Chris Kapusta, an executive at Softcat-owned GDT, explains how Nvidia’s AI financing boom can benefit the channel, but says solution providers, including his company, are grappling with the ongoing global components crunch and growing U.S. data center backlash.
Chris Kapusta has seen AI infrastructure deals with neocloud companies “stall” or “fizzle out” before because such customers sometimes lack the financing muscle to procure the expensive data center equipment they need.
But the executive at U.S. solution provider giant GDT, now owned by U.K.-based Softcat, said that’s starting to change as Nvidia plays a growing role in providing financing or credit support to neoclouds and other customers in need of funding assistance.
[Related: How Mark III Won ‘One Of The Largest’ Nvidia AI Factory Enterprise Deals In The Channel]
“What I think Nvidia has quickly realized is the market has to catch up to meet that demand, and with funding for some of these newer neoclouds sometimes being challenging, this is a way to help secure that at a more successful rate and help accelerate the capacity of GPU- and AI-driven infrastructure that’s coming online,” said Kapusta, vice president of advisory and transformation, in an interview with CRN last month.
Intel CEO Lip-Bu Tan and Broadcom CEO Hock Tan gave credence to the need for financial support to keep deals flowing this week, with the former citing “capital and the cost of capital” as a “challenge” and the latter saying, “the biggest constraint is capital.”
While Kapusta (pictured) said Nvidia’s financing moves could help the channel overcome these issues with customers, he noted that solution providers are still grappling with a global components shortage that is being driven by the overall flood of AI data center activity.
“Nothing feels sustainable about the current supply chain,” he said.
The GDT executive also acknowledged that the growing U.S. backlash against data centers means companies like his will need to work with customers “in how we get more creative [about] what some of these data centers even look like.”
Why Kapusta Thinks The Channel Could Benefit From Nvidia’s Moves
Kapusta said he believes Nvidia’s growing influence as a financial player in AI infrastructure deals presents a big opportunity for solution providers who sell, integrate and even manage AI server clusters in data centers.
“There’s a downstream effect there for the channel in that perspective of being able to secure more of these deals and make sure there’s funding behind them,” said Kapusta, who added that “there’s a lot of room for the channel” in the growing neocloud market.
While Nvidia has been making equity investments for years in customers, the vendor has been expanding its financing activities over the past year to more complex arrangements.
These include deals in which GPUs are treated as an “investable asset” and Nvidia provides a financial backstop in the form of a guarantee to help customers finance AI infrastructure purchases. In some cases, it has provided minimum-revenue guarantees in exchange for customers sharing a percentage of their revenues above the minimum threshold.
With GDT counting about 20 neoclouds as customers, Kapusta said roughly 25 percent of those companies have received equity investments from Nvidia, which has helped his company get AI infrastructure deals with such customers across the finish line.
While Kapusta said that none of GDT’s neocloud customers have entered the more complex financial arrangements that Nvidia is starting to pursue, a couple of those customers “are in very early discussions” for such deals.
“Through these new programs, we’re seeing more and more interest from the neoclouds in participating in that,” he said.
The Biggest Challenge Kapusta Sees With The AI Buildout
While the growing influence of Nvidia’s financial muscle with customers has attracted scrutiny from critics who call it another sign of an AI bubble, Kapusta said he believes the moves are sustainable because there is enough demand from enterprises who need to access compute without standing up new data centers.
“It’s an easier way of enabling those capabilities for the enterprise without dealing with some of the frustrations that come with the state of the market,” he said.
But enterprises seeking to build their own AI data centers are facing a major challenge that has plagued the industry at large since last year: the ongoing components supply crunch, reflected by shortages in not just memory chips but also GPUs, CPUs and other kinds of chips.
Echoing others in the channel, Kapusta said these shortages are causing cost increases and long lead times for data center products sought out by enterprises.
“If I have workloads and capabilities that I want to bring online now or even within the next quarter or two, given the state of ordering, architecting and running some of that on-prem, I may not get some of that for up to a year,” he said.
Enterprise customers also must grapple with the “different ways” manufacturers prioritize orders for such deals, according to the GDT executive.
“Some do it by size of order. Some do it by customer spend. Everybody’s kind of different there, so even just placing an order these days doesn’t guarantee you’ll get it in kind of that first-in, first-out fashion,” he said, noting that lead times can vary depending on which components are in supply and which aren’t.
With vendors varying in the way they allocate orders, Kapusta said a big part of GDT’s value proposition is to help customers, both enterprises and neoclouds, “manage that supply chain because it can be overwhelming, confusing—and change on a daily basis.”
How Kapusta Sees Data Center Backlash Impacting Customer Deals
With the U.S. data center backlash becoming a major political issue for the country’s midterm elections next month, Kapusta said he has seen “a little bit” of impact on GDT’s ability to find sites for customers to build AI clusters, though no projects have been outright cancelled.
This backlash has mainly come in the form of some states and localities putting into place moratoriums or other types of restrictions on new data center development, which he said has resulted in a “shift of focus into where organizations are looking to put capacity.”
It’s also making GDT and its customers think more creatively about what types of projects can get approved, which means “looking more at modular data centers and things that aren’t the behemoth of a build that’s an eyesore and tends to get public backlash,” he added.
“Are there ways to do more distributed type of data centers that tend to be more visually appealing and tend to be more accepted by communities?” Kapusta said.
But the executive pointed out that there are other challenges that are informing how enterprise customers approach new data center development. These issues include liquid cooling, which is becoming necessary for many state-of-the-art AI servers, as well as “some of the other heavy environmental requirements that are coming,” he added.
“We’re starting to see clients look at the cost of retrofitting an existing data center, saying, ‘Hey, this is way too expensive,’” Kapusta said.
If they don’t retrofit existing facilities, however, these customers may have to “fight for capacity” inside of a colocation center or another data center because “the neoclouds and hyperscalers are buying up all that capacity,” he added.
It’s this fight for physical data center space that is also putting the onus on enterprise customers to consider “more modular approaches” to development, according to Kapusta, which he believes can help speed up broader adoption in the enterprise.
“Are there ways we can put AI infrastructure next to your existing data center? [It would] be built ready for the power, cooling and other requirements that you need, but it saves you on a costly retrofit from that as well,” he said.
