I recently took part in a DCD-led discussion on the role of the Independent Power Producer (IPP) in the data center world, alongside Rhett Weiss, chief development officer at Exus Digital. A couple of years ago the topic would have struggled to fill a session. Today it sits near the top of almost every commercial development conversation I have.

The reason is simple: the availability of reliable, secure, and preferably low-carbon power. Power is no longer a line in a development plan but increasingly sets the pace of the whole development and the capital cycle.

Year zero

You can date the change with some precision. It runs from late 2022, when ChatGPT went live and AI moved into public consciousness. Look at the IEA’s 2023 World Energy Outlook and data centers barely register as a source of demand. Look at the 2025 edition and there is a whole chapter on what they will do to power systems.

The industry had grown steadily since the 1990s, mostly in the background. What brought it into relief was the power intensity of AI workloads. Talk about a gigawatt campus and you are talking about a city the size of Munich, Germany, drawing that power around the clock. For two decades, electricity demand across America and Europe was flat to declining, and grid planning continued to make that assumption. Until the AI token factories started talking about building gigawatt-scale facilities in a matter of months.

The step-change caught a system that was not built to grow at this speed by surprise. AFRY’s latest independent market report for Switzerland gives a sense of the scale: power demand from Swiss data centers (one of the previously ignored locations but currently attracting a lot of interest) rises from around 6.1 TWh in 2030 to 10.8 TWh in 2040 and 15.3 TWh in 2050. Put another way, by 2050 Swiss data centers would require around two and a half times as much electricity as in 2030. In power-system terms, this is the emergence of a new industrial demand category that has to be planned for over decades.

The timing gap

My fellow panelist, Rhett Weiss, put his finger on the practical problem. There is now, in his words, a “major timing gap between when the grids can provide power and when the data centers are ready for the power.” In many markets that gap is several years.

This is where the IPP has changed roles. For a long time, these producers were discussed mainly in the context of renewable procurement and sustainability targets. Now they are part of the core delivery model for data center power, especially in the US where demand is greatest. As Rhett put it, “IPPs are also now serving a role of potentially closing or shortening that gap, able to provide power sooner than grid power.”

I want to be clear that the issue of grid connection has not gone away. The hyperscalers and their preferred developers still want a traditional high-voltage connection. The catch is that it may not arrive when the project needs it, so the energy plan has to cover several phases: power before the grid, power at energization, and power as the site scales.

Bring your own power, and everything around it

BYOP “Bring your own power” has become useful shorthand, though the real list of options runs far longer: bridge power, private wire, behind-the-meter generation, grid reinforcement, brownfield redevelopment, renewables, gas, fuel cells and, further out, nuclear.

The clearest large-scale example is Project Kilby in West Texas. The Microsoft and Chevron project, over two gigawatts built on a 20-year gas PPA with no grid connection at all. It works: the US has the gas, the land, and the regulatory approach for a scheme on that scale.

Much of Europe does not, which is why the answer there tends to be more local and more case-by-case. In markets such as Portugal, where 706 zero or negative price hours were recorded in the first half of 2026 alone while 2,150 MW of BESS capacity is already in process, the opportunity is increasingly about flexibility as much as it is about generation.

Site selection has gone local

One of the clearest trends is the move away from a narrow cluster mindset. In Europe, the FLAP-D markets, Frankfurt, London, Amsterdam, Paris and Dublin, have become badly constrained by grid capacity, land availability and permitting pressure. Attention is therefore shifting to markets such as Italy, Poland, Greece and Croatia, where developers may find more headroom and more accommodating policies, provided they understand local grid rules, permitting processes and power-market conditions.

This is a more pragmatic view of the whole picture: power, regulation, permitting, and the plain ability to get things done. Brownfield sites, or already “powered land” are part of this story. As Rhett noted, they were largely in disfavor for a long time when redevelopment looked too complicated. An existing connection changes that. “The fact that those connections already exist makes them a lot more valuable now than they used to be,” he said.

Why capital alone does not solve it

We tend to assume that companies with this much financial firepower can do anything. However, working at a local level tells a different story. Grid connection processes are being rewritten market by market. They are intensely local. Rhett described filing two applications in Germany just to cover two readings of an ambiguous process. Permitting, language, local relationships and the practicalities of getting things done all take time, and no data center company is set up to manage that across several continents at once.

That is the opening for the IPP. The role of the IPP varies across geography and for good reason: demand density, how power is regulated in front of and behind the meter, and the local availability of resources all differ. A fuel cell needs a gas line; where there is no gas, that option falls away. The value lies in taking local energy complexity off the developer’s plate and turning it into a delivery path that is bankable.

A relationship that runs both ways

This works in both directions, as data centers need power in several forms and cannot rely on the grid alone. IPPs need customers strong enough to underpin new generation. Rhett called it a “symbiotic relationship,” and made the financing point plainly: an IPP needs a bankable customer, a customer with high creditworthiness. Large data center operators provide exactly that. There is a system benefit too. IPPs are sometimes seen as adding complexity to the grid. In practice, they relieve constrained networks and help sites energize while operators work through their queues.

Different speeds, same direction

The energy and data center industries will work together more closely, out of necessity, yet they will keep running at different tempos. Rhett noted that “the power industry, for good reason, has operated at different – slower – speeds compared to the high-tech industry.” That caution is deliberate. When the power system fails, the consequences are severe, and it carries obligations around safety, reliability and public service that digital capital does not.

What gives me confidence is that the technology leaders now understand this. Microsoft CEO Satya Nadella talks about tokens per dollar per watt as a core measure of Microsoft’s success and indeed of future GDP growth. Few big-tech chief executives reach for the unit “watt” when they speak to investors. President and CEO of NVIDIA, Jensen Huang, described an AI stack as a ‘Five-Layer-Cake’, whose foundation is energy. If you get that wrong, then the rest is academic, since you cannot turn the switch on.

For developers and operators, a few things follow:

• Put power strategy at the very start of development, not late in the process.
• Treat optionality as the goal. Needs will change quickly and in ways none of us can predict.
• Judge sites on power, regulation and community acceptance together, not on demand alone.
• Choose partners who understand generation as well as site risk, local rules, grid limits and customer credit.
• Think and act locally.

A final thought

The last decade of digital infrastructure was a story of near-frictionless scaling. The next one will be shaped by how quickly power can reach the right place, at the right time, at acceptable cost. Vacancy rates and future leases show that demand is not in doubt. What matters now is optionality, the ability to keep a project viable as conditions shift, and the local knowledge to make it real. That is where the independent power producer has quietly become one of the most valuable partners in the room.

This article is a reflection following a DCD-led broadcast titled ‘The role of the IPP in data center growth’. Watch the full conversation here.