Surging demand for artificial intelligence (AI) and computing power is driving data centers to ramp up their cooling capabilities.
McKinsey & Company estimates that global demand for data centre capacity could grow by 19–20% annually through 2030, pushing yearly consumption to between 171 and 219 gigawatts by the end of the decade.
Burlington-based Alloy Enterprises is meeting this rising demand with its American-made direct liquid cooling (DLC) cold plates. These are designed to improve GPU server rack efficiency, reducing power consumption by 21% and cutting pumping pressure by a factor of four.
Access more expert insights at Additive Manufacturing Advantage: Energy on September 17th. Reserve your free ticket now.
Alloy uses its proprietary metal additive manufacturing technology, Stack Forging, to produce its devices. This process fuses layers of laser-cut aluminum and copper sheets into fully dense, high-performance parts. According to CEO Ali Forsyth, the method can create internal microchannels as small as 50 µm.
In a recent interview with 3D Printing Industry, the Harvard University graduate estimated the 2024 serviceable addressable market for GPU and manifold cooling at $1.1 billion, with a projected CAGR of 36%. Forsyth called Stack Forging the ideal additive process to meet soaring demand for data center cooling and revealed that Alloy is scaling its DLC solution for mass production.
The company can produce 15,000 components per month with a 92% part yield. It uses aluminum that costs less than $10 per kilogram, 20 times cheaper than the powder-based 6061 aluminum used in laser powder bed fusion (LPBF). Alloy has also integrated copper into its DLC solution, increasing thermal performance by approximately 30% compared to aluminum alone.
Ultimately, Forsyth emphasized the urgent need to cut the carbon emissions of data centers. She cited cleaner power sources, like nuclear fusion, as key to averting the looming energy crisis.
What is Stack Forging?
Alloy Enterprises was founded in early 2020. The company sought to develop a process capable of producing fully dense aluminum components at higher throughput and lower cost than existing metal 3D printing methods.
“Our goal was just to move a lot of metal through the machine and make high-performance components,” Forsyth explained. “We believed that if we could create a manufacturing process that could do that, we could open up some newer market opportunities at higher volumes for series production.”
Forsyth’s company now runs two shifts on the production floor, with some equipment operating 24/7. Over the past two years, Alloy has experienced growing demand for advanced thermal management, driven by customers seeking better cooling for high-power devices. While initial interest came from photonics and defense primes, it quickly expanded to semiconductor equipment, data centers, and GPU cooling.
Today, the firm focuses on thermal management hardware, especially for data center server racks. It ships certified components to 15 customers worldwide, including some of the “biggest folks in the industry,” Forsyth explained. Alloy’s main clients are server rack OEMs, hyperscalers, and chip manufacturers.
At the core of Alloy’s technology is its proprietary Stack Forging process. This begins with thin sheets of 6061 aluminum alloy or C110 copper, typically 350 to 400 µm thick. Each layer is laser-cut to define the part’s geometry, including intricate internal channels and multi-part nesting.
An inhibition agent, functioning like a mold release, is then selectively printed onto the surface before the layers are stacked and placed into a “Bond Machine.” This provides heat and pressure in a controlled environment to create solid, fully dense components. Finally, the parts are heat-treated to enhance strength and hardness.
Notably, these monolithic components are leak-free, a crucial advantage when using liquid to cool GPU racks that can be worth millions of dollars. “Reducing the likelihood of leaking on this really expensive and valuable hardware is absolutely central,” Forsyth said.
Another key differentiator is Alloy’s precision and ability to create enclosed, intricate geometries that legacy manufacturing methods can’t achieve. Finer microchannels deliver higher thermal performance, a critical factor in optimizing high-powered electronics.
Forsyth emphasized that stack forging outperforms other metal additive manufacturing processes. She stated that the smallest comparable channels made using LPBF can reach around 300 µm, while stack forging can achieve 50 µm, “and we’re pushing down to 10 µm.” Additionally, because Alloy’s method avoids melting, sintering, or fusing, the resulting components are free of porosity and deliver superior mechanical strength, Forsyth added.
Stack Forging also offers significant advantages in printing speed and throughput. According to Forsyth, the process delivers nine times the throughput per laser compared to LPBF. On the materials side, Alloy’s aluminum sheets are 20 times cheaper than powder-based 6061, while being significantly easier to handle.
Unlike metal powders, which require energy-intensive atomization and complex post-processing, Alloy’s sheet-based feedstock has a lower carbon footprint and eliminates the risk of trapped powder in narrow internal channels.
Meanwhile, waste material, mostly support sections, can be easily collected and recycled. “We’ve designed our process around reusability,” Forsyth added. “Both aluminum and copper scrap can be remelted and turned back into sheet with minimal waste.”
Enhancing thermal management for data centers
Keeping data centers cool is an urgent challenge. As digital infrastructure expands, servers are processing ever-larger volumes of data and emitting massive amounts of heat. AI-ready facilities are especially power-hungry, driven by the high average power densities of their server racks.
According to McKinsey, average power densities more than doubled from 8 kW in 2022 to 17 kW in 2024, and are expected to reach 30 kW by 2027. Training models like OpenAI’s ChatGPT can require over 80 kW per rack. During this year’s GTC AI conference, NVIDIA’s CEO, Jensen Huang, revealed his company’s new NVL576 rack, which is capable of using up to 600 kW of power.
“Chipmakers are producing larger, hotter processors, making efficient cooling a critical concern for data centres,” said Forsyth. She noted that cooling these systems can consume more than 20% of a facility’s total power, making thermal efficiency invaluable.
Alloy Enterprises aims to tackle these challenges with its direct liquid cooling (DLC) solution. The cold plates feature intricate internal microcapillaries that enhance cooling efficiency. According to the company, these complex architectures reduce thermal resistance by up to 50% and improve Power Usage Effectiveness (PUE) by 18%.
