Leaseweb: AI Infrastructure Is Not One Size Fits All Anymore, Podcast

“Create your value and your wealth as a strategic consultant instead of being a technology vendor.” In this Technology Reseller News podcast, Tim Mandell of Leaseweb launches an ongoing infrastructure series focused on how MSPs can build successful AI practices—and why AI infrastructure can no longer be treated as a one-size-fits-all proposition. Leaseweb is a 29-year-old, private and sovereign-by-design infrastructure-as-a-service provider. The company works with MSPs by supplying the underlying infrastructure while allowing the service provider to own the customer relationship and deliver the higher-value solution. Mandell says one of the biggest mistakes organizations make is starting with the hardware. The AI conversation is often dominated by GPUs, including the latest high-performance systems, but the most powerful hardware is not automatically the right infrastructure for every workload. “You have to understand what problem you’re trying to solve before assuming the technology is going to solve it,” Mandell says. AI training, inference and production workloads can have very different requirements. For many MSP customers, inference is particularly important because performance depends heavily on latency, data location, storage and connectivity. Mandell points to an MSP that was losing customers because infrastructure had been placed in the wrong geographic location, creating latency problems that degraded the customer experience. GPUs may get most of the attention, but Mandell says storage can sometimes become the larger and more complicated infrastructure challenge. Organizations can underestimate both the amount of data required and how quickly that data must be retrieved. The same problem applies to sizing. MSPs can overengineer an environment and spend unnecessarily, or build something that lacks the flexibility required as customer demand grows. “There really is no such thing as one AI infrastructure,” Mandell says. “Different applications require different compute, bandwidth, storage and placement.” For MSPs, that creates an opportunity to move away from simply selling infrastructure and toward becoming strategic advisers. Leading with a particular server, GPU or technology stack turns infrastructure into a commodity where price becomes the primary differentiator. Instead, Mandell recommends beginning with the customer’s business objective, asking where the organization is headed over the next several years and then designing the infrastructure around those outcomes. “The infrastructure is going to be what it needs to be based on the business outcome,” Mandell says. “But if you lead with infrastructure, that becomes the foundation of the relationship.” His takeaway for MSPs is straightforward: ask more questions, understand the workload and build value around the customer journey rather than the technology itself. Visit Leaseweb.com to learn more.