Immersion cooling tank acceptance field guide for AI data centers

Immersion cooling, and what acceptance proves

Immersion cooling submerges servers in a tank of dielectric fluid, a non-conductive liquid that carries heat off the components by direct contact instead of blowing air across them. The fluid touches the chips, the boards, and the power supplies, and because liquid holds far more heat per unit volume than air, a tank can cool densities no air-cooled rack can reach. AI and HPC hardware now puts out more heat than you can move with any volume of air you can push through a chassis, and immersion is one of the answers.

Tank acceptance is the commissioning work that proves the tank, the fluid, the heat rejection, and the life safety are all correct before a single server goes into the bath. It is not the same as accepting a rack. Once hardware is submerged you cannot get back to most of it without pulling it dripping out of the fluid, so the things that are hard to fix later get proven up front: the floor load under a full tank, the fluid chemistry, the leak path, and the fire-code treatment.

The discipline reorders around the fluid. Everything that makes immersion work, the direct contact, the heat capacity, the silence of a tank with no fans, also makes it unforgiving. Wrong fluid, an incompatible material left on a server, a floor that was never checked for the load, or a fire-code question nobody asked the authority having jurisdiction, and the problem shows up after the hardware is wet and worth more than the building system around it. The cooling pillar covers the air and chilled-water side of the hall. This guide stays on the immersion tank and what it takes to accept one.

What is the difference between single-phase and two-phase immersion cooling?

Single-phase and two-phase immersion differ in whether the fluid boils. In single-phase immersion the dielectric fluid stays liquid the whole time. A pump, or natural convection as the warm fluid rises, moves it past the hot components and over to a heat exchanger, and the fluid never changes state. It is the more common deployment today, and the tank runs open or loosely covered because nothing is evaporating.

Two-phase immersion lets the fluid boil right at the chip. The fluid is chosen with a low boiling point, so the heat off a hot processor turns it to vapor at the surface. The vapor rises, hits a condenser coil in the top of the tank, gives up its heat, condenses back to liquid, and rains back down. It carries heat by the latent heat of vaporization, which is a large amount of heat per unit of fluid, and it needs a sealed tank to keep the vapor in. That sealing is also what makes two-phase harder to service. You cannot just lift a lid.

The distinction matters at acceptance because the two run on different fluids and carry different regulatory weight. Verify which one is in front of you before you write a procedure, because the leak mode, the fire treatment, and the fluid handling all change with it. As of this 2026 review, single-phase is the more widely deployed of the two on new large jobs, and two-phase carries an open regulatory question the next section covers. Confirm the architecture against the equipment submittal rather than assuming.

Two-phase fluids and the PFAS question

The fluids that made two-phase immersion work have been fluorinated liquids, and those fluids are under heavy regulatory pressure. The low-boiling dielectrics two-phase relied on, the engineered fluorocarbons sold under names like Novec and Fluorinert, are per- and polyfluoroalkyl substances, the PFAS family often called forever chemicals because the carbon-fluorine bond does not break down in the environment.

The supply side moved first. 3M, the dominant maker of these fluids, announced it would exit PFAS manufacturing, and industry reporting puts the last orders for those Novec fluids around early 2025 with production winding down after. That sits on top of tightening EPA and international PFAS rules and a multi-billion-dollar settlement over water contamination. The practical effect is that the fluid two-phase immersion was built around is becoming hard to source, and a clean replacement that boils where you need it, stays non-conductive, and is not a regulated PFAS has not clearly arrived.

So treat a two-phase proposal with that question open. Confirm the fluid, its long-term availability, and its regulatory status before the design is locked, because a tank you cannot get fluid for in five years is a stranded asset. This is a fast-moving area and the specifics shift, so verify the current fluid availability and the PFAS rules in the project's jurisdiction rather than trusting any single snapshot. Single-phase hydrocarbon fluids do not carry the same PFAS exposure, which is part of why most new large deployments have leaned single-phase.

The dielectric fluids: synthetic, hydrocarbon, and the properties that matter

The fluid is the heart of an immersion tank, and it is not yours to pick on the jobsite. Single-phase immersion runs mostly on hydrocarbon and synthetic dielectric fluids: mineral-oil-based fluids, synthetic hydrocarbons such as polyalphaolefins and gas-to-liquid base stocks, and synthetic esters. They are non-conductive, they carry a high flash point, and they are chosen for how they behave across years in contact with hot electronics. Two-phase runs on the low-boiling fluorinated fluids covered above.

A handful of properties decide whether a fluid is right, and they pull against each other. Dielectric strength is the fluid's ability to resist conducting electricity, and it has to stay high or the fluid stops being safe around live boards. Contamination, water, and broken-down thermal grease all drag it down. Viscosity sets how easily the fluid moves heat and how hard the pumps have to work, and it tends to climb with flash point, so a safer high-flash fluid is often a thicker one that needs more pumping. Thermal capacity and conductivity set how much heat the fluid carries off the silicon. Flash point sets the fire classification the life-safety section covers.

The numbers belong to the fluid manufacturer's data sheet, not to memory or a rule of thumb. At acceptance you confirm the delivered fluid matches the specified fluid by name and grade, and you pull a baseline sample so there is a documented starting point for dielectric strength, water content, and acidity to trend against later. A fluid that arrives off-spec, or gets cross-contaminated with the wrong product during fill, is a problem you want to catch in the drum, not in the tank.

Material compatibility: what the fluid attacks

A dielectric fluid is a solvent, and it spends years in contact with everything in the tank, so material compatibility is a real acceptance concern, not a footnote. The fluid can swell, soften, or extract certain plastics and elastomers. It can lift adhesives and float labels off equipment. It can wick into porous materials. Standard thermal grease is the classic casualty: the fluid breaks it down, which both ruins the thermal joint and contaminates the fluid, dropping its dielectric strength as the dissolved grease loads up.

The compatibility list lives with the fluid and the equipment vendors, and it is specific. Common guidance points to compatible elastomers such as EPDM or fluoroelastomers for seals and warns off materials a particular fluid attacks, but which materials are fine and which are not depends on the exact fluid chemistry. A material that shrugs off a synthetic hydrocarbon may degrade in an ester. Optical components are a known watch item, because fluid can wick along a fiber and into the coatings, which is why immersion deployments use transceivers and cables rated for submersion rather than standard parts.

At acceptance this shows up two ways. You confirm the tank's own wetted materials, its seals, gaskets, hoses, pump components, and the heat-exchanger coil, are on the fluid's compatibility list. And you confirm the IT hardware was prepared for the fluid, which is the next section. Get a label or a gasket wrong and you do not find out at commissioning. You find out months later when the fluid has darkened, the dielectric strength has slid, and you are chasing why.

What has to change on a server before immersion

A standard server is built to be cooled by air, and several of its parts do not belong in a fluid bath, so the hardware gets prepared before it goes in. This is vendor-governed work, and the immersion fluid maker and the server OEM both have a say, but the recurring changes are consistent enough to know going in.

Thermal interface material comes first. The standard thermal grease between a chip and its heat sink gets replaced, commonly with an indium foil, a thermally conductive epoxy, or a solder joint, because the fluid dissolves ordinary grease, wrecks the thermal path, and contaminates the bath. Fans come out. They do nothing useful in a liquid and only add drag and waste power, and the firmware usually has to be told not to alarm or shut down on the missing fans, a BIOS or BMC change that is easy to forget until a node refuses to boot. Air baffles and shrouds, the plastic ducting that steered air through the chassis, come out so the fluid can move freely.

Spinning hard drives are a problem because they are vented and the fluid gets into them. The usual move is sealed drives or a swap to solid-state. Optical transceivers and active optical cables need immersion-rated versions, since standard ones can wick fluid. Some labels and adhesives have to be swapped for fluid-compatible types so they do not float off and foul the bath. The point at acceptance is easy to state and easy to skip: confirm every server going into the tank was prepped to the vendor's immersion procedure, because one node with the wrong thermal paste degrades the fluid for the whole tank.

Structural and floor loading under a full tank

A full immersion tank is heavy, and the floor under it is the first acceptance check, not an afterthought, because it is the one thing you cannot fix after the fluid is in. The fluid alone is a large mass, and with the tank, the hardware, and the fluid together a single immersion tank can weigh several times the air-cooled rack it replaces. Industry figures put a full tank in the rough range of a couple of thousand pounds and up, but the real number is the tank's filled weight on its actual footprint, which is the data sheet figure that matters.

The check is a structural one, and it belongs to the structural engineer, not the mechanical contractor's eyeball. The concern is point load and distributed load. A heavy tank on a small footprint concentrates the load, and on a raised access floor the question is whether the floor system and the slab below carry the filled tank plus a margin. Older raised floors in particular were never designed for immersion loads and often need reinforcement or a structural slab. The floor-loading limits for the room are their own topic worth checking against the building's structural documents.

Verify the filled-tank weight against the manufacturer's data, the footprint and any pedestal or load-spreader the design calls for, and the floor's rated capacity from the structural engineer of record. This is a place to be blunt. A tank that overloads a raised floor is not a deficiency you correct with a punch-list item. Confirm the load path before the tank is set, because once it is filled the only way to relieve it is to drain it.

Tank acceptance: leak integrity, fill, and the loop to facility water

With the floor proven, the tank itself gets accepted as an assembly: the vessel, the fill, the circulation, and the heat exchanger that ties it to the building. Start with leak integrity. The tank holds a standing volume of fluid for years, so every seam, drain valve, fitting, and the heat-exchanger penetration has to be proven tight. Depending on the design this is a fluid-level hold, a pressure or leak check on the piped portions, and a visual on the welds and seals. The piped secondary side that carries fluid to a heat exchanger or pump falls under the same pressure-test logic as any cooling loop, governed by the applicable piping code by topic and the project spec.

The circulation and heat rejection get checked next. A single-phase tank moves fluid past the hardware and over to a heat exchanger, either with a pump and an external coolant distribution unit or by convection within the tank, and a two-phase tank relies on the condenser coil and the boil. Confirm the pump, the flow path, and the heat exchanger or CDU match the design, and that the secondary loop to facility water is connected, filled, and isolated correctly. The interface to the building chilled water is the same primary-loop relationship the liquid cooling loop guide covers in depth, and the OCP immersion work defines that facility-water interface by topic.

Fill level and trapped air round it out. The fluid has to cover the components the design says it covers, at the level and with any expansion volume the manufacturer specifies, and the loop has to be vented so an air pocket does not sit against a hot board or starve a pump. Confirm the level cold and again at temperature, because fluid expands as it warms and a level that looked right cold can change.

Fluid handling, filtration, and the baseline sample

The fluid has to go into the tank clean, and it has to stay accounted for, so fluid handling is its own acceptance step. New fluid is not automatically clean enough. It can carry water, particulate, or the wrong product from a mislabeled drum. The fill is done through filtration to the manufacturer's particulate target, and the transfer is kept clean so the fill itself does not introduce contamination. Plain shop water, a dirty pump, or a hose used for something else first are how a clean fluid arrives dirty.

The baseline sample is the record that makes the rest of the tank's life legible. Pull a sample of the fluid as filled and document the starting dielectric strength, water content, acidity or acid number, and appearance against the manufacturer's acceptance values. That baseline is what tells operations, two years on, whether a fluid reading is a new problem or normal aging, and it is the question you cannot answer later if nobody captured it at the start. The fluid is a maintenance item the owner inherits, and it begins with this number.

Spill containment closes it out, because the volume in an immersion tank is large. Confirm the containment, bunding, or drip provisions sized to the fluid volume the tank holds, and the spill-response materials and procedure, so a leak or an overfill has somewhere to go that is not the floor and the room below. The fluid volume per tank is a number worth knowing for both the containment and the fire-code conversation.

Heat rejection and the warm-fluid advantage

The heat the fluid picks up has to leave the tank, and how it leaves is a core acceptance check. In a single-phase tank the warm fluid passes through a heat exchanger, a coil or a plate exchanger, that hands the heat to the facility water loop, often through a coolant distribution unit with its own pump and controls. In a two-phase tank the vapor condenses on a coil in the top of the tank and the condenser rejects that heat to facility water. Either way there is a heat exchanger and a facility-water interface, and that interface is where you confirm flow and temperature.

The number that tells you the heat path is healthy is the approach temperature, the gap between the facility water coming in and the fluid leaving the exchanger. A wide or widening approach means the exchanger is undersized, fouled, or starved of flow, and it shows up as the tank fluid running warmer than design even when the facility water is fine. Record the design approach and the as-found flow at acceptance so operations has a baseline to trend, the same way the liquid cooling loop guide treats the CDU approach.

