Castrol’s move from selling cooling fluid to selling responsibility for how that fluid performs over a facility’s life. Rather than supplying a product that operators must integrate with hardware from other vendors, Castrol says it will act as a single point of contact, coordinating a network of partners in products, infrastructure, design and services. The company attributes the shift to its own industry study, which it says found rising demand for integrated thermal management, plus concerns about supply chains and the high cost of failures.

    The launch matters because liquid cooling, circulating fluid close to or around hot chips instead of relying on air, is becoming a standard requirement for AI and high-performance computing. When cooling moves from fans to fluids, the chemistry of the fluid becomes part of the uptime equation. A supplier that tests and monitors that fluid on site, across more than 150 countries, is offering something closer to an operational guarantee than a commodity.

    What Castrol has not disclosed is how much of this is already contracted. It names no customers, deployed capacity, pricing or performance commitments, so CORE is best read as a clear statement of strategy with one completed pilot behind it.

    Why a Coolant Maker Is Selling Accountability

    The core of the announcement is a change in what the customer is paying for. Castrol says CORE covers the full lifecycle, from early design and specification through deployment, commissioning, operation and maintenance, and that its service network in more than 150 countries enables ongoing testing, validation and on-site monitoring to reduce the risk of costly interruptions. That is the mechanism behind the headline: the fluid is no longer the whole product. It is the anchor for a recurring relationship built around keeping the cooling loop healthy.

    This follows from how liquid cooling works. In an air-cooled hall, a failing fan is a swappable part. In a liquid-cooled system, the fluid’s condition, including contamination, degradation or incompatibility with pumps, seals and cold plates, can affect every server on the loop. Castrol’s own study, it says, flagged the high cost of failures as a major industry concern. A supplier that samples and validates fluid on site throughout a facility’s life is effectively selling a reduction in that risk, and the geographic reach of the service network is what makes that promise credible for operators with sites on several continents.

    The people most affected are data center operators and the engineering firms that design their facilities. For operators, a single interlocutor could simplify procurement and clarify who is responsible when something goes wrong. For design and integration firms, a fluid supplier moving upstream into design is a new participant in work they have typically led. Castrol frames this as partnership, saying CORE brings together specialists across products, infrastructure, design and services, but the balance of who leads a project will depend on terms it has not yet published.

    What Chipmaker Validation Does and Does Not Establish

    Castrol’s strongest technical claim is that its cooling technology has been tested and validated by leading chip and hardware makers, including NVIDIA and Intel. For buyers, validation from the companies whose processors are being cooled carries real weight, because running an unapproved fluid can raise questions about hardware compatibility and support.

    The claim is also broad. Castrol does not specify which products or platforms were validated, whether the fluids are for direct-to-chip cooling (where liquid flows through plates mounted on processors) or immersion cooling (where servers sit submerged in a non-conductive fluid), or what the validation tested. Operators evaluating CORE will want that detail, since compatibility with one generation of hardware does not automatically carry over to the next.

    Asia-Pacific as the Proving Ground

    Castrol has set up a regional data center design team in Asia-Pacific, supported by local specialists, and describes the region as adopting new cooling infrastructure quickly. Its one disclosed field result is a completed proof-of-concept program with a leading data center operator in China, with similar projects now under way with global operators.

    A proof of concept demonstrates that the approach can work in a real facility; it is not a commercial rollout. The combination of a Singapore launch, an Asian design team, joint labs across Asia and a China pilot suggests Castrol sees the region as its early market for the integrated model. Whether CORE scales from pilots into multi-site contracts is the question that will determine whether the service shift is a strategy or a business.

    The Trade-Off in One-Stop Cooling

    Integrated offers solve a real coordination problem, but they shift risk in both directions. For Castrol, taking on design and lifecycle responsibility means absorbing some of the liability that previously sat with integrators and operators, and managing a partner network it does not fully control. For customers, a single supplier simplifies accountability but can raise switching costs if fluids, monitoring and design are tightly bundled.

    Castrol also says CORE is a platform that should gradually extend its thermal management offering beyond data centers. That signals the company views thermal management as a growth line in its own right, not a side market for lubricants, though it has not said which applications come next or on what timeline.

    Background

    Castrol is one of the world’s largest lubricant brands, with more than 127 years of experience formulating fluids for automotive, marine, industrial and energy applications. Its data center work sits within a dedicated Data Centers and Thermal Management division, led globally by president Peter Huang, who says customers are looking for experienced partners to handle the full range of thermal challenges rather than another standalone product.

    The broader context is the rise of liquid cooling. As AI and high-performance computing push chip heat output higher, operators are adopting systems that use specialized fluids to carry heat away from processors. That has turned fluid suppliers into participants in data center design and operations, a role Castrol is now formalizing with CORE.

    Sources

    Source: Castrol lance « Castrol CORE », une offre intégrée de gestion thermique destinée aux centres de données (Castrol launches Castrol CORE, an integrated thermal management offer for data centers), Castrol’s announcement of its end-to-end data center cooling offer, distributed via PR Newswire on September 29, 2026.

  • Vertiv’s 2nd Fluid-Services Deal in Under 10 Months: AI Cooling Risk Sits at Handover

    Vertiv’s 2nd Fluid-Services Deal in Under 10 Months: AI Cooling Risk Sits at Handover

    TL;DR · 30-second read

    The Short Version

    Vertiv, an American company that makes the power and cooling equipment inside data centers, is buying King Environmental Services, a Dublin firm whose crews clean, fill and test the liquid cooling systems used by artificial intelligence computers.

    Those chips run so hot that many new buildings pipe liquid right to them, much like a car’s radiator. A little dirt or trapped air in the pipes can overheat very expensive machines.

    It is Vertiv’s second purchase of this kind in under a year. The bet: the hard part is getting the plumbing right on day one.

    Vertiv Holdings Co (NYSE: VRT) said on September 24, 2026 that one of its wholly owned subsidiaries has signed a definitive agreement to acquire King Environmental Services Ltd. (KES), a fluid management, commissioning and load-testing specialist founded in 2002 and headquartered near Dublin, Ireland, with hubs in North Dublin and Frankfurt. Vertiv disclosed the deal in an 8-K filed the same day, with its press release furnished as Exhibit 99.1.

    The transaction is expected to close in the fourth quarter of 2026. Vertiv did not disclose financial terms and said the deal is not expected to be material to its financial results. It follows Vertiv’s December 2025 acquisition of PurgeRite, which the company describes as the foundation of its North American fluid management business.

    Executive Summary

    Vertiv is extending into Europe, the Middle East and Africa (EMEA) a services line it began building in North America less than a year ago: preparing, cleaning, balancing and testing the liquid loops that carry heat away from high-density AI servers. KES brings pipework flushing, filtration and water treatment, testing, adjusting and balancing (TAB, the process of tuning flow so every branch of a cooling system gets the fluid it was designed for), and large-scale thermal and electrical load testing, in which heaters and load banks stand in for real servers so a facility can be proven before it goes live.

    The deal is small by Vertiv’s own account, but the pattern is the story. Two consecutive fluid-management acquisitions, PurgeRite and now KES, put Vertiv’s capital into field crews rather than factories. CEO Gio Albertazzi framed the rationale directly: as computing density rises, cooling and power performance “will depend as much on execution in the field as on engineering and system design.”

    For data-center builders and operators in Europe, the practical consequence is that one of the largest thermal equipment vendors now intends to own the commissioning and lifecycle fluid work around its systems, offering what it calls a single accountable partner from construction through operation.

