Delays in data center construction are severely impacting the global AI capacity pipeline as backlogs in grid interconnections are hindering new power connections, while increasing lead times for power transformers are forcing Software as a Service (SaaS) founders to rethink their cloud pricing strategies. This infrastructure crunch may quietly reshape their financial planning.

A recent report on the growth of large data centers and “AI factories” reveals that developers are looking to add at least 16 gigawatts of new power capacity by 2026. However, progress has been slow, with construction starting on only about 5 gigawatts so far.

The report raises concerns that as much as 30 to 50 percent of the anticipated 2026 capacity could face delays, pushing some of this critical infrastructure into 2027 or later, all while demand for AI keeps climbing.

Big Tech executives are speaking out about a significant hurdle they’re facing: power supply.

Marsden Hanna, Google’s global head of sustainability and climate policy, mentioned that “transmission barriers are the number one challenge we’re encountering on the grid.”

He pointed out that utilities often require four to ten years to connect new loads, and in one instance, a utility told Google it could take a staggering 12 years just to evaluate an interconnection request. This situation reflects the pressing challenges the tech industry faces as it strives to keep up with the growing demand for energy to support AI advancements.

Why “announced” capacity keeps vanishing

The widening gap between what’s promised and what’s actually delivered in data centers can largely be attributed to interconnection queues that don’t align with the timelines for building these facilities. Recent data from Berkeley Lab highlights a troubling trend: the average wait time for interconnection has more than doubled.

For projects finished between 2000 and 2007, it was less than two years, but for those built between 2018 and 2024, it now exceeds four years. Alarmingly, by the end of 2024, only 13 percent of the capacity that applied for interconnections from 2000 to 2019 will have become operational.

Moreover, research from Berkeley Lab shows that the majority of interconnection requests don’t even get completed—over 70 percent are withdrawn, and only 19 percent (amounting to 14 percent of the total capacity) submitted from 2000 to 2018 were realized by the end of 2023.

This queue situation is particularly significant for software leaders, as the cloud supply often hinges on the same limited metropolitan areas and substations. In Northern Virginia, the largest data center market globally, a recent CBRE report revealed a colocation vacancy rate of just 0.72 percent during the first half of 2025, with preleasing extending into areas expected to be operational by 2027 and beyond.

Delays in data center development also have a ripple effect, influencing AI capacity, grid interconnections, cloud pricing, and even SaaS founders and power transformers. Utilities treat these facilities like small cities, continuously drawing power and pushing every dependency to the brink.

The quiet culprit: transformer lead times

Power bottlenecks don’t just exist in the realm of permits and studies; they also extend into the factories that produce the hardware that moves electricity. According to experts, lead times for large power transformers now average around 128 weeks, while generator step-up units can take about 144 weeks to deliver, with some shipments taking even longer.

This delay creates a tough reality for developers. They can lay the foundation, erect the building, and yet still find themselves waiting for years for the last piece of equipment that will energize their site. In the meantime, the grid struggles to handle an ever-increasing demand without additional transmission infrastructure. Reporting that cited Google noted that our aging infrastructure often fails to keep pace with the concentrated demands placed on it.

Marc Ganzi, CEO of DigitalBridge, summed up the situation during a 2024 earnings call, stating, “Power is really the constraining factor.” He emphasized that this issue would become more apparent in the next two years as demand continues to rise.

The delays in data center construction link together AI capacity, grid interconnection, cloud pricing, SaaS founders, and power transformers, creating a disconnect in the industry. Software develops in weeks, but the delivery of new power takes years, leaving a significant gap between technological advancement and infrastructure readiness.

What it changes for global SaaS companies

For most SaaS companies, the challenges won’t manifest as simply “no cloud.” Instead, you’ll likely notice a significant reduction in flexibility.

First, as capacity tightens, many customers are starting to prelease more, make longer reservations, and choose fewer “nice-to-have” regions. A JLL report from 2025 highlights this trend in Northern Virginia, revealing an impressive 87 percent preleasing rate for inventory scheduled to come online in 2025-26. This indicates that large customers are securing their space well in advance of when they actually need it.

Second, as big buyers reserve more supply, smaller customers face tough decisions about costs, location, and guaranteed availability. This is especially true for AI-heavy workloads that require dense racks and stable power.

Finally, founders and finance teams should prepare for infrastructure planning to increasingly influence product strategies. Enterprise buying behavior indicates that many companies are likely to concentrate their AI spending with fewer vendors in 2026. This shift could further strain limited infrastructure resources, benefiting the largest platforms while sidelining smaller providers.

As data center delays continue to tighten AI capacity, issues such as grid interconnection, cloud pricing, and power constraints become even more critical. Hyperscalers can finance their power strategies in ways smaller firms cannot, widening the access gap in a landscape where supply remains scarce.

What Founders Should Do

Founders might not be able to fix the power grid, but they can definitely lessen their exposure to unexpected shortages. Instead of focusing solely on infrastructure, it’s crucial to develop a comprehensive “capacity plan.” Treat factors such as region, reservation terms, and ramp-up timelines as essential constraints, just as you would with latency and compliance.

Be sure to pressure-test vendor promises against real-world progress. For instance, compare the ambitious 16-gigawatt target for 2026 with the mere 5 gigawatts currently under construction. This comparison offers a straightforward way to gauge the credibility of timelines that seem too smooth to be true.

Design for portability from the start. While multi-region and multi-provider architectures might require a higher initial investment, they can safeguard your launch schedules if a single region faces high demand.

Keep a close eye on power equipment lead times, just as you would with chips. Delays in transformers can halt entire campuses, even after construction is complete.

Data center delays will continue to shape AI capacity, grid connections, cloud pricing, SaaS founders, and power transformers until grid studies accelerate, equipment backlogs are addressed, and developers stop selling ambitious timelines based on best-case power-delivery scenarios.

Follow USTechTimes on Facebook, Twitter and Linkedin for in-depth news of market trends, funding updates, and regulatory changes affecting startups in USA.

We Recommend: