The current AI-driven data center landscape is packed with more of everything: more density, heat, digital tools, connectivity and, crucially, more challenges.
Amid this complex web of interconnected pressures, it can be difficult to identify just one issue as the defining concern. Yet today, more than ever, power is topping that list and steering critical decisions from where and when to build, to how enough energy can be delivered to support increasingly power-hungry compute.
As a result, power infrastructure has quickly transformed from a supporting utility function into a strategic enabler of deployment and growth. Power availability constraints are not only driving valuable innovation, but also accelerating the diversification of power strategies.
At their core, power strategies themselves are being expanded, reshaped, and redefined in order to transform the power bottleneck into a competitive advantage at the very heart of operational approaches.
In light of this transformation, Bhargava Srikantha, global marketing leader of data centers at Cummins, explores how AI-driven growth is colliding with power, grid, and execution constraints to shape the data center landscape, what the diversification of power solutions looks like in practice – plus how the company is enabling operators to navigate this landscape sustainably and at scale.
The new power reality
“AI has fundamentally changed the relationship between compute and infrastructure,” says Srikantha. “Typically, power infrastructure was designed to support predictable, relatively stable workloads. But AI is now driving unprecedented power density, faster deployment timelines, and much larger campus-scale requirements.”
In the data halls of just a few decades ago, there was a strong correlation between the white and gray space – both operated in relatively stable harmony. Today, the scale and speed of AI growth have completely redefined this relationship.
This pressure is actively forcing the industry to rethink long-standing assumptions around data center infrastructure.
“This is where innovation in the gray space, especially power infrastructure, is so critical,” says Srikantha. “Supporting the demand for AI is essential, and the industry is realizing that power availability has become one of the primary gating factors for growth.”
Increasingly, operators are able to secure land, financing, and compute roadmaps, yet power availability remains the central hurdle.
“The most common challenges we’re seeing are interconnection delays, transmission constraints, permitting complexity, and overall utility readiness for large AI campuses,” continues Srikantha. “Existing grid infrastructure in many markets wasn’t originally designed to support this unprecedented level of demand from hyperscale and AI-driven deployments.”
These extended timelines – whether related to substation upgrades or transmission delays – are significantly impacting project schedules and driving the necessity for a new landscape defined by diversified power strategies.
Whether it’s onsite power, energy-as-a-service, standby power where the grid remains the primary source, backup via diesel or alternative fuel options, or fully islanded power models, these approaches are increasingly being relied upon as foundational operational cogs – designed in from the outset rather than added on as a last-minute afterthought.
Diversification on multiple fronts
This reality is not only driving the diversification of power, but the expansion of the data center market itself. In some cases, the search for power is taking operators beyond traditional hubs and into emerging markets where land and energy are more accessible.
While no single location can tick every box, many rising secondary and tertiary markets that may lack immediate utility access can instead offer more favorable conditions for onsite or hybrid power generation.
“In the last few years, we've seen a significant increase in onsite power deployments across North America and other key markets,” says Srikantha. “Power is being delivered through various emerging technologies not just as backup infrastructure anymore, but as the primary source of energy in regions where the grid cannot provide access quickly enough.”
Similarly, the scale of AI growth and data center demand is actively remodeling market dynamics. Hyperscalers remain central, but growth is increasingly flowing across key partnerships and new customer segments, as Srikantha explains:
“What we’re seeing is that some hyperscalers are depending on AI growth through neoclouds or AI infrastructure developers. Instead of building everything themselves, they’re identifying their core competency as compute, for instance, and in regions where power access is constrained, they may instead buy compute or GPU capacity as a service, or lean on AI developers to build AI data centers where there’s no immediate grid access.”
Staying ahead in this competitive landscape will increasingly be defined by the ability to match the right power strategy to the right project – whether that’s grid-first, onsite generation or a hybrid model – without compromising reliability or scalability.
From isolated to integrated
The crux of this conversation inevitably comes down to securing enough reliable power as quickly as possible. Central to striking the right balance is early engagement with power specialists.
“Engineering collaboration is becoming much more strategic,” says Srikantha. “A few years ago, the question was simply: can you meet the spec and deliver the backup power needed? Now, we’re increasingly hearing: 'I don’t have access to utility power for several years. What technologies can you deploy onsite to give me access to power faster?'”