    Why Liquid Cooling Risk Concentrates at Handover

    Air-cooled data centers are forgiving: if a fan underperforms, the room runs a little warmer. Direct liquid cooling is less forgiving. Heat moves from chips into a closed loop of treated water or coolant, and Vertiv’s own description of what can go wrong is specific: contamination, trapped air, corrosion, improper chemistry and flow imbalances can reduce heat-transfer efficiency, degrade equipment and create commissioning and operational risks. Every one of those failure modes is introduced or caught in the same window, when a newly built system is flushed, filled, balanced and brought under load for the first time.

    That window is the handover between construction and live operation, and it is exactly where KES sits. The company describes itself as bridging construction and live operation, validating that cooling systems are clean, balanced, design-compliant and ready for AI workloads, including through thermal load testing before a facility enters service. PurgeRite plays the equivalent role for Vertiv in North America. Buying two such firms in under ten months, December 2025 and September 2026, is a clear signal of where Vertiv believes customer risk now sits: not in whether a coolant distribution unit or cold plate works on a test bench, but in whether the loop around it was installed and prepared correctly on site.

    This matters to three groups. Operators carry the downtime risk if a contaminated loop fouls hardware. General and mechanical contractors carry schedule risk if commissioning uncovers problems late. And hardware vendors carry reputational risk when a field problem is blamed on their equipment. Vertiv’s pitch of “a single accountable partner for commissioning, testing, fluid management and ongoing operations” is aimed squarely at that last point, collapsing a chain of hand-offs into one party that owns the outcome.

    A Small Check With a Strategic Signal

    Vertiv says the deal is not expected to be material to its results, and it furnished the announcement under Item 7.01, the Regulation FD disclosure item, in its September 24 8-K. In plain terms, this is a capability purchase, not a revenue purchase. Vertiv is buying people, procedures, regional permits and an operating footprint that would take years to build organically.

    The economics also point toward recurring work. The release stresses that beyond initial commissioning, customers need lifecycle fluid management for availability, thermal performance and equipment longevity. Fluid chemistry drifts, filters load up and systems get expanded, so a firm that commissions a loop is well placed to maintain it. For an equipment maker, attaching a service relationship to each installed liquid-cooling system is a way to stay involved long after the hardware sale.

    Vertiv’s own risk language underlines what it is really acquiring. Among the transaction risks listed in the release are its ability to maintain relationships with KES’s customers and suppliers and to retain KES management and key employees. In a field-services business, those relationships and people are the asset.

    What EMEA Adds, and What Remains to Be Proven

    KES operates from North Dublin and Frankfurt, two established European data-center hubs, and serves customers across EMEA. The release also highlights compliance with regional fluid handling, treatment and disposal requirements, a less glamorous but real advantage: treated coolant is a regulated waste stream, and a firm with local disposal routes and regulatory familiarity removes friction that a North American team would face entering the market.

    The broader claims deserve a measured reading. Vertiv describes “significant customer benefits” and KES’s “strong reputation” but offers no figures on KES’s headcount, number of projects, load-testing capacity or customer base, so the scale of the European execution capacity being added cannot yet be assessed. Reach is also a fair question: two hubs in Ireland and Germany can serve a lot of Europe, but supporting the Middle East and Africa at the same standard is a different logistical task.

    There is also a structural question for buyers. Commissioning has traditionally been valued partly for its independence from the equipment supplier. When the firm validating a cooling system is owned by the company that made parts of it, operators may want clarity on how test results, disputes and non-Vertiv equipment are handled. That is not a criticism of either company; it is the natural question any procurement team will ask as vendors move deeper into services.

    Background

    Vertiv, headquartered in Westerville, Ohio, and doing business in more than 130 countries, supplies power, thermal management and IT infrastructure equipment and services for data centers, communication networks and commercial and industrial facilities. As AI and high-performance computing push more heat into each server rack, operators are increasingly adopting liquid cooling, in which coolant is piped to or near the chips instead of relying only on moving air through the room.

    Liquid cooling shifts a new kind of work onto construction sites and operations teams: the cleaning, filling, balancing and ongoing treatment of cooling loops. Vertiv entered that business in North America with its December 2025 acquisition of PurgeRite. KES, founded in 2002, has specialized in the same class of work in Europe and serves customers across EMEA from North Dublin and Frankfurt.

    Sources

    Source: Vertiv Announces Agreement to Acquire King Environmental Services Ltd., Expanding Global Fluid Management Services, Vertiv’s announcement of its planned acquisition of the Dublin-based fluid management and load-testing firm.

    Primary sources: Vertiv Holdings Co Form 8-K, filed September 24, 2026 (SEC EDGAR); Exhibit 99.1 to Vertiv’s September 24, 2026 8-K: press release on the King Environmental Services acquisition.

  • AI Density Without a New Build: OData’s $630M Liquid Cooling Retrofit in Brazil, Mexico

    AI Density Without a New Build: OData’s $630M Liquid Cooling Retrofit in Brazil, Mexico

    TL;DR · 30-second read

    The Short Version

    The computer chips that run artificial intelligence get so hot that blowing cold air over them is no longer enough. The fix is to pipe liquid right up to the chip.

    OData, a data center operator active across Latin America, is spending 630 million dollars to add that liquid plumbing to two buildings it already runs, one in Brazil and one in Mexico. The work has started.

    Why it matters: the region can host the newest, hottest artificial intelligence machines by upgrading buildings that already exist, instead of waiting years for new ones to be built.

    Data Center Dynamics reported that OData is installing DeltaFlow, a proprietary liquid cooling system developed by its parent company Aligned Data Centers with the power and cooling vendor Munters, at its SP04 data center in Brazil and its QR03 facility in Mexico. The project is priced at $630 million across the two sites, and phase one installation and testing is already underway at both.

    Both facilities already run Aligned’s Delta Cubed air cooling system. DeltaFlow will let them operate as hybrid halls — air cooling for conventional servers, liquid cooling for the high-density racks that AI accelerators require. The DeltaFlow units are manufactured at Munters’ plant in Araucária, in the Brazilian state of Paraná.

    Executive Summary

    This is a retrofit, not a groundbreaking. OData is adding direct liquid cooling — coolant carried to a plate sitting on the processor itself, rather than chilled air pushed across a room — to two data centers that are already built, already powered and already in service. That distinction is the whole economic argument. Liquid cooling is what allows a rack to draw the tens of kilowatts that AI accelerators consume, and the fastest route to that capability is almost always a building that already exists.

    The $630 million figure is unusually large for a project described in terms of cooling, which suggests scope well beyond swapping chillers: retrofits of this type typically pull in electrical distribution, floor loading, piping, heat rejection and controls. The release does not break the number down, and it does not state how much liquid-cooled capacity the spend produces.

    The second thread is manufacturing. DeltaFlow is built at a Munters factory in Paraná, inside Brazil. Long-lead mechanical equipment has been one of the more stubborn schedule constraints in data center construction, and supplying a Brazilian deployment from a Brazilian plant removes ocean freight and customs from that path.

    The Retrofit Is the Point

    OData’s two target sites are not new. The release states that both already operate Aligned’s Delta Cubed air cooling, and that DeltaFlow will enable them to run as hybrid air-and-liquid facilities catering to high-density AI racks. In plain terms: the buildings were designed for servers that a stream of cold air could keep in range, and they are being upgraded for servers that it cannot. A modern AI accelerator rack concentrates heat in a volume where moving enough air becomes impractical and expensive long before it becomes impossible, which is why coolant is brought to the chip instead.