The 3D printed components also deliver an exceptionally low pressure drop, which Forsyth revealed is four times lower than the nearest competitor. This reduces pumping resistance and improves liquid flow rate, meaning less energy is required to circulate cooling fluid around data center racks.
Crucially, improved thermal performance allows data centers to cool more powerful chips using higher-temperature water, which is typically 44°C. “You want to be able to circulate warmer water throughout your facility,” Forsyth explained. This, in turn, allows dry chillers to discharge heat outside, sidestepping the need for energy-intensive refrigerated air conditioning systems. Forsyth described this as “absolutely critical for power savings.”
Forsyth argued that “100% liquid cooling is the way forward,” eliminating the need for bulky, energy-intensive heat sinks and fans required for air-based cooling. The company’s liquid-focused components also enable higher cooling density, allowing rack builders to pack more electronics.
In addition to GPUs, these peripheral components include memory, SSD hard drives, power supplies, QSFP transceivers, and Network Interface Cards (NICs). “We have a lot of customers looking for part consolidation, helping to cool more than one item at a time with a single cooling plate,” explained Forsyth.
In modeled scenarios for 2030, Alloy’s technology reduces total power use in a 75 MW data center by 21%, compared to the next best option. “It’s incredibly remarkable how much power can be saved with higher performance cooling components,” Forsyth added.
Alloy Enterprises began producing 3D printed GPU cooling devices using 6061 aluminum sheet. Earlier this year, it used the same material to create a liquid cold plate for NVIDIA’s H100 PCIe card.
This high-performance data center GPU is designed to accelerate AI and advanced computing. Alloy developed the cooling plate using topology optimization design software from nTop. This process allowed the team to optimize the device’s geometry using simulation data, including heat flux, fluid flow, and structural constraints.
In recent months, Alloy has moved beyond aluminum, unveiling its copper DLC solution in June. By using C110 copper, the company has reportedly improved cooling efficiency by about 30% compared to aluminum. The material also meets ASHRAE chemical compatibility standards, reducing the risk of corrosion, degradation, and leaks.
Mass production with metal 3D printing
As demand for AI compute surges, Alloy is ramping up production to keep pace. Its current facility can produce 15,000 components per month. Capacity can be expanded by adding additional modular Stack Forging cells to the production line.
“The scalability of Stack Forging was core to the development of the technology,” explained Forsyth, who is targeting “mass production” of her company’s DLC devices. “We have no problem servicing customers that need tens of thousands of components annually,” she revealed.
That scale matters, as the market for data center cooling continues to surge. “The demand for compute is skyrocketing. The demand for power to fuel that compute is skyrocketing. And the demand for cooling components is right alongside it,” Forsyth added.
The CEO, formerly an Engineering Manager at 3D printer OEM Desktop Metal, said she plans to scale output to meet rising customer demand. The focus is squarely on next-generation data centers, with no plans to pursue retrofits of existing facilities.
I asked Forsyth why Alloy opts to produce DLC components in-house rather than sell its Stack Forging machines to customers. She cited the significant challenges of training customers to reliably operate 3D printing systems and design parts for additive manufacturing.
“We had a choice between the first five machines at five customer sites, or the first five machines in our facility run by our employees,” said Forsyth. “In order to ensure reliability and quality, we felt the best path was to put our money where our mouth was.”
This in-house approach is paying off. Alloy has steadily improved equipment efficiency in recent years, reaching a 92% part yield in May 2025, an impressive benchmark in industrial 3D printing. “I’m incredibly proud of the engineering and production teams for having achieved that milestone,” Forsyth said.
The future of 3D printing for GPU cooling
Looking to the future, Forsyth acknowledged that power availability for data centers is not limitless, making improved thermal efficiencies essential. “Any kilowatt you can save can be used to generate more tokens or run more computations,” Forsyth said. “That translates into higher top-line revenue or reduced CO₂ emissions.”
Indeed, the carbon footprint of data centers is facing growing scrutiny amid tightening environmental regulations and emissions targets. According to the International Energy Agency (IEA), data centers consumed approximately 415 Terawatt-hours in 2024, roughly 1.5% of global electricity use. The IEA estimates that, to achieve its Net Zero targets, data‑center emissions must halve by 2030.
Many of Alloy’s customers, particularly hyperscalers, have committed to cutting their carbon footprints. For Forsyth, DLC’s efficiency gains are key to meeting those goals. She highlighted its ability to deliver more compute per kilowatt and run on uncooled 44°C water critical. She also noted that Stack Forging’s sheet-based feedstock generates significantly less CO₂ than the metal powder used in LPBF, while being far easier to recycle and reuse.
The tech CEO also urged a shift toward cleaner energy to power data centers. She highlighted Google’s recent agreement to buy 200 megawatts of electricity from ARC, a compact nuclear fusion plant being developed by MIT spin-off Commonwealth Fusion Systems (CFS). Located in Chesterfield County, Virginia, ARC aims to deliver clean energy by the early 2030s.
“I think our whole society will benefit from that,” exclaimed Forsyth. “If it’s data centers that push those innovations, then I’m okay with that, because the whole world’s gonna need it.”
Registrations are now open for Additive Manufacturing Advantage: Energy on September 17th. Reserve your free ticket now.
Want to help select the winners of the 2025 3D Printing Industry Awards? Register to join the Expert Committee today.
Subscribe to the 3D Printing Industry newsletter to keep up with the latest 3D printing news.
You can also follow us on LinkedIn, and subscribe to the 3D Printing Industry Youtube channel to access more exclusive content.
Featured image shows Alloy Enterprises’ copper cold plates. Photo via Alloy Enterprises.
