When it comes to AI, we talk a lot about models, use cases and agents. But once initial tests need to become production applications, a very concrete question arises: Where will it all actually run?
Especially when sensitive business data, internal know-how or business-critical processes are involved, I also see this as a question of sovereignty. Companies should be able to decide for themselves which models they use, where their data is processed and how they secure their AI environment.
Having your own infrastructure can also make economic sense, particularly when AI applications run regularly and at high utilisation. It can also be tailored to your specific requirements. An internal AI assistant for technical documentation needs something different from an environment in which large language models and multiple AI agents run in parallel.
Moving from experimentation to action
From my perspective, however, the challenge today lies elsewhere: Many companies already have use cases that work technically, deliver clear value and are ready for the next step.
These are exactly the use cases we should start with.
With AI, a lot can be planned in advance. But the experience that really matters only comes from running it in production: Which model works for my requirements? How does usage evolve? What performance do I actually need? Where do bottlenecks arise? And what does that mean for infrastructure, operations and costs?
You cannot learn that from PowerPoint.
That is why I see little value in trying to plan the perfect AI infrastructure for the next five years today. What matters more is starting with use cases that are ready for production, building first-hand experience and using it to determine the next steps.
Getting started should not turn into an infrastructure project
Yet this is often where the next hurdle appears. Powerful GPUs are in demand, lead times are not always easy to predict, and servers, software and services need to come together in a working environment.
But companies that want to get started with a specific AI use case should not have to spend months assembling the infrastructure they need first.
With Lenovo AI Express, we aim to simplify that first step. The programme offers pre-validated configurations for different requirements, from getting started with inference and assistant applications to more demanding generative and agent-based AI scenarios.
For me, this is explicitly not about deploying the largest possible infrastructure straight away. The initial infrastructure investment must match the first step into production.
Speed is part of the solution
Availability is another important factor. For eligible Lenovo AI Express configurations, the target shipping timeframe is 15 to 25 business days, starting once the required ordering and approval processes have been completed.
At first, that may sound like an operational detail. But it is not.
An AI solution delivers no business value until it is running in production.
If companies want to build first-hand experience, they need to move into implementation. That requires both technical performance and rapid availability.
A GPU server alone is not an AI environment
Hardware alone is, of course, not enough. Lenovo AI Express combines validated ThinkSystem configurations with NVIDIA technology, enterprise AI software and services that support the transition into production. Experienced partners also play an important role in quickly turning infrastructure into a solution that can be used in production.
The goal is simple: Companies should not have to select every component themselves and integrate everything from scratch before they can start their actual AI project.
And above all: The first configuration is not a decision for all time.
Starting with a manageable use case in production builds first-hand experience. That experience makes it much easier to decide what is needed next: larger models, more users, additional applications or greater performance.
The infrastructure can then be developed further in a targeted way through the Lenovo Hybrid AI Factory.
Start. Learn. Scale.
For me, this is precisely the key point.
We need to move faster from AI experimentation to action.
Rather than trying to anticipate every requirement of tomorrow today, we should start with the use cases that are ready for production. Gain first-hand experience. Understand what is really needed. And then scale in a targeted way.
Lenovo AI Express provides a fast and predictable way to get started with your own AI infrastructure. This enables companies to build their applications under their own control and, above all, do what matters now: put AI into production and use their own experience to determine the next step.