Immersion shares the warm-water advantage that makes liquid cooling pay off on energy. The fluid does not need to be cold to cool the chips, because the temperature difference between a hot junction and a warm fluid is still large, so the facility can often reject the tank's heat to the outdoors with a dry cooler or an economizer for much of the year instead of running mechanical chillers. The supply temperature the design targets, tied to ASHRAE TC 9.9 liquid-cooling guidance and the chip vendor's limits, is what you commission to, not the coldest the plant can make.

What does fire code require for an immersion cooling tank?

Fire and life safety turn on the fluid's flash point and how the fire code classifies it, and it is a question to put to the authority having jurisdiction early, not at inspection. Single-phase dielectric fluids are combustible liquids, not flammable ones, because their flash points are high. But combustible is not the same as ignored. They burn if you get them hot enough, and an open tank holds a large standing volume of the stuff.

The data center fire standard treats this directly. Recent editions of NFPA 75, the standard for the protection of information technology equipment, set a flash-point floor for immersion-cooling fluids: the insulating liquid is to be noncombustible or have a closed-cup flash point at or above 135 C, which is 275 F, a stricter bar than the generic combustible-liquid cutoff. NFPA 30, the flammable and combustible liquids code, classifies these high-flash fluids as Class IIIB combustible liquids, and the International Fire Code adopts NFPA 30 by reference. The exact section numbers and the flash-point threshold can shift between editions, so confirm them against the adopted edition rather than trusting a remembered number.

What this means on the job: confirm the fluid's flash point meets the standard the AHJ enforces, confirm the fluid volume and the room's fire detection, protection, and any suppression are treated for that volume and class, and get the AHJ's position in writing before the tank is filled. The fire-code treatment of an immersion tank is jurisdiction-dependent and the adopted edition controls, so the AHJ conversation is the acceptance step, not the code book on your shelf.

Serviceability: pulling a hot server out of the fluid

Immersion changes maintenance more than any other liquid approach, because there is no dry server to work on. To service a node you lift it out of the tank, and it comes out hot, dripping fluid, and needing somewhere to drain. A single-phase node pulls up into the air and drips back into the tank if you give it time on a rail or a drip tray. The fluid is oily and it gets on tools, gloves, and the floor if you rush it. A two-phase tank has to be opened, which means managing the vapor and the seal, and that is slower and more involved.

This is the part operators underrate at acceptance. The tank has to support real service: a lift or a hoist for heavy nodes, drip management so a pulled server is not making a slick on the floor, and the space and procedure to handle a wet, oily piece of hardware. Confirm the service access, the lifting provision, and the drip and handling arrangement are there and workable, because a tank you cannot service safely is one that gets serviced badly.

The trade-off is real and worth naming. Immersion buys density and quiet and drops the fan energy to nearly nothing, and it costs you the easy swap. A technician used to sliding a rail-mounted server out in a minute is now pulling a dripping node out of oil. Plan the maintenance reality into the acceptance, not into the first outage.

Monitoring and controls

A tank that runs unwatched is a tank that fails quietly, so the instrumentation and its path to the building system get verified at acceptance. The core points are fluid temperature, fluid level, and leak or level alarms. Fluid temperature confirms the tank is rejecting heat and the supply to the hardware is in band. Level matters because a low level can uncover components or pull air into a pump, and a high level on an expanding or two-phase fluid can be its own problem. Leak and level sensing catches fluid going where it should not.

The acceptance work is to prove the alarms actually fire and actually arrive. It is not enough that a temperature sensor and a level switch are installed. Verify the readings are sane against a reference, then exercise the alarms. Drive a level or leak condition and confirm the alarm reaches the building management system and reads correctly, and that any automatic response the design calls for, a pump action or an alarm escalation, actually happens. An alarm that lives only on a local panel is an alarm nobody sees at 3 a.m.

Tie the monitoring to the integrated test below, because the controls have to hold the tank through an upset, not just report a calm day. A reading you confirmed once at commissioning, point-to-point to the BMS, is the difference between operations trusting the tank and operations guessing.

Environmental handling, disposal, and the maintenance the owner inherits

The fluid does not disappear at turnover. It becomes a standing inventory and an eventual waste stream the owner has to manage. Single-phase hydrocarbon and synthetic fluids are generally handled as an oil: stored, spill-contained, and disposed of through the channels the local rules and the fluid's safety data sheet specify, which is not the storm drain. Two-phase fluorinated fluids carry the heavier environmental and regulatory weight described in the PFAS section, and their handling and disposal are governed accordingly. The fluid's safety data sheet and the local environmental rules govern. Confirm them rather than assuming.