    What that changes operationally is the pacing item. A new data center is gated by land, permits, construction and — most painfully in the current cycle — a grid connection, the queue to get utility power delivered to the site. A retrofit inherits all of those. The gate becomes mechanical installation and commissioning inside a shell that already has power, floor, network and security in place. That is measured in a fit-out schedule rather than a development schedule, and it is the reason a hybrid upgrade can put dense capacity into a market faster than any greenfield project competing for the same customer.

    The people this reaches are buyers in Brazil and Mexico who want to train or serve AI models on machines physically located in the region — for latency, for data residency, or simply because capacity elsewhere is spoken for. The release confirms the capability is coming and that work has begun. It does not say how many megawatts of liquid-cooled capacity results, or what rack density in kilowatts the halls will support, so the size of that advantage cannot yet be measured from the outside.

    $630 Million Is a Building Number, Not a Chiller Number

    The figure is given for the project across both sites, without itemisation. It is worth being precise about what a cooling retrofit actually involves, because the headline number only makes sense at that scope. Introducing liquid cooling into a live air-cooled hall generally touches the electrical distribution feeding denser racks, the structural loading of the floor, the piping and manifolds that carry coolant, the heat rejection plant outside the building, the water or refrigerant management behind it, and the control systems that have to run two cooling regimes at once without either interfering with the other — all while the existing customers stay online.

    What is substantiated here is a capital commitment of a stated size, a named technology, two named facilities and work that has begun. What is not substantiated is the unit economics. Without a capacity figure, there is no cost per megawatt, and cost per megawatt is precisely the number that would tell a buyer or an investor whether upgrading beats building. That is a disclosure question for OData and Aligned, not a flaw in the underlying project.

    There is a strategic read in how the system is sourced. Aligned designed DeltaFlow itself, with Munters, rather than buying merchant coolant distribution units off a catalogue. For a group operating across several countries, a single in-house platform means common spares, common training and a repeatable commissioning process — real advantages when you are staffing facilities in four national markets. The trade-off is concentration: the platform’s roadmap, and the factory behind it, become a dependency.

    The Factory in Paraná Is Part of the Infrastructure Story

    DeltaFlow is manufactured at Munters’ plant in Araucária, Paraná, in southern Brazil. That detail does more work than it first appears to. Across the industry, the binding constraint on data center schedules has repeatedly been long-lead mechanical and electrical equipment rather than chips or capital. Supplying a Brazilian deployment from a Brazilian factory takes ocean freight, port handling and import clearance out of the critical path for at least one of the two sites.

    Mexico is still a cross-border shipment, so the benefit is not symmetrical. But the same production line and the same platform serve both, and Munters’ Keith Dunnavant framed the company’s global manufacturing footprint as what allows it to support Aligned and OData as they expand across Latin America. A manufacturing presence in-region also tends to bring service technicians and spare parts inventory with it, which matters more for liquid systems than air ones: a leak or a pump failure in a liquid loop is a faster-moving problem than a failed fan.

    Neither company has quantified the plant’s output or said whether any of it is reserved for this project, so the supply-chain advantage is directional rather than measured.

    What It Signals About Latin American Colocation

    OData was founded in 2015 by the Brazilian private equity firm Patria Investments and acquired by Aligned Data Centers in May 2023. It runs facilities in Chile, Colombia, Brazil and Mexico with more in development. This deployment is, in effect, technology transfer from a North American operator into a regional platform it now owns — the same pattern the company followed with Delta Cubed on the air side.

    OData chief executive Ricardo Alário described the goal as establishing DeltaFlow as the market benchmark for liquid cooling in the region. That is an ambition rather than a disclosed market position, and it should be read as one. What the announcement does establish is concrete and checkable: a stated sum, two identified sites, a named manufacturing source and construction in progress. What it does not establish is demand. No anchor customer, pre-lease or contracted capacity has been disclosed, and a retrofit can be built ahead of committed tenants as easily as behind them.

    For competing operators in the same two markets, the practical effect is that rack density moves onto the specification sheet. Once one regional provider can quote liquid-cooled halls, requests for proposals start asking every provider what they can support per rack — and the answer for an air-only site is a hard ceiling that no amount of floor space compensates for.

    Background

    OData was founded in 2015 by Patria Investments, a Brazilian private equity firm, and grew into one of the region’s larger colocation platforms — operators that lease space, power and cooling to companies that prefer not to run their own facilities. Aligned Data Centers, a US operator known for developing its own cooling technology rather than buying it, acquired OData in May 2023. OData now runs facilities in Chile, Colombia, Brazil and Mexico with further sites in development, and has been the route by which Aligned’s in-house systems reach Latin American markets: the same two data centers receiving DeltaFlow already run Aligned’s Delta Cubed air cooling.

    The wider context is the shift in what a data center has to physically accommodate. For two decades, racks drew a few kilowatts each and chilled air sufficed. AI accelerators have pushed rack power an order of magnitude higher, past the point where air can economically carry the heat away. That has made cooling — historically an engineering afterthought to power and space — a gating factor on whether a given building can host AI workloads at all, and it has turned retrofits of existing, already-powered facilities into one of the faster routes to new AI-capable capacity.

    Sources

    Source: OData to deploy Aligned’s DeltaFlow liquid cooling system at Brazil and Mexico data centers — reporting on the $630 million deployment of Aligned’s DeltaFlow liquid cooling at OData’s SP04 facility in Brazil and QR03 facility in Mexico, including comment from OData and Munters.

  • Two Degrees Decide Whether an AI Cooling Unit Is 5 MW or 3 MW — and How Many You Buy

    Two Degrees Decide Whether an AI Cooling Unit Is 5 MW or 3 MW — and How Many You Buy

    TL;DR · 30-second read

    The Short Version

    Artificial intelligence runs inside warehouse-sized buildings packed with computers that get extremely hot. Getting that heat out has become one of the hardest and costliest parts of building them.

    On September 19 in Santa Clara, California, the Italian air-conditioning maker Clivet showed a new range of cooling equipment built for those buildings, including its largest heat-removal unit yet.

    The catch is in the fine print. That unit only reaches full strength if operators let the cooling water run slightly warmer. Ask it for water two degrees colder and it carries about 40 percent less heat — so you need more machines, and more floor space to put them on.

    Clivet said in a September 19 press release that it made the global debut of its next-generation magnetic levitation compressor at AI Infra Summit 2026 in Santa Clara, California, under a “DUO POWER BOOST” theme. The compressor uses a dual-rotor, two-stage architecture in which each rotor and impeller varies speed independently to suit different flow rates and pressure ratios.

    Alongside it, the company showed a data center cooling portfolio spanning air-cooled magnetic levitation chillers, a “magnetic CDU” that merges the cooling source with liquid-cooling distribution, high-capacity coolant distribution units, fan walls and terminal units. The largest CDU on display is rated up to 3 MW at a 3K approach temperature difference or 5 MW at 5K, with predictive control holding temperature fluctuation within ±0.5°C and key components replaceable online. Clivet also staged a joint power-and-cooling coordination demonstration with CLOU, which supplied energy storage and power delivery.

    Executive Summary

    This is a product debut rather than a contract announcement: no customer, no order book, no ship date. What makes it worth reading closely is a single line in the specification. The same coolant distribution unit — the box that takes heat from the water loop running through server racks and hands it to the building’s chilled-water system — is rated at 3 MW or 5 MW depending on nothing more than how close the two water temperatures are allowed to get.