More and more, operators are relying on the right partnerships across engineering firms, consultants, utilities, and power providers to support deployment timelines. And as AI infrastructure continues to scale, coordinated engineering and infrastructure planning are becoming competitive advantages rather than optional steps in the process.
Bridging the power gap with collaboration extends far beyond people and into the tech itself. Prioritizing the best individual products that function effectively in isolation is no longer enough.
Instead, integrated power solutions are delivering infrastructure that functions as a coordinated ecosystem rather than a collection of standalone components.
For hyperscalers and colocation providers, this means thinking beyond standalone equipment and focusing instead on how generation systems, controls, cooling, and switching infrastructure can work together to support faster deployment.
“As we see more onsite power, there’s much greater collaboration and integration required between different components,” says Srikantha. “And as environments become more power dense, integration becomes critical for successful deployments.”
For Cummins, this end-to-end approach extends further again, directly to its customer relationships and working model:
“We consider ourselves a partner to our customers,” continues Srikantha. “This means providing early access to capacity and delivering integrated power solutions, whether that’s backup power or bridging power through natural gas assets integrated with other emerging technologies.”
This integrated strategy – where both systems and stakeholders work together as one – is becoming increasingly important for managing not only the power bottleneck, but also the volatile nature of AI workloads which introduces unique transient and power quality challenges, increasing the need to validate how generation, energy storage, controls, and other onsite assets perform together under dynamic operating conditions.
Cummins brings this collaboration to life through its advanced testing facilities such as the Power Integration Center (PIC) lab in Fridley, Minnesota – a 20,000-square-foot microgrid lab where customers work side-by-side with engineers to simulate real-world power scenarios before deployment.
“One advantage of having a power integration center lab at our facilities is that it helps validate how onsite assets can support different workload densities,” says Srikantha. “We work closely with hyperscalers and key customers to validate AI workloads on our integrated infrastructure solutions and predict failures ahead of time.”
The sustainability factor
Rising sustainability expectations introduce an additional layer of complexity. Operators are increasingly evaluating how to balance power resiliency and scalability with environmental objectives at the same time as energy demand continues to accelerate.
“As operators build higher gigawatt-scale sites, they’re also asking how they can remain environmentally conscious,” says Srikantha. “This requires two things. First, carbon accounting and transparency around how much carbon footprint is produced in manufacturing products. Second, how suppliers and the wider ecosystem are mitigating sustainability challenges.”
Cummins is advancing this transition through an end-to-end sustainability strategy that spans manufacturing, product development, and power system innovation, helping customers balance resiliency, scalability, and environmental objectives.
The company’s Destination Zero initiative reflects its long-term sustainability commitment, integrating 2030 environmental goals and 2050 targets into a broader strategy focused on decarbonization, responsible material use, and community impact.
“That's why we've enabled our products to use HVO fuel, which can help customers reduce lifecycle greenhouse gas emissions compared with conventional diesel, depending on feedstock source and accounting methodology,” adds Srikantha.
In addition, Cummins offers Tier 4 final solutions for applications where stricter onsite emissions requirements apply.
Cummins’ business and sustainability strategies are intentionally and intricately aligned through Destination Zero – the company’s commitment to sustainability and helping its customers navigate the energy transition while growing the business.
This strategy builds on the company's long-standing commitment to environmental sustainability, with focused efforts on three priority areas: decarbonization, materials, and communities.
Tomorrow’s energy landscape
Srikantha’s view of the path ahead is clear: “We expect the future data center energy landscape to become increasingly diversified and flexible. Traditionally, you may have seen only three or four major players in the power space. But as AI-driven demand continues to grow, we expect to see a much broader portfolio of solutions across the industry.”
Grid power will continue to play a central role, but onsite generation, energy storage, and advanced controls are expected to become even more important as operators navigate AI growth, resiliency requirements, and evolving market dynamics.
Infrastructure planning itself is also becoming more ecosystem-oriented. Operators are increasingly evaluating how different energy sources, operational strategies and resiliency models can work together to support scalability, reliability, and long-term flexibility.
“Our focus is helping customers navigate these transitions through broader power expertise, integrated infrastructure capabilities, and global execution support,” concludes Srikantha.
The AI-driven data center industry is entering a period where infrastructure flexibility and speed to power will soon become just as important as compute itself.
Navigating the growing energy access challenge demands a new approach built on innovation, diversification, and collaboration from the outset, actively transforming this constraint into a competitive advantage.
Comments