Bitget App
Trade smarter
Buy cryptoMarketsTradeFuturesStocksEarnInstitutionAI & More
This company’s metal towers might offer a solution to the significant heat challenges faced by AI.

This company’s metal towers might offer a solution to the significant heat challenges faced by AI.

Bitget-RWA2025/11/05 18:45
By:Bitget-RWA

When Nvidia introduced its Rubin GPU lineup in March, it made headlines by revealing that racks equipped with the Ultra chip—anticipated for a 2027 launch—could consume as much as 600 kilowatts of power. That’s almost double the output of today’s most powerful EV chargers.

As data center racks demand ever more energy, a major challenge will be managing their cooling. One startup believes the solution lies in stacks of metal.

Alloy Enterprises has created a method that transforms copper sheets into robust cooling plates for GPUs and other peripheral chips—components like memory and networking devices that make up about 20% of a server’s cooling requirements. 

“That 20% wasn’t a big concern when racks were only 120 kilowatts,” said Ali Forsyth, Alloy Enterprises’ co-founder and CEO, in an interview with TechCrunch. But now, as racks reach 480 kilowatts and are on track for 600, engineers must find ways to liquid cool everything from RAM to networking chips—areas where current solutions don’t exist.

Alloy’s solution leverages additive manufacturing—building up objects layer by layer—to fabricate cold plates that fit into compact spaces and endure the high pressures required for liquid cooling.

This company’s metal towers might offer a solution to the significant heat challenges faced by AI. image 0 Image Credits:Alloy Enterprises

However, the company doesn’t rely on 3D printing. Instead, it bonds metal sheets together using heat and pressure. This approach costs more than traditional machining but is less expensive than 3D printing.

The outcome is a cold plate that is essentially a single, solid piece of metal. Unlike machined products, there are no seams, and unlike 3D-printed versions, the metal is not porous. “We achieve the properties of raw material,” Forsyth explained. “The copper is as strong as if it had been machined.”

Most cold plates are produced by machining, which involves carving out features with large tools, requiring each half to be made separately. The halves are then joined through sintering—a process that fuses metal powders with heat—creating a seam that could leak under high pressure. Alloy’s method, a form of diffusion bonding called “stack forging,” produces seamless cold plates. 

Stack forging also enables the creation of finer features, down to 50 microns—about half the width of a human hair—allowing more coolant to circulate. According to Forsyth, Alloy’s cold plates offer 35% better thermal efficiency than competing products.

Due to the intricacies of stack forging, Alloy handles most of the internal design work. Clients provide essential specifications and measurements, and the startup’s software converts these into a design compatible with their manufacturing process.

At Alloy’s facility, copper rolls are first prepared and cut to the required dimensions. A laser then carves out the necessary features. Areas that shouldn’t bond are coated with an inhibitor. Once ready, each layer of the cold plate is aligned and stacked before being placed in a diffusion bonding machine, where heat and pressure fuse the stack into a single metal unit.

Forsyth mentioned that the company is collaborating with “all the major players” in the data center industry, though she declined to name specific partners. 

The technology was originally designed for a common aluminum alloy, but as interest from data centers grew, Alloy adapted the process for copper, which is highly conductive and resistant to corrosion. After Alloy unveiled the product in June, “interest skyrocketed,” Forsyth said.

0
0

Disclaimer: The content of this article solely reflects the author's opinion and does not represent the platform in any capacity. This article is not intended to serve as a reference for making investment decisions.

Understand the market, then trade.
Bitget offers one-stop trading for cryptocurrencies, stocks, and gold.
Trade now!

You may also like

VIPAI Asset Sell-Off and Selective Rotation in Crypto

1. Technology stocks fell sharply as investor interest shifted toward energy and selected defensive assets. From October 10 to 16, the Nasdaq 100 declined 5.49% and the S&P 500 fell 1.95%, while the Dow gained 0.67% and WTI crude rose 10.41%. BTC, ETH, and SOL declined 1.27%, 2.65%, and 4.44%, respectively. This week's technology earnings will be an important test of whether revenue growth, margins, and cash flow can justify current valuations. Better-than-expected results could help stabilize markets in the near term. 2. Crypto continues to see pockets of strong activity, but a broad-based altcoin season has yet to materialize. CoinMarketCap's (CMC) Altcoin Season Index stood at 54 as of the October 18 UTC daily observation, below the altcoin-season threshold of 75, while BTC dominance rose to 60.08%. Our proxy leverage ratio fluctuated between 4.00 and 4.07 during the week, showing no sustained increase and suggesting limited appetite for additional leverage. Meanwhile, as major exchanges expand access to traditional financial assets, the recent sell-off in AI-related equities could attract short-term speculative capital away from crypto. Such capital rotation could make it more difficult for a new altcoin season to take hold. 3. Papertrade's loss-to-mine model has generated significant interest across the crypto community. The protocol combines high-leverage derivatives trading with zero slippage and a mechanism that rewards trading losses with PAPER tokens. Across the eight complete UTC days from October 10 to 17, cumulative reported revenue reached approximately $56.08 million, including $51.90 million distributed to token holders. With growing participation from key opinion leaders (KOLs), Papertrade has reportedly ranked among the top three platforms by profitability in the broader market recently. Assets to watch: BTC, ETH, SOL, ARB, OP, SKHYNIX, SKHY, META, TSLA, LRCX, and SPCX. Key research priorities this week: Cash flow performance in technology earnings; the relative discount and holding costs of the two SK Hynix contracts; and U.S. initial jobless claims and PMI data. Weekly market data cover the period through October 16, while research information was verified through October 18. Observation periods are specified separately for each chart.

Bitget•2026/10/19 06:51
AI Asset Sell-Off and Selective Rotation in Crypto