    That gap, called the approach temperature difference, is normally treated as a plumbing detail. At AI rack densities it is a capacity decision. A two-kelvin change in the specification moves roughly 2 MW of heat rejection per unit, which translates directly into how many CDUs an operator buys, how much floor space they consume and how much pipework threads the room.

    The wider signal is a European HVAC manufacturer re-pointing core product lines — chillers, compressors, air handling — at AI thermal loads, and pairing them with an energy-storage partner so that power and cooling arrive as one conversation. Clivet has disclosed an architecture and a capacity envelope. It has not disclosed efficiency ratings, availability, pricing or a single deployment.

    The Spec Sheet, Not the Steel, Sets the Megawatts

    Read the CDU rating again: up to 3 MW at 3K approach temperature difference, or 5 MW at 5K. Approach temperature is the gap between the facility water a chiller plant delivers and the warmer technology-side water that actually circulates through cold plates on the processors. A heat exchanger’s duty scales with the temperature difference driving it, so widening the approach from 3K to 5K — a factor of 1.67 — lifts the rating from 3 MW to 5 MW, also a factor of 1.67. The ratings track heat-exchanger physics almost exactly. Nothing inside the cabinet changed.

    The operational consequence is the part buyers underestimate. An operator whose chip vendor mandates cold inlet water, and who therefore specifies a tight 3K approach, is buying a 3 MW machine at the price and footprint of a 5 MW one. Sizing a 30 MW hall needs ten units instead of six, with the extra floor area, valve sets, pipe runs, commissioning hours and maintenance headcount that go with them. Conversely, an operator who can accept warmer technology-side water recovers roughly 67 percent more capacity per unit — and typically unlocks more free-cooling hours upstream, because warmer facility water lets the chiller’s compressors idle for more of the year.

    That uplift is not free, and Clivet’s release does not price it. Moving 5 MW instead of 3 MW through the same unit means proportionally more flow at a given temperature rise, which means pump power, larger secondary pipework and higher pressure drop. Flow rates, pump curves and pressure-drop figures were not published. The honest reading is that the 5 MW headline number is real but conditional, and the condition is set by the silicon vendor’s inlet temperature limit, not by the cooling vendor.

    Why Magnetic Levitation, and Why the Compressor Is the Announcement

    A magnetic levitation compressor spins its impellers on magnetic bearings rather than oil-lubricated ones. Removing oil removes a chronic efficiency tax: over time, oil migrates into heat exchangers and films the tube walls, degrading heat transfer in exactly the equipment whose job is heat transfer. Oil-free machines also tolerate frequent starts and hold efficiency well at part load — which matters because a data center chiller plant almost never runs at design load.

    Clivet’s specific claim is architectural. The DUO POWER BOOST design puts two rotors in a two-stage compression path and lets each vary speed independently, so the machine can trade flow against pressure ratio as ambient conditions and IT load move. In principle that widens the efficient operating envelope, which is the right target for a facility whose load profile swings with training runs and whose lift changes with the seasons.

    What is substantiated here is a design description. What is not substantiated is performance. The release contains no coefficient of performance, no part-load efficiency figure such as IPLV or NPLV, no capacity range for the chillers, no refrigerant, no sound data and no third-party certification. Those are the numbers a consulting engineer actually specifies against, and until they exist the compressor is an architecture rather than a comparable product.

    Selling Power and Cooling as One Problem

    The joint demonstration with CLOU — energy storage and power delivery on one side, thermal management on the other — is commercially more interesting than it looks. Clivet’s own framing is that power configuration, thermal design, space layout and maintenance planning have to be settled together in a project’s earliest stages. That is a fair description of where high-density projects slip: the electrical and mechanical designs are often procured by different teams on different schedules, and the mismatch surfaces at commissioning.

    Packaging across that boundary is a recognisable vendor strategy in this market, and it is a response to buyer behaviour rather than a technical breakthrough. Operators building AI halls increasingly want a single party accountable for the megawatt from the switchgear to the cold plate, because integration risk is what delays revenue.

    The limits should be stated plainly. What was shown was a coordinated demonstration, not a disclosed commercial arrangement. Neither company published a joint product, a reference architecture, a territory split, an exclusivity term or a customer that has bought the combination. Until one of those appears, this is a booth, not a venture.

    A European HVAC Line Re-Aimed at AI Heat

    The portfolio’s shape is the story for the wider industry. Clivet showed facility-side air-cooled chillers, a magnetic CDU that folds the cooling source and the liquid distribution loop into one machine, high-capacity CDUs for secondary-side distribution, fan walls and terminal units — coverage from outdoor heat rejection to the rack. That is a general-purpose comfort-and-process HVAC catalogue being re-cut for a single vertical, and it puts Clivet into a tier already crowded with Vertiv, Schneider Electric’s Motivair, Boyd, nVent, Munters and Modine.

    The magnetic CDU is the genuinely differentiated item in the list, because combining chiller and distribution in one enclosure removes an interface — and interfaces are where commissioning time and finger-pointing live. It also concentrates risk, which is presumably why the release stresses online replacement of key components and ±0.5°C predictive control. Tight temperature stability matters for cold-plate systems because dew point and thermal cycling both sit close to the operating envelope; a loop that overshoots risks condensation on cold plates or stress on solder joints.

    For buyers, the practical takeaway is that the competitive question in this tier is no longer whether a vendor offers liquid cooling. It is whose ratings survive the approach temperature their silicon demands, and who will put efficiency and availability numbers behind the architecture.

    Background

    Clivet is an Italian heating, ventilation and air-conditioning manufacturer based in Feltre, in the Veneto region, and has been majority-owned by China’s Midea Group since 2016. Its traditional business spans chillers, heat pumps and air-handling equipment for commercial and industrial buildings — a catalogue it is now re-aiming at data centers, where thermal design has become a primary constraint on how much computing capacity a building can hold.

    AI Infra Summit, held in Santa Clara in the heart of Silicon Valley, has become a gathering point for the supply chain behind AI buildout: chips, power, networking and increasingly cooling. The shift to direct-to-chip liquid cooling has drawn in vendors from adjacent markets — industrial refrigeration, automotive thermal, building HVAC — because a coolant distribution unit is, at bottom, a pump, a heat exchanger and a controller, all of which these companies already build at scale. What they must prove to data center buyers is uptime, serviceability and certified performance, which is why the specification sheet, not the stage demonstration, is where these launches are ultimately judged.

    Sources

    Source: DUO POWER BOOST: Clivet Debuts Next-Gen Magnetic Levitation Technology and AI Cooling Portfolio at AI Infra Summit 2026 — the company’s September 19, 2026 announcement detailing its magnetic levitation compressor architecture, CDU capacity ratings and a joint power-and-cooling demonstration with CLOU.

  • SenseTime’s 80% Tokens-Per-Watt Gain Pulls the Grid Inside AI’s Efficiency Math

    SenseTime’s 80% Tokens-Per-Watt Gain Pulls the Grid Inside AI’s Efficiency Math

    TL;DR · 30-second read

    The Short Version

    Artificial intelligence computers draw electricity in sudden bursts, which can jolt the power grid the way a heavy factory switching on and off does. At a conference in Shanghai, Huawei showed gear meant to let large AI computing sites soak up those swings with batteries, and even help steady the grid instead of straining it.

