Power, performance and partnerships: Sharon AI’s US data centre play

High performance computing service provider Sharon AI has teamed up with Lenovo and NVIDIA to develop its AI Supercluster offering, providing a full-stack sovereign AI cloud.

However, despite establishing technology within data centres, its current strategy is to colocate in Australia and build up its own data centre in the US.

Speaking to ARN, Sharon AI COO and cofounder Andrew Leece said the provider goes beyond GPUs by “closing the gap for researchers, enterprise, and government from ideation to implementation.

“We built an enormous library of tools, resources and automations for a mid-level engineer from any type of organisation in Australia,” he said.

In Australia, Sharon AI colocates within a NextDC facility to provide its services. Additionally, it also holds partnerships with Equinix and CDC and plans to offer its solution domestically through the three data centre operators.

Its offering is appealing enough on a global scale to draw attention in the US, but instead of colocating abroad, Sharon AI is on its way to build its own data centre in what Leece calls a “particularly competitive” market.

Construction officially kicked off in November last year in Texas with Sharon AI forming a joint venture with industrial gas production company New Era Helium to establish the facility.

As of June, the pair announced a memorandum of understanding with PowerForward Energy Solutions for the manufacturing, installation and operation of 250 megawatt generators for the site and in July signed non-binding letters of intent with multiple parties for land acquisition and potential power purchase agreements.

Leece referred to the US as the “hub of the ecosystem that’s building around AI”. In Sharon AI’s efforts to find capacity, he said that the US proved to be a “very tight” data centre market despite also being the largest and most mature.

“As a result of that, we strategically identified the opportunity to go and procure land but with a slightly different angle,” he said.

“We are not only building the data centre, but we’re also building the energy generation asset that’ll sit on site to eliminate some of the variability in power prices and give us the control to ensure that it’s a net zero asset and a net zero supply of power.”

Meanwhile, the Australian market is, in Leece’s own words, “entirely different”; while it has “phenomenal” amounts of high-quality data centre capacity and power supply and has contextually similar characteristics geopolitically speaking, it also has a much smaller addressable market up until recently.

He added that Australia has had great industrial capabilities over the years, but it has dissipated in recent times. As such, other regions and countries have become more efficient by comparison. That being said, the COO is still looking to establish its presence and export its capabilities to Asia and the rest of the world.

Working with hardware partners

A core part of Sharon AI’s journey is its partnership with Lenovo, which saw the establishment of a formal partnership between the two in the middle of 2021, roughly six months after it opened its doors.

“The relationship with Lenovo was a somewhat accidental relationship initially and has grown into a very strategic alliance from our perspective,” Leece said.

“Lenovo have treated us as a genuine partner, not just a customer taking orders. They engage with us globally with customers, they lead us into customers – we even have an opportunity right now, we’re working on in Japan, hand in hand.

“It goes way beyond just the provision of hardware.”

Lenovo vice president and worldwide general manager of AI and high-performance computing Scott Tease said Sharon AI’s services are providing “last mile capability” to put AI technology in the hands of everyone.

“These systems are quite different from your traditional CPU-based system, and Sharon makes it easy for customers to take advantage of it however they want,” he said.

“It’s not like it’s just access to metal itself – it’s all of the capabilities that they bring to allow the customer to bring their data in and make use of it by applying different large language models they get from outside.”

It’s not just Lenovo that’s helping out with Sharon AI’s service development, as there’s also NVIDIA technology at play, but the computing and processing vendor came into the equation further down the track.

“We initially worked with Lenovo to design the systems and then procure the systems. We did so initially without the support of Nvidia,” Leece said.

“It was only actually through our relationship with Lenovo that we were introduced directly to Nvidia from a local and now global perspective.

“It was quite interesting when we were first introduced directly to Nvidia. They said, ‘How have you guys gone and done all of this already without our support there? So many people come to us saying, ‘We will do this if you give us XYZ’; you guys have just gone and done it.’ It started our relationship with NVIDIA on a very different trajectory because we were already building … systems in country.”

Since then, Sharon AI’s partnership with the vendor has grown and is currently an NVIDIA Cloud Partner.

Tease echoed Leece’s sentiments, adding that NVIDIA’s partnership is what enabled Sharon AI to scale up considerably.

Going back to five years ago, Tease said GPU discussions revolved around PCI cards, then in the last few years shifted to server level AI, which necessitated various components. When this year rolled around, a migration took place towards rack-level AI, he continued.

“As you get bigger building blocks, it gets more and more complicated to make sure it all fits together properly. What Sharon AI is buying is super cluster-sized amounts of IT,” he said.

“So, having the person that designed the GPU and the networking be a part of everything that we do is super critical to make sure that we’re creating a perfectly balanced system that’s up to whatever we’re doing today, but also what what’s going to be needed tomorrow based on what NVIDIA is seeing from the really big players like Meta, Google or Microsoft.

“Having their partnership is what’s allowed us to move so quickly from such a small scale to the big scale that Sharon AI is doing today.”