The maintenance the owner inherits is real and should be named at acceptance, not discovered later. The fluid ages: dielectric strength can drift, water and acidity can climb, and contamination from a poorly prepped node loads it up. Filters load and need changing. The heat-exchanger approach drifts as it fouls. A coolant-sampling schedule, a filter-change interval, and the acceptance baselines are part of the turnover package, not something operations is left to invent.

Hand a tank over without that baseline and that schedule and you have handed over a system that degrades silently until a chip throttles and nobody knows why. The maintenance plan is a commissioning deliverable. The fluid sample you pulled at fill is the first entry in it.

What the integrated test proves under load

The integrated test puts the tank under real or simulated heat load and runs it through the failure cases, because a tank that holds temperature on a calm commissioning afternoon has not been proven against the day something breaks. Load the tank with IT load or with heaters that mimic the chip heat, bring it to design fluid temperature and flow, and then start breaking things on purpose. The point is to find the gap between the design and the as-built tank while there is still time and no production workload is riding on it.

The scenarios are the ones that happen. Lose facility water and watch how long the tank rides through before the fluid temperature climbs out of band, because the fluid has thermal mass but a tank full of hot silicon heats up when the heat stops leaving. Fail a pump, where the design has redundancy, and confirm the backup holds flow. Trip a leak or level alarm under load and confirm the alarm and any programmed response fire. Lose utility power and confirm the cooling rides through the same gap the power plant has to ride through.

That last one is why the tank's integrated test does not stand alone. Cooling cannot ride through a power event the electrical side has not proven, so on a real facility the mechanical and electrical integrated tests run together, the power QA pillar covering the utility-loss transfer and the generators and UPS that have to carry the cooling through the gap. The immersion tank is the last link in that chain and the one sitting closest to the hardware.

What are the acceptance criteria for an immersion cooling tank?

The acceptance criteria fall into four areas, and every one is hedged to the governing document because the hard numbers belong to the fluid maker, the equipment vendor, and the project spec, not to a generic table. The areas are fluid, leak integrity, thermal performance, and structural and life safety. A tank passes when each is proven against its governing criterion and the result is documented.

Treat the table below as the shape of the acceptance, not the values. The dielectric strength, water content, particulate target, test pressure, design flow, supply temperature, flash point, and floor capacity are all real acceptance points, but each one's number comes from the fluid data sheet, the equipment submittal, the piping code by topic, the structural engineer, and the AHJ. Where a result lands close to a limit, treat it as a second look, not a pass, the same discipline as any commissioning hold point.

Field example: a single-phase tank that ran warm at the top

A single-phase immersion tank came up at commissioning with the fluid temperature on spec at the pump but the top row of nodes reading warmer than the bottom under load. The first instinct on the floor was that the heat exchanger was undersized, because the tank as a whole was carrying its heat and the facility water was on spec. The fluid temperatures by depth said otherwise.

Measured fluid temperature near the top of the tank ran several degrees above the bottom, and the flow pattern was the reason. The circulation was short-cycling along the bottom of the tank and not turning over the fluid around the upper nodes, so the hottest hardware sat in the least-moved fluid. The heat-exchanger approach was on target, so the exchanger was not the problem. The fluid distribution inside the tank was. It was the same failure as a poorly balanced loop, one region rich and one region starved, except the imbalance was inside the bath instead of across a manifold.

Correcting the internal flow path and the fluid level brought the top nodes into band, and the tank carried full load within its supply-temperature spec with no change to the heat exchanger. The lesson it left is the one this guide keeps returning to: the tank average can look fine while a region of hardware runs hot, so you verify the condition where the hardware actually sits, not the number at the pump. The as-found and as-corrected fluid temperatures went into the record as the baseline operations would trend against.

What to document

A tank that was proven but never documented hands operations a system with no baseline, and the first warm node becomes a guessing game. The record is what tells the next engineer whether a reading is a new problem or how the tank has always run. Capture the tank and fluid identity, the fluid chemistry baseline, the cleanliness and fill result, the leak-integrity result, the flow and heat-exchanger approach, the supply temperature under load, the structural sign-off on the floor load, the fire-code position from the AHJ, the leak and level alarm test, and the signoff.