    It also pushed a different scorecard. Instead of measuring how much power a building wastes, measure how much useful artificial intelligence work each unit of electricity buys. The Chinese firm SenseTime says it has improved that figure by 80 percent.

    Huawei used the HUAWEI AIDC Facility Summit, held alongside HUAWEI CONNECT 2026 in Shanghai on September 20, 2026, to launch what it calls a grid-interactive AIDC solution — a package of power and cooling equipment for AI data centers built around what the company labels “3+1” innovations. The announced version 1.0 combines a grid-friendly uninterruptible power supply, intelligent lithium batteries and grid-forming energy storage, alongside an AI-managed liquid cooling system that predicts the condition of its working fluids to schedule maintenance before failures. In the power domain, Huawei also outlined a longer-horizon “MIMO” power supply architecture.

    Hou Jinlong, Director of the Board of Huawei and President of Huawei Digital Power, opened the summit with four design arguments: pairing new power architecture with liquid cooling, using multi-layer hybrid storage to absorb local power fluctuations, coupling electricity and compute so data centers stop behaving as rigid loads, and shifting to modular, prefabricated delivery to cut construction lead times. Speakers from VNET Group and SenseTime shared the stage; SenseTime’s Lin Hai said end-to-end optimization had improved SenseCore’s tokens per watt by 80%.

    Executive Summary

    The substance of the announcement is architectural rather than financial. Huawei is proposing that an AI data center should be designed as a participant in the power system — able to hold its own voltage and frequency behavior steady through grid-forming storage, and to buffer the second-to-second swings that large training and inference clusters impose — rather than as a fixed block of demand that a utility must simply serve.

    That framing matters because grid connection, not construction, is increasingly what sets the calendar for AI capacity. A site that arrives at the utility as a controllable, power-quality-friendly asset is a different negotiation from one that arrives as raw megawatts of volatile load. Huawei’s pitch, delivered by Digital Power Vice President Bob He, is that as renewables raise the share of power-electronic sources on the grid, compute and energy infrastructure can no longer be planned independently.

    The second half of the pitch is a measurement change. Huawei and SenseTime both argued for scoring AI facilities on tokens per watt — useful model output per unit of electricity — instead of power usage effectiveness, the long-standing ratio of total facility power to IT power. That is a bigger change than it sounds, because it moves the boundary of the efficiency calculation outward to the grid and inward to the model at the same time.

    Why a Data Center Would Want to Look Like a Grid Asset

    A conventional data center is, from a utility’s point of view, a well-behaved but inflexible customer: it wants a large, constant, highly reliable supply and offers nothing back. AI clusters break the “well-behaved” half. Synchronized training jobs can ramp thousands of accelerators together, producing what the release describes as rapid and severe power fluctuations, and Huawei’s speakers named the consequences utilities worry about — disconnection events and unstable wideband oscillation, meaning sustained electrical ringing across a range of frequencies that neither the site nor the grid damps out.

    The engineering answer Huawei describes is layered storage sitting between the workload and the grid. Batteries and grid-forming inverters absorb the swing locally, so the fluctuation never reaches the interconnection point; grid-forming units go further by actively setting voltage and frequency rather than following the grid, which is what allows a site to support a network with a high share of inverter-based renewable generation instead of leaning on it. VNET Group Founder and Chairman Josh Chen extended the idea to campus scale, describing direct green power connections, active microgrids and direct-current distribution inside buildings as the way to serve gigawatt-class AI campuses.

    The commercial logic follows from scarcity. When interconnection capacity is the binding constraint on AI buildout, the cheapest new megawatt is often the one a developer does not have to request. Equipment that lets a site ride through disturbances, shave its own peaks and behave predictably at the meter is, in effect, a substitute for grid capacity that has not been built yet — and a bargaining chip in queue discussions. Huawei has not published evidence on how much connection time or capacity this saves in practice, which is the number that would turn the argument into a business case.

    Tokens Per Watt Redraws the Boundary of the Efficiency Question

    Power usage effectiveness has governed data center efficiency for two decades. It is a facility-boundary ratio: total power into the building divided by power reaching the IT equipment. It says nothing about whether the servers do anything useful. Once a site is spending most of its energy on inference, that blind spot becomes the main event — a facility can post an excellent PUE while running inefficient hardware on badly scheduled work.

    Tokens per watt closes that gap by measuring output rather than overhead, and this is where the 80% figure does its work. SenseTime’s Lin Hai described SenseCore’s improvement as end-to-end, spanning power plants and grids, rack services in the data center, compute cloud services and model services. That is the whole chain, not the mechanical plant. A number of that size is therefore not evidence that any single piece of equipment got 80% better; it is evidence that when you draw the measurement boundary that wide, most of the available gain sits in layers the facility team has never been scored on — and, at the far end, in the grid supplying the site. Huawei’s Hou Jinlong put the same point commercially: maximizing tokens per watt and minimizing cost per token will be the core competitiveness of what he called compute factories.

    Who this affects is concrete. Colocation operators sell space, power and a PUE commitment; they do not control the model, the scheduler or the chip, so they cannot be held to a tokens-per-watt number without a very different contract. Chip vendors and inference-software teams gain, because their contribution finally appears in the facility scorecard. Utilities gain a vocabulary for load that varies in usefulness as well as size. The catch is standardization: tokens differ by tokenizer, model and precision, so an 80% improvement is a credible internal engineering result and not yet a figure that can be compared across operators. No published methodology accompanied the claim.

    The Delivery Clock and the Question of Who Can Buy It

    Hou’s fourth argument — modular, prefabricated, product-based delivery to shorten lead time — is the least glamorous and arguably the most operationally consequential. In a market where the constraint is time to first megawatt, converting site-built electrical and cooling systems into factory-built blocks shifts the schedule risk from a construction crew to a production line. It also suits Huawei’s structural position: the company is selling the power chain and the cooling chain and the compute stack, and Bob He explicitly named end-to-end energy plus full-stack AI as the differentiator.

    That integration is genuinely hard for point-product vendors to match, and it is also the strongest argument against buying it. A grid-interactive architecture in which the uninterruptible power supply, the batteries, the grid-forming storage, the liquid cooling controller and the workload scheduler are co-designed is, by construction, difficult to unbundle later. Buyers evaluating it should ask what the interfaces are and whether any layer can be swapped — questions the launch materials do not address.

    Market context also shapes who this is for. The named practitioners on stage — VNET and SenseTime — are China-based, and the summit sat inside a conference where Huawei also presented an upgraded Stellar AI Network solution and released its Intelligent World 2035 report series setting out ten key directions. The ideas here travel further than the product does: measuring output per watt, buffering workload volatility on site and offering grid services are design responses any operator facing a long interconnection queue will recognize, whatever equipment ends up in the building.

    Background

    Huawei Digital Power is the Huawei division that sells power conversion, energy storage, solar inverter and data center power and cooling systems — a business adjacent to, but distinct from, the company’s networking and computing lines. Its data center pitch has increasingly merged with Huawei’s AI compute strategy, on the argument that the electricity system and the compute system are now a single engineering problem. HUAWEI CONNECT is the company’s annual flagship conference; the 2026 edition in Shanghai also produced an upgraded Stellar AI Network solution and a set of Intelligent World 2035 research reports laying out ten key directions.

