On-Prem

AI frenzy continues as Macquarie commits up to $5B for Applied Digital datacenters

Bubble? What bubble?


Fears of an AI bubble have yet to scare off venture capitalists and private equity firms from pumping billions of dollars into the GPU-packed datacenters at the heart of the machine-learning craze.

Case in point, on Tuesday Macquarie, one of Australia's top financial services firms, revealed plans to inject up to US$5 billion into American datacenter operator Applied Digital to finance the construction of new facilities.

The Dallas, Texas-based operator is a relative newcomer to high-performance computing, though that hasn't stopped it from trying to replicate the successes of its contemporaries.

The biz has followed a similar trajectory to CoreWeave breaking into the HPC services space circa 2022 after a bout of cryptocurrency mining. Like CoreWeave and Lambda, another major rent-a-GPU outfit, Applied offers a variety of datacenter and cloud services powered by Nvidia's chips.

And with Macquarie's investment, which includes an initial $900 million round of funding, Applied plans to move forward with its Ellendale High Performance Computing campus in North Dakota while also recovering roughly $300 million in equity previously invested in the 400-megawatt facility.

The campus is the one of several Applied Digital datacenters that Macquarie intends to plow money into. Under the agreement, the financial services firm has been granted first right of refusal for all future AI datacenters Applied may pursue over the next 30 months up to $4.1 billion.

"At today's build costs, we will have a significant portion of the equity needed to construct over two gigawatts of HPC datacenter capacity, including our Ellendale Campus," Applied Digital CEO Wes Cummins said in a statement.

Applied hasn't said when, where, and how many datacenters will ultimately be built, though the press release does seem to suggest that these facilities will be liquid cooled. It is, however, a fairly safe bet that, just like its North Dakota site, future datacenters will utilize Nvidia accelerators — perhaps even its 120 kilowatt liquid-cooled GB200 NVL72 rack systems.

You may recall, Nvidia along with a handful of other investors poured $160 million into Applied Digital back in September to fuel its datacenter build out. It probably doesn't hurt that much of Nvidia's investment is likely to return in the form of GPU orders and AI Enterprise subscription fees.

In exchange for its investment, Macquarie has effectively bought itself a 15 percent stake in Applied HPC business with the datacenter operator retaining the remaining 85 percent share.

"As an owner and manager of datacenter platforms, we see this as a highly attractive opportunity to help build an industry leading HPC datacenter company well positioned in these high growth segments of the market," Anton Moldan, senior managing director at Macquarie, said in a canned statement.

According to our sibling site The Next Platform investing $1.5 billion in infrastructure — enough to build and network a datacenter with roughly 16,000 H100-class GPUs to rent out — will net you about $5.27 billion in revenues within four years, assuming a decent mix of spot instances and long-term commitments from customers.

Given the estimated return on investment, it's not hard to see why so many capital management firms are clamoring for a piece of the action. It probably doesn't hurt either that GPUs are in such high demand they're now considered fungible. This has enabled datacenters to use them as collateral for massive loans.

In fact, last April Macquarie helped to drive half a billion dollars of debt financing backed by Nvidia GPUs into rival DC operator Lambda.

Among the biggest beneficiaries of this trend has been CoreWeave, which according to Crunchbase raked in $9.9 billion in funding and debt financing in 2024 alone.

Some investment groups now appear to be cutting out the middleman and pursuing datacenter developments. Last fall, US investment giant Blackstone said it would plow £10 billion (US$12 billion) into a massive AI datacenter in northeast England. ®

Send us news
11 Comments

Intel has officially missed the boat for AI in the datacenter

But it still has a chance at the network edge and the PC

Where does Microsoft's NPU obsession leave Nvidia's AI PC ambitions?

While Microsoft pushes AI PC experiences, Nvidia is busy wooing developers

Datacenter energy use to more than double by 2030 thanks to AI's insatiable thirst

Shocking research warns electricity shortages could create construction bottleneck

France, UAE to drop €50B on AI mega-datacenter. Still nowhere near America’s $500B bet

Oh look, a mini Stargate, how quaint

Nvidia deprecates CUDA support for aging architectures

Maxwell, Pascal and Volta, oh my! But fear not, driver support is still safe

Intel knocked off global chip revenue top spot after rotten 2024

Missed the AI processor boat, split with CEO savior, lost #1 seat to Samsung

IBM seeks $3.5B in cost savings for 2025, discretionary spend to be clipped

Workforce rebalancing? Yes, but on the plus side, the next 12 months are all about AI, AI, and more AI

DeepSeek means companies need to consider AI investment more carefully

But Chinese startup shakeup doesn't herald 'drastic drop' in need for infrastructure buildout, say analysts

Who's afraid of DeepSeek's impact on AI hardware sales? Not AMD CEO Lisa Su

Predicts more efficient ML architectures will drive adoption, see Instinct sales accelerate, shares dive

South Carolina's abandoned nuclear reactors positioned to fuel the AI datacenter boom

VC Summer units 2 and 3, abandoned in 2017, are looking for a buyer; owners say tech industry needs are a perfect fit

Ubuntu upgrade had our old Nvidia GPU begging for a downgrade

If you need legacy drivers, you might want to keep your older OS version – possibly indefinitely

Not even Nvidia's Jensen Huang can talk President Tariffs out of chip import taxes

GPU giant could just wait eight minutes for Donald to change his mind