Lenovo’s New AI Inferencing Servers Will Help Businesses and the Channel: Analysts

A trio of new AI inferencing server models unveiled by Lenovo at the recent Consumer Electronics Show will bring significant changes to the fast-moving world of enterprise AI as businesses move from AI training to AI inferencing over the next few years.

That is the opinion of two IT analysts who talked with ChannelE2E about how the industry will continue to evolve from training AI models to using those models to infer data to generate real-world, useful results.

The new Lenovo inferencing servers include the Lenovo ThinkSystem SR675i, which is built to run full large language models (LLMs) for the largest workloads and accelerated simulation in manufacturing, critical healthcare, and financial services environments; the Lenovo ThinkSystem SR650i, with accelerated AI inferencing power, high-density GPU compute, broad scalability, and easy deployment in existing data centers; and the Lenovo ThinkEdge SE455, a compact and rugged machine built for retail, telecom, and industrial environments that can bring AI inferencing capabilities anywhere data is located.

These new servers will allow businesses and channel partners to deploy them in traditional data centers, where they can do their work directly without having to rely on massive models from hyperscalers and hyperclouds, Jack E. Gold, president and principal analyst with J. Gold Associates, LLC, told ChannelE2E.

“While huge AI data center systems may get much of the press, we estimate that in the next two to three years more than 85% of AI workloads will be inference-based and not training-based, which is the primary use today,” said Gold. “The huge Nvidia GPU data center systems are not appropriate to bring AI to the masses, as AI needs to run more locally on edge compute and near the end use case.”

That is where more modest AI inference servers from Lenovo and other vendors, including Dell and HPE, “will be mission-critical for many organizations that want to run specialized AI models trained with custom data,” he said. “Companies also want to buy systems from their current suppliers, so the traditional server suppliers have a built-in relationship with customers they can leverage.”

In addition, said Gold, “the payback on these smaller-scale systems can be much higher than trying to ‘rent’ compute from the hyperscalers, particularly for edge-based systems.”

Moving From Building AI Models to Generating Business Value

Another analyst, Anurag Agrawal, founder and chief global analyst at Techaisle, said the new AI inferencing servers “are immediately useful because they address the cloud cost fatigue that many businesses are currently facing. As companies move AI from experimental pilots to 24/7 production, they reach a cost inflection point where the recurring expense of cloud inference becomes unsustainable.”

But by buying their own inferencing servers and deploying them in their own data centers, businesses will be able to bring their steady-state workloads on-premises, “offering a cheaper alternative to renting cloud capacity for constant tasks,” said Agrawal. “Furthermore, because models like the SR650i are built on standard enterprise form factors, MSPs can deploy them into existing data centers without requiring clients to overhaul their facilities, making them a practical solution for immediate cost control.”

At the same time, the powerful new servers will also allow businesses “to shift from building AI models to putting those models to work to generate business value,” he said. “In retail and hospitality environments, companies will use edge devices like the SE455 to run visual AI locally, enabling real-time loss prevention and frictionless checkout systems where data is processed instantly at the source rather than traveling to the cloud. This eliminates latency and bandwidth costs while improving responsiveness.”

Big Shifts Coming for MSPs and Other Channel Partners

For channel partners, Lenovo’s new AI inferencing servers will bring other significant changes, said Agrawal.

“The critical story lies in the shift for partners, specifically the mandate to specialize vertically or risk obsolescence,” he said. “Selling generic AI capacity is a race to the bottom; the real opportunity for MSPs is to build vertical-specific context on top of Lenovo’s hardware, selling outcomes like regional banking fraud detection rather than just servers.”

Another substantial change with the new inference servers, said Agrawal, is a transition to liquid cooling, which he called a major differentiator that partners need to understand.

“As next-generation AI racks approach megawatt density, traditional air cooling is becoming obsolete,” he said. “Lenovo’s Neptune liquid cooling technology allows partners to deploy high-density AI infrastructure into clients’ older, air-cooled facilities without risking leaks or requiring expensive retrofits. Combined with consumption-based models like Lenovo TruScale Infrastructure as a Service, this enables partners to offer a flexible financial model with the performance and security benefits of on-premises hardware.”

More Details on Lenovo’s New AI Inferencing Servers

The new inferencing servers are designed to help enterprises move from trained LLMs into real-world, autonomous applications while connecting data across cloud, data center, and edge environments, enabling AI to run wherever it delivers the greatest business impact, according to the company.

The AI inferencing server news builds on Lenovo’s newly unveiled AI Cloud Gigafactory with NVIDIA, a gigawatt-scale AI factory program designed to accelerate enterprise AI from development to production. Lenovo’s enterprise AI hardware is the foundation of what the company calls its Lenovo Hybrid AI Factory, a validated modular framework for building and using AI applications across a broad range of tasks.

The new AI inferencing servers are available with a choice of operating environments, including Lenovo Hybrid AI Inferencing with Nutanix AI, Red Hat AI, or Canonical Ubuntu Pro.