    The wider context is a buildout constrained by electricity rather than silicon supply alone. Racks designed for AI accelerators draw many times what general-purpose servers did, pushing operators to liquid cooling and to far larger grid connections, while utilities in many markets face multi-year queues to study and approve new large loads. That has pushed the industry toward on-site generation, storage and demand flexibility, and toward efficiency metrics that account for useful work rather than facility overhead. VNET Group is a Chinese carrier-neutral data center operator; SenseTime is a Chinese AI company whose SenseCore platform runs its own intelligent computing centers.

    Sources

    Source: Huawei Grid-Interactive AIDC Solution: Shaping the New Paradigm of AI Infrastructure — Huawei’s September 20, 2026 announcement from the HUAWEI AIDC Facility Summit in Shanghai, detailing the grid-interactive AIDC 1.0 solution and remarks from Huawei, VNET Group and SenseTime executives.

    Primary sources: Huawei представляет инновационное решение в области иммерсивного телеприсутствия, которое предлагает эффект погружения и интеллектуальные услуги (Huawei’s September 20, 2026 immersive telepresence announcement from the same conference); Huawei представляет обновленную сеть Stellar AI для обеспечения безопасного и интеллектуального подключения к агентскому миру (the upgraded Stellar AI Network solution presented at HUAWEI CONNECT 2026 on September 18, 2026); В последнем отчете Huawei предлагаются 10 ключевых направлений Интеллектуального мира-2035 (Huawei’s September 18, 2026 launch of its Intelligent World 2035 report series).

  • Equinix’s $5B-$7B AI Buildout Shifts the Colocation Bottleneck From Cabinets to Megawatts

    Equinix’s $5B-$7B AI Buildout Shifts the Colocation Bottleneck From Cabinets to Megawatts

    TL;DR · 30-second read

    The Short Version

    Equinix rents out space, electricity and network connections inside the buildings where companies keep their computers. It says it now plans to spend 5 to 7 billion dollars every year building more of them, because demand for artificial intelligence computing is climbing.

    That is bridge-and-power-plant money, not server-room money. And the hard part is no longer pouring concrete. It is finding enough electricity, the heavy gear that delivers it, and the cooling for machines that run far hotter than the ones they replace.

    Yahoo Finance reported that Equinix plans an annual data center buildout of $5 billion to $7 billion, citing accelerating artificial intelligence demand as the driver. Equinix is one of the largest operators of carrier-neutral data centers worldwide, and the figure describes a sustained annual rate of construction spending rather than a single project.

    No individual site, market, customer or completion date accompanied the spending range. The disclosure sets a capital pace; it does not yet say where the capital lands.

    Executive Summary

    A $5 billion to $7 billion annual construction budget is a statement about build rate, not about any one building. For a company whose historical identity is retail colocation — selling cabinets and cross-connects to thousands of enterprises, carriers and cloud on-ramps inside dense interconnection hubs — spending at that tempo implies a different physical product: fewer, larger, denser halls built for accelerated computing.

    That matters because retail colocation and artificial-intelligence-scale capacity consume capital in opposite ways. Retail capacity is built incrementally and monetised through recurring interconnection revenue that compounds as ecosystems thicken. AI-scale capacity is built in large power blocks with long lead items and is monetised mainly through space and power. Committing this much annually pushes the mix toward the second without abandoning the first.

    For the rest of the market, the number functions as a pace-setter. Interconnection density is a cumulative advantage: the value of being inside a hub rises with how many networks and clouds are already there. An operator that builds through the AI cycle at this rate is also defending that density. Competitors that build more slowly risk seeing new ecosystems form somewhere else.

    The Bottleneck Moves From Cabinets to Megawatts

    Colocation growth used to be gated by commercial motion: sales cycles, fit-out of cabinets, provisioning of cross-connects — the short patch cables that link one tenant to another inside a facility and generate high-margin recurring revenue. Those constraints scale with people and process. At a $5 billion to $7 billion annual spend, they are no longer the binding limit. The limits become physical and external: grid interconnection, where a site waits in a utility queue for the right to draw large loads; long-lead electrical equipment such as transformers, switchgear and generators; land that already has power or a credible path to it; and the plumbing, heat-rejection and manufacturing capacity behind liquid cooling, which accelerated servers increasingly require because air cannot carry the heat away fast enough.

    The mechanism is straightforward. Capital converts into revenue only when a hall is energised. If a company can deploy several billion dollars a year but its sites cannot be energised on that schedule, the money queues rather than compounds. That is why a spending number of this size is best read as a procurement commitment: to sustain it, an operator has to be placing equipment orders and power arrangements well in advance of demand it has not yet contracted.

    The people affected sit downstream. Utilities and grid operators see a customer asking for large, firm, long-duration load. Electrical-equipment manufacturers and engineering and construction firms see order books extended. Cooling vendors see demand shift from air handling toward liquid distribution. And enterprise buyers of ordinary colocation may find that in the most power-constrained metros, the scarce resource they are competing for is no longer floor space but the megawatts behind it.

    What a Build Rate Says That a Project Never Does

    Single-project announcements are easy to make and easy to quietly defer. A recurring annual range is a harder statement, because it implies a pipeline of land, power and permits deep enough to absorb that capital every year. It also implies a funding plan that repeats annually rather than once.

    This is where structure matters. Equinix operates as a real estate investment trust, a tax structure that requires distributing most taxable income to shareholders. That limits how much cash can simply be retained and recycled into construction, and pushes large sustained programmes toward some combination of debt, equity issuance, asset recycling and joint ventures with institutional partners — the model the industry has used for hyperscale-scale halls, where a partner supplies much of the capital and the operator supplies development and management.

    The economics then hinge on the spread between the cost of that capital and the stabilised yield on newly built capacity. AI-scale halls generally sell power at scale with less interconnection revenue attached per megawatt than a mature retail hub produces. Building a lot of them can raise revenue while diluting the blended margin profile, unless the operator uses that capacity to pull more networks and tenants into the same campuses. Whether this spend is accretive is a question about yields and funding mix, not about the headline figure.

    Interconnection Density Is the Asset Being Defended

    The strategic logic behind spending through an AI cycle is not only that AI capacity is rentable. It is that where compute lands, networks follow, and where networks concentrate, the next tenant has a reason to choose the same campus. Interconnection density compounds: each additional carrier, cloud on-ramp and peering participant makes the location more valuable to the next arrival. It is slow to build and correspondingly slow to erode.

    Artificial intelligence workloads complicate that picture because training clusters can sit far from population centres, chasing cheap power rather than low latency. Inference — running trained models for users — behaves more like traditional cloud traffic and rewards proximity to networks and data. An operator with existing dense hubs has a real advantage in the second category, provided it can add power-dense capacity in or near those hubs rather than only in remote, power-rich locations.

    For competitors, including other large global platforms and the wholesale developers serving hyperscale demand, the practical consequence is timing. Power in the strongest metros is allocated years in advance. A rival that matches this pace later does not merely arrive late to the same opportunity; it arrives to a smaller pool of available grid capacity in exactly the places where interconnection value is highest.

    Background

    Equinix is one of the largest carrier-neutral data center operators in the world, with facilities across major metropolitan markets in the Americas, Europe and Asia-Pacific. Carrier-neutral means it is not owned by a telecom operator, so tenants can connect to any network present in the building. That neutrality is the foundation of its interconnection business: over decades, the densest sites accumulated carriers, internet exchanges, cloud on-ramps and enterprise tenants, each arrival making the next one more likely.

    The industry’s demand base has shifted repeatedly — from enterprise server consolidation, to cloud, and most recently to accelerated computing for artificial intelligence, which consumes power at densities the previous generations of halls were never designed for. That shift has changed the economics of construction: the scarce inputs are increasingly electricity, long-lead electrical equipment and cooling capacity rather than land or building shell, and capital commitments are being made years ahead of the demand they serve.

    Sources

    Source: Equinix Plans $5B-$7B Annual Data Center Buildout as AI Demand Accelerates — Yahoo Finance report on Equinix’s planned annual data center construction spending amid accelerating AI demand.

  • CoreWeave Brings Up Multi-Rack Vera Rubin NVL72: What It Asks of Operators

    CoreWeave Brings Up Multi-Rack Vera Rubin NVL72: What It Asks of Operators

    TL;DR · 30-second read

    The Short Version

    CoreWeave, a company that rents out computing power for artificial intelligence, says it has switched on several of Nvidia’s newest server cabinets and wired them together to work as one giant machine.

    Each cabinet packs 72 of Nvidia’s top chips. Linking hundreds of those chips lets companies build and run bigger, more capable digital assistants.

    Why it matters: machines this crowded need a lot of electricity and liquid cooling, so the buildings that house them have to change quickly. CoreWeave has not said who is using the system or when customers can rent it.

    CoreWeave (Nasdaq: CRWV) has brought up a multi-rack NVIDIA Vera Rubin NVL72 cluster on CoreWeave Cloud, joining hundreds of NVIDIA Rubin GPUs into a single scale-out cluster aimed at agentic AI, Light Reading reported on September 16, 2026, publishing the company’s announcement. “Bring up” is the industry term for powering on, configuring and validating new hardware so it can run real workloads.

    Alongside the compute news, CoreWeave introduced two additions to CoreWeave AI Object Storage: cross-region write acceleration, which lets a job write data locally while it is copied to another region in the background, and a new Archive tier for colder data.

    Executive Summary

    Each Vera Rubin NVL72 rack combines 72 Rubin GPUs, 36 Vera CPUs, NVIDIA’s NVLink 6 chip-to-chip interconnect, ConnectX-9 SuperNICs (high-speed network cards) and BlueField-4 DPUs (processors that offload networking and security tasks). CoreWeave says it has joined several of these racks over NVIDIA Spectrum-X Ethernet so they behave as one cluster for training, inference and reinforcement learning. The company attributes the result to automation in its Mission Control software, including a Rack LifeCycle Controller that handles hardware detection, firmware, validation, power and cooling.

    The significance is less about any single rack than about what comes with it. Rack-scale systems shift the unit of deployment from the server to the entire cabinet, and that concentrates power, heat and networking demands in a way that tests data center design. CoreWeave is positioning itself as an early operator of NVIDIA’s next generation, and signaling that speed of bring-up is now a competitive metric in its own right.

    The announcement is detailed on architecture but light on the facts that determine commercial impact: how many racks are running, where, for whom, and when customers can buy the capacity.

    The Rack Is the New Computer

    For most of cloud computing’s history, the basic building block was a server: a box with a few processors that could be installed, swapped and scaled one at a time. NVIDIA’s NVL72 designs change that. Inside one rack, 72 GPUs are tied together by NVLink so they can share work almost as if they were a single, very large chip. That is known as scale-up. Joining multiple racks over a network, as CoreWeave describes with Spectrum-X Ethernet, is scale-out.

    The combination matters because the largest AI models and the most demanding inference jobs no longer fit comfortably inside one rack. CoreWeave frames the multi-rack cluster around agentic AI, meaning systems that chain many model calls and tool uses together to complete a task. Its argument is that delays compound across those steps, so tightly coupled compute and fast data access pay off. That reasoning is plausible, though the company did not publish benchmarks showing how much faster agentic workloads run on the new cluster.

    The networking claims are ambitious. CoreWeave says each Rubin GPU gets two ConnectX-9 SuperNICs, providing 1.6 terabits per second of scale-out connectivity, and that the fabric supports roughly 128,000 GPUs per rail in a non-blocking design, meaning traffic should not queue for lack of capacity. That figure describes what the network architecture can accommodate, not what has been deployed; the cluster announced here is described as hundreds of GPUs.

    Deployment Speed Becomes a Competitive Weapon

    The most revealing part of the announcement is how much of it is about process rather than silicon. CoreWeave describes racks arriving as hardware that still has to be connected and validated, and says its Rack LifeCycle Controller automates that work, with software components it calls Racky handling rack control and Valvey executing cooling actions. Racks then go through NVIDIA field diagnostics plus full-rack workload testing, and only those that pass as a system move into production.

    This matters economically. Every week a costly rack sits installed but unvalidated is capacity that earns nothing while its financing costs keep accruing. For GPU cloud providers competing for the same scarce next-generation hardware, the ability to turn delivered racks into billable capacity quickly can be as important as securing the allocation in the first place. Automating firmware updates, power sequencing and cooling control is how an operator compresses that interval.

    The claim that “every GPU performs at its best” is marketing language and should be read as such. What the release does substantiate is a described validation workflow; it does not provide failure rates, burn-in durations or time-to-production figures that would let buyers compare CoreWeave’s process with competitors’.

    What Next-Generation Racks Demand of Facilities

    CoreWeave explicitly lists cooling and power among the layers that had to be engineered for the cluster to perform as one system, and it built a dedicated software component to execute cooling actions. That is a clear signal of where the operational burden now sits. Dense rack-scale AI systems of this class have relied on liquid cooling, piping coolant directly to the chips because air alone cannot remove heat fast enough.

    For data center developers and colocation providers, the implication is that readiness for the rack era is not just about floor space. It requires facility cooling loops, electrical distribution designed for highly concentrated loads, and structural and networking plans that allow new racks to be added without rework. CoreWeave’s point that its modular network topology lets racks be added without redesigning the fabric at each expansion speaks to that last requirement.

    Operators whose buildings were designed for traditional air-cooled servers face costly retrofits or the prospect of being passed over for this generation of hardware. Suppliers of liquid-cooling equipment, power distribution gear and high-capacity optical networking stand to benefit, while the release gives no per-rack power figures that would let facility planners size requirements precisely.

    Storage Moves Closer to the Loop

    The storage announcements address a less glamorous bottleneck: GPUs sitting idle while they wait for data. With cross-region write acceleration, CoreWeave says a job writes locally while replication to another region happens in the background, so intermediate results, retrieved context and outputs are not held up by long-distance transfers. The new Archive tier gives customers a place to keep colder data within the same storage service.

    Background replication is a well-established technique, and its value depends on details the company has not provided, such as how far replicated copies can lag, how conflicts are handled and what happens if a region fails before replication completes. For buyers, the practical question is whether keeping training data, checkpoints and archives within one provider’s storage stack is worth the added dependence on that provider.

    Background

    CoreWeave is a Nasdaq-listed cloud provider (ticker CRWV) that specializes in renting GPU computing capacity for artificial intelligence workloads, marketing itself as “The Essential Cloud for AI.” Unlike general-purpose clouds, its business is built around large clusters of NVIDIA accelerators plus the networking, storage and orchestration software needed to run them for AI training and inference.

    NVIDIA’s Vera Rubin platform is the successor to its Grace Blackwell generation, pairing Rubin GPUs with Vera CPUs. The NVL72 designation refers to rack-scale systems in which 72 GPUs are interconnected with NVLink, a design approach that has moved AI infrastructure from individual servers toward entire racks engineered as single units, with corresponding increases in the power, cooling and networking demands placed on data centers.

