Torrential rain is not the ideal weather condition for the launch of 5,072 solar panels on the roof of a factory. Nevertheless, on a meteorologically miserable day in April, DCD braved the slippery conditions of a mildly terrifying external spiral staircase to see the culmination of the latest development at Lenovo’s factory in Budapest, Hungary.

In 2020, the site the factory now sits on was a cornfield. The external shell of the building was built over 10 months at the height of the Covid-19 pandemic, with a further seven months needed to get all the relevant licenses in place and construct the manufacturing lines.

The factory now totals 49,000 sqm (527,400 sq ft) and ships between 600,000-800,000 Lenovo products, including commercial and enterprise servers and storage, high-performance computing (HPC) solutions, and water-cooled servers. Lenovo sends these products to 2,500 customers across 70 countries each year.

From day one, the factory was conceived with sustainability in mind. In the winter months, cold air is pumped in from the outside to keep the factory cool, while heat generated from its servers is used to warm water that circulates through the building's radiators.

The newly unveiled solar panels provide 3MW in solar energy capacity which, though not enough to power the whole factory, does supply enough energy to run its HPC and AI Innovation Centre which sits at the heart of the operation.

The Innovation Centre provides Lenovo ThinkSystem and ThinkAgile hardware and solutions to support demos, proof of concepts, and benchmarking for the company’s EMEA customers, with each Lenovo ThinkSystem SR680a V3 and ThinkSystem SR780a V3 able to support eight Nvidia H100 and H200 Tensor Core GPUs.

The center also offers remote VPN access to customers and partners, in addition to providing Lenovo’s first liquid cooling testing and demonstration facility within Europe.

Similarly, the company’s global HPC and AI test facility is supported by HPC Cluster Lennox architecture, again offering benchmarking, proof of concept, and solution demo capabilities to the company’s global customer base.

A second AI Innovation Centre to serve customers in North America is housed at Lenovo’s so-called “megasite” facility in Monterrey, Nuevo Leon, Mexico.

Speaking to DCD in London ahead of the factory tour, Scott Tease, who leads Lenovo’s supercomputing and high-performance computing and AI systems team, says one of the biggest challenges facing data center providers in the age of AI is power constraints, as many data centers are currently ill-equipped to deal with the kind of density these workloads require.

“Most of our data centers are designed for eight to 15kW per rack, but one of these generative AI servers is 10kW, just a single one. So you can imagine having a 40-unit tall rack with a single server inside of it. It’s got to feel like you’ve wasted some of the most expensive real estate in the world inside of a data center,” he says.

More powerful hardware means companies can now run workloads that used to require 42 servers on 15 to 18 instead. But Tease is unsure if the increased efficiency is worth the power these servers are now consuming, an issue that, he explains, is best demonstrated by the proliferation of GPUs.

“With traditional CPU servers, the most you could hit would be maybe a 40kW rack,” Tease says. “That would really be pushing the envelope, really doing something pretty amazing to get that dense.

“With GPU systems, we're already shipping GPU server racks that are at 80-85kW per rack.”

Tease notes that, as a result of supply chain constraints, GPUs are now being sourced from other vendors, as well as Nvidia. It now has competition in the shape of the AMD MI300X and Intel’s Gaudi3, both of which are due to start shipping later this year.

“[AMD and Intel are offering] a very attractive accelerator for these big systems,” he says.

That is one of the benefits provided by the AI Innovation Centre. In partnership with AMD, Intel, and Nvidia, customers are able to trial new data center solutions and test the hardware before deciding to invest a not insignificant amount of money or join a lengthy waitlist.

Built to perfection

Back in Budapest, speaking to DCD on the day of the solar panel unveiling, Brian Jaeger, supply chain services director at Lenovo, explained that the AI Innovation Centre has the capacity to support around 20 racks worth of proof of concepts (POC), meaning the number of customers that could be taking advantage of the center’s provisions at any one time could vary depending on how many servers their POC requires.

“Typically a proof of concept will last between 30 and 90 days,” he says. “That’s the window we’re trying to work with.”

Lenovo is looking to increase this capacity over the coming year but did not divulge by how much.

Jaeger added that one of the benefits of having the Innovation Centre based inside of the factory is that it allows the company to offer almost its entire product portfolio for inclusion in POCs, because all of Lenovo’s supply chain materials are housed at the same facility.

“We can basically build the exact configuration a customer is looking for,” he says, something that wouldn’t be possible if it was instead housed at a Lenovo development center as access to inventory would be much more limited.

“We're able to actually build exactly what the customers think they want and basically offer a 'try before they buy' type of environment,” Jaeger adds.

This also allows them to build configurations using all the necessary component parts, instead of testing a setup that uses a customer’s AI accelerator of choice but a basic version of a storage solution.

The Innovation Centre was built specifically to support AI workloads, and it’s therefore unsurprising that Jaeger says a lot of focus on AI right now and it’s the space in which Lenovo’s customers are making the biggest investments.

“Our focus here over the last 12 to 24 months has been building out and continuing to expand our AI portfolio,” he says, adding that the data center market has seen a whole series of trends over the last 20 years before, most recently, focusing its attention on AI.

Despite some recent skepticism around the long-term promise of generative AI specifically – a Goldman Sachs report from earlier this year said big businesses are set to spend $1 trillion on AI hardware in the coming years with no clear plan to get a return on their investment – Jaeger believes the AI hype will endure for some time yet.

“We are continuing to expand our portfolio to support that and make major investments in these very high-end AI systems that are that are going to be to be coming out, so absolutely I think [AI] is here to stay,” he says.

“Now, will it shift into a different direction in 10 years? Probably. But for several years, I think we'll continue to see the focus around building out AI capabilities.”