Two records carry the most weight later. The fluid baseline sample proves the starting chemistry you can never reconstruct once the fluid has aged, and the structural and fire-code sign-offs prove the two things that are hardest and most consequential to fix after the tank is filled. A turnover package missing those leaves the owner trusting that the steps that matter most got done.

Common mistakes

• Filling with a fluid that does not match the specified product, grade, or concentration, or cross-contaminating it during a sloppy fill.
• Leaving incompatible materials on a server, standard thermal grease, vented hard drives, or the wrong labels, that degrade the fluid for the whole tank.
• Setting a tank without a structural check of the filled-tank weight against the floor's rated capacity.
• Skipping the fluid baseline sample, so there is no starting chemistry to trend aging against.
• Not getting the AHJ's fire-code position on the fluid flash point and volume before the tank is filled.
• Proposing a two-phase tank without confirming the fluid's long-term availability and PFAS regulatory status.
• Accepting the tank on the average fluid temperature while a region of hardware runs hot.
• Installing leak and level alarms but never driving a condition to prove they reach the BMS and fire the design response.
• Handing the tank over with no fluid-sampling schedule, filter interval, or maintenance baseline for operations.

Field checklist

Standards and references

ASHRAE Technical Committee 9.9 is the central thermal reference. Its thermal guidelines and liquid-cooling guidance set the facility-water supply temperatures and the water-quality and material-compatibility framework the immersion heat-rejection side is built around, the same guidance the liquid cooling loop work draws on. The Open Compute Project publishes the immersion-specific requirements many deployments design to, including the OCP immersion requirements, a base specification for immersion fluids, and design guidelines for immersion-cooled IT equipment, covering the tank, the fluid, the facility-water interface, and component preparation. Name the OCP document by topic and confirm the current revision, since the documents revise on their own cycle.

The fluid itself is governed first by the fluid manufacturer's specification and safety data sheet, which set the dielectric strength, flash point, compatibility list, and handling, and which override any general guidance where they are stricter. Fire and life safety run through NFPA 75 for the protection of information technology equipment, which in recent editions sets a flash-point floor for immersion-cooling fluids, and NFPA 30, the flammable and combustible liquids code, which classifies the high-flash fluids and is adopted by reference in the International Fire Code. The piped portions of the loop fall under the applicable pressure-piping code, commonly an ASME B31 section by topic, with the project spec setting the test parameters.

Edition numbers, the exact flash-point threshold, the water-class values, and the OCP revision levels move between cycles, and the fire code is adopted and amended by jurisdiction, so confirm all of them against the published documents, the fluid and equipment manufacturers, and the AHJ before citing them on a submittal. ASHRAE TC 9.9 and OCP give the framework. The fluid manufacturer's specification, the structural engineer, and the AHJ control the limits you accept the tank to.

Units, terms, and acronyms

Immersion cooling borrows vocabulary from HVAC, the fluids industry, the IT side, and the fire code, so the same idea can read differently across a tank submittal, a fluid data sheet, and a code review. The terms below travel across the immersion scope.

Single-phase immersion

Immersion where the dielectric fluid stays liquid and is pumped or convected past the hardware to a heat exchanger

Two-phase immersion

Immersion where the fluid boils at the chip and condenses on a coil, carrying heat as latent heat of vaporization in a sealed tank

Dielectric fluid

The non-conductive liquid the hardware is submerged in; it carries heat off the components by direct contact

Dielectric strength

A fluid's ability to resist conducting electricity; it must stay high and drops with contamination or water

Flash point

The temperature at which a fluid gives off enough vapor to ignite; it sets the fire-code classification

GPM / LPM

Gallons or liters per minute, the fluid or facility-water flow rate confirmed against the design

kW per tank

The heat load a tank carries; immersion tanks reach densities far above an air-cooled rack

Approach temperature

The gap between facility water in and fluid out at the heat exchanger; a widening approach signals fouling

DECS

Datacom equipment cooling system, the OCP term for the fluid, heat exchangers, circulation, and controls within the tank or rack

PFAS

Per- and polyfluoroalkyl substances, the fluorinated chemistry behind two-phase fluids, under heavy regulatory pressure

Related tools

Calculators and readiness checks for this work

Compare your options

FAQ

People also ask

• What is bypass airflow in a data center?
• What is direct-to-chip liquid cooling?
• What is two-phase cooling?
• What is direct-to-chip liquid cooling?
• What is in-row cooling?