The AI arms race has become obsessed with models. Every week brings another breakthrough, another benchmark, another announcement that promises faster reasoning, smarter agents, or more capable automation.
Yet while the software grabs the headlines, a quieter battle is unfolding beneath the surface—one that could have a far greater influence on who ultimately wins.
That battle is being fought in data centers, on power grids, across fiber networks, and inside the facilities that keep AI running around the clock.
In his latest video, Andrej Radonić, founder and CEO of interSales AG, argues that the industry’s biggest challenge isn’t building more intelligent AI, but building the infrastructure capable of supporting it.
It’s easy to think of AI as just another software layer. Users type a prompt, receive an answer, and never see the extraordinary amount of infrastructure required to make that exchange happen. Behind every response, however, GPUs are processing workloads, servers are consuming power, cooling systems are removing heat, and networks are moving enormous volumes of data between applications and users.
Watch the full video here:
Five lessons for the Cloud community
1. AI is bringing infrastructure back into the spotlight
AI isn’t replacing the importance of infrastructure—it is dramatically increasing it. As AI workloads become embedded in search, productivity software, customer service, and enterprise applications, infrastructure is no longer simply supporting innovation. It is becoming one of the defining factors that determines how quickly organizations can innovate in the first place.
This dynamic will feel familiar to those who watched the Cloud evolve and grow: big, heavy physical elements can be ignored when they’re not right in front of your face.
2. Access to power is becoming a competitive advantage
The Cloud industry has always cared about location, but AI is changing what makes one location more valuable than another. Connectivity, regulation, and customer proximity remain important, yet the availability of reliable electrical capacity is rapidly moving to the top of the list.
As the video explains, announcing a new AI data center is relatively easy. Building one that actually has access to sufficient grid capacity, substations, transformers, permits, and local infrastructure is considerably harder. That distinction between announced capacity and operational capacity may become one of the biggest differentiators in the years ahead.
For Cloud providers, this represents a strategic shift. Access to power is no longer just an operational concern—it has become a business asset.
3. Efficiency may drive greater sustainability
Every increase in computing density creates another engineering challenge. High-performance GPU clusters generate far more heat than traditional workloads, forcing operators to rethink how they design and manage modern data centers. Technologies such as liquid cooling, direct-to-chip cooling, and new rack architectures are becoming essential rather than experimental.
What makes this particularly interesting is that efficiency is no longer simply a sustainability metric. The organizations that can deliver more compute from every watt of electricity, every rack, and every square meter will enjoy meaningful commercial advantages. In an increasingly competitive AI market, operational efficiency is becoming directly linked to profitability.
4. Connectivity could be just as important as compute
Training large language models may happen wherever power and land are available, but running AI applications in real time is a different challenge altogether. Enterprise AI depends on fast access to business systems, databases, APIs, and customer information, making latency a critical part of the user experience.
That creates opportunities beyond the hyperscalers. Regional Cloud providers, colocation facilities, and specialist infrastructure operators may not build the world’s largest GPU clusters, but they can deliver low-latency services, regional compliance, access to sovereign data, and proximity to customers that many AI deployments require. Specialization may ultimately prove more valuable than scale for many providers.
5. The Cloud industry has an opportunity—not just a challenge
It’s easy to view AI’s infrastructure demands as another problem to solve, but the video ends on a more optimistic note. AI is making the physical foundations of the internet strategically important again, highlighting the value of expertise in data centers, networking, operations, and infrastructure management.
For years, much of the industry’s attention focused on the software layer. AI is shifting that balance, creating fresh opportunities for organizations that can build resilient infrastructure, optimize operations, and deliver the reliable services that increasingly intelligent applications depend on.
The winners of the AI era may not simply be those with the best models, but those capable of powering, cooling, connecting, and operating them at scale.
AI, Cloud, and the MSP community
Although much of the conversation focuses on CSPs, the implications extend throughout the technology ecosystem.
For example, customers are already asking MSPs about AI strategy, and those conversations are quickly expanding beyond model selection to include performance, sovereignty, resilience, security, and cost optimization.
That means successful MSPs will increasingly need to understand the infrastructure that underpins AI as well as the applications themselves. Advising customers on where workloads should run, how latency affects user experience, or what infrastructure choices best support compliance will become an increasingly valuable part of the managed services proposition.
Those are exactly the conversations shaping the future of the industry.
MSP GLOBAL is where those who build and manage Cloud infrastructure can continue the conversation with those who use it.
Join fellow technology leaders, MSPs, and infrastructure experts to explore the trends that will define the next generation of managed services.
The two-day event takes place near Barcelona, 21-22 October. Head over to the MSP GLOBAL Knowledge Hub and sign up for the newsletter to get your free registration code. See you in Spain!
Andrej Radonić is the founder and CEO of interSales AG, a German software and technology company he established in 1998. Under his leadership, interSales has evolved into a provider of enterprise e-commerce solutions, ERP software, and cloud-based business platforms for customers across multiple industries.
Alongside his entrepreneurial work, Andrej has spent more than 25 years as an independent IT journalist and industry analyst. He has written for leading German technology publications including Linux Magazin, IT-Administrator, WindowsPro, and heise / c’t, covering enterprise IT, data center solutions, hosting infrastructure, cloud computing, and open-source technologies. He is also the author of a book on the Xen virtualization platform.