    Sources

    Source: CoreWeave brings up multi-rack NVIDIA Vera Rubin NVL72 cluster — Light Reading, September 16, 2026: CoreWeave announces a multi-rack Vera Rubin NVL72 cluster and new AI Object Storage capabilities.

  • Bitdeer’s 65.1MW Malaysia Deal Tests a Bitcoin Miner’s AI Cloud Pivot

    Bitdeer’s 65.1MW Malaysia Deal Tests a Bitcoin Miner’s AI Cloud Pivot

    TL;DR · 30-second read

    The Short Version

    Bitdeer, a company that got its start mining bitcoin, is betting heavily on renting out powerful computers used to build and run artificial intelligence.

    It signed a 10-year deal for a new building in southern Malaysia that will use enough electricity to run tens of thousands of homes. If the building earns what a similar Bitdeer site does, it could bring in about $1.1 billion a year.

    That yearly figure is about two-thirds of what the whole company is worth on the stock market today. The catch: the power does not switch on until the second half of 2027.

    Bitdeer Technologies Group’s AI cloud division, Bitdeer AI, has signed a 10-year service agreement for A202, a 65.1-megawatt data center in Johor Bahru, Malaysia, Cryptopolitan reported. The facility sits on the same campus as Bitdeer AI’s existing 21.7MW A201 site, bringing the campus to 86.8MW of critical IT load (the power available to the computing equipment itself, excluding cooling and other overhead). Power is expected in the third quarter of 2027, and the addition lifts Bitdeer AI’s secured AI cloud capacity to about 206.5MW, roughly 59% of its target of up to 350MW by the first quarter of 2028.

    Bitdeer expects A202 to generate revenue per megawatt similar to its A102 facility, where contracts covering the next five years carry more than $800 million in expected revenue across 9.5MW. Applied to A202’s size, that works out to about $1.1 billion a year, the figure Cryptobriefing highlighted, while blockchain.news reported that the deal pushes Bitdeer’s AI pipeline toward $7 billion. Shares were quoted at $11.97, for a market capitalization of about $1.64 billion, though Stocktwits reported the stock slipped as a broader tech sell-off weighed on the sector.

    Executive Summary

    Bitdeer’s A202 agreement is the company’s largest single capacity addition in Southeast Asia and its clearest signal yet that a business built on bitcoin mining intends to become a long-term provider of AI computing. The 65.1MW facility is designed for liquid-cooled, rack-scale NVIDIA systems, including the GB300 NVL72 and the upcoming Vera Rubin platform, and will support both GPU cloud services (renting access to AI chips) and data-hosting workloads.

    What makes the announcement significant is the combination of scale and funding model. Bitdeer says it aims to have customer prepayments cover more than half of the capital spending for each facility, with the rest financed from contracted and operating cash flow. If that works, the company can grow its AI business without leaning as heavily on new debt or share issuance. If demand or prepayments fall short, the revenue math behind the headline numbers becomes considerably less certain.

    For the wider market, the deal adds another data point that operators with existing land, power and cooling infrastructure are competing hard for the next generation of AI hardware, and that Johor remains a focal point for that capacity in Southeast Asia.

    The Revenue Math Behind the $1.1 Billion Figure

    The headline revenue number is a derived estimate, not a disclosed contract value. Bitdeer’s A102 facility has more than $800 million in expected revenue over five years across 9.5MW, which is roughly $17 million per megawatt per year. Bitdeer says it expects A202 to earn similar revenue per megawatt; multiplied across 65.1MW, that yields approximately $1.1 billion annually once the facility is fully running.

    That per-megawatt figure is high because GPU cloud is a different business from traditional data center leasing. A landlord that rents out space, power and cooling earns rent on the building. A GPU cloud operator also owns the chips and sells computing time on them, so revenue per megawatt is far higher, but so is the capital required. The comparison with Bitdeer’s roughly $1.64 billion market capitalization is striking, but it sets potential future revenue against current equity value and says nothing about the hardware costs, operating expenses or margins behind that revenue.

    The estimate also rests on an assumption that A202 pricing will match A102’s. A102 is a 9.5MW deployment with contracts covering five years; A202 is nearly seven times larger, arrives in late 2027 on newer hardware, and sits inside a 10-year agreement. Pricing for AI computing can shift quickly as new chip generations arrive and supply catches up with demand.

    Prepayments as the Financing Engine

    Bitdeer’s stated plan is for upfront customer payments to cover more than half of each facility’s capital spending. Prepayment-funded buildouts have clear appeal: they reduce the need to raise outside capital, and a customer willing to pay in advance is signaling genuine commitment. For a company whose legacy business is exposed to bitcoin price swings, that is a meaningful source of discipline.

    The trade-off is concentration risk. Prepayment structures typically depend on a small number of large customers with the balance sheets to pay ahead, and the remaining funding depends on contracted and operating cash flow materializing on schedule. Over a 10-year horizon, the hardware inside A202 will likely be refreshed at least once, raising the question of who funds the next generation of chips after GB300 and Vera Rubin systems age.

    The Brownfield Advantage in Johor

    A202 is being built on land Bitdeer AI already controls, next to the existing A201 facility. That lets the company extend power, liquid-cooling and network infrastructure it has already built, which Bitdeer says should shorten the time to full operation. In an industry where grid connections and specialized cooling are often the slowest parts of a project, expanding an existing campus is a genuine competitive edge.

    Liquid cooling matters here because rack-scale systems like the GB300 NVL72 pack so much computing power into a single rack that air cooling cannot remove the heat effectively. Facilities built for these systems are scarcer than general-purpose data center space, which supports CFO Michael G. Potter’s claim that demand for “liquid-cooled, rack-scale AI Cloud capacity” for 2027 is outpacing supply. The claim is backed by A102 having been fully booked before it was energized; A201, by contrast, is in advanced negotiations rather than signed contracts.

    From Miner to AI Landlord: Halfway There

    With 206.5MW secured across Malaysia, Norway and the United States, Bitdeer is about 59% of the way toward its target of up to 350MW of AI-ready capacity by early 2028. Status as a preferred NVIDIA Cloud Partner gives it credibility in hardware allocation, a practical advantage when the newest systems are in short supply.

    The company is running a hybrid model rather than abandoning mining, which keeps an existing revenue stream but also means two capital-hungry businesses competing for the same balance sheet. The market’s reaction captured the tension: a transformative-sounding pipeline, but a share price still moving with sentiment across the broader technology sector. Investors will judge the pivot less on secured megawatts than on how many of those megawatts turn into signed, paying contracts.

    Background

    Bitdeer Technologies Group is listed on Nasdaq under the ticker BTDR. It began as a bitcoin mining operation founded by Jihan Wu, the former CEO of Bitmain, one of the best-known makers of bitcoin mining machines. Like several miners, Bitdeer has sought to repurpose its experience securing large amounts of power and running dense computing facilities toward AI infrastructure, placing that effort in a separate division, Bitdeer AI, while continuing to mine.

    Bitdeer AI’s capacity spans Malaysia, Norway and the United States, and the company has been named a preferred NVIDIA Cloud Partner. Its Malaysian campus is in Johor Bahru, across the causeway from Singapore, a location that has drawn significant data center investment as demand for AI computing in Southeast Asia has grown.

    Sources

    Source: Bitdeer signs 10-year deal for 65.1MW AI data center in Malaysia (Cryptopolitan), covering Bitdeer AI’s A202 agreement in Johor Bahru, its revenue expectations, prepayment funding plan and 350MW capacity target.