The Skyrocketing Cost of Powering AI: Tech Giants Grapple with Natural Gas Plant Expense and Delays

The insatiable demand for electricity to fuel the burgeoning artificial intelligence (AI) revolution is forcing major technology companies, including industry titans like Microsoft and Meta, into an increasingly expensive embrace of natural gas-fired power plants. While these companies have historically favored renewable energy sources for their data centers, a confluence of surging electricity needs, public opposition to new infrastructure, and critical supply chain bottlenecks is pushing them towards fossil fuels, albeit at a significantly higher cost. A recent report from BloombergNEF has illuminated just how steep this price hike has become, revealing a staggering 66% increase in the cost to construct a new natural gas power facility over the past two years.

This escalating expense comes at a time when the underlying commodity, natural gas, remains relatively affordable in the United States, even amidst geopolitical tensions like the ongoing conflict in Iran. However, the cost per kilowatt of generating capacity for a new combined cycle gas turbine (CCGT) plant has surged dramatically. In 2023, building such a facility cost less than $1,500 per kilowatt. By last year, this figure had ballooned to $2,157, a substantial increase that impacts the economic calculus for these massive infrastructure projects. Furthermore, the timeline for bringing these plants online has also lengthened, with construction now taking approximately 23% longer than it did just a few years ago. This combination of rising costs and extended project durations presents a significant challenge for companies racing to meet the immense power demands of AI development and deployment.

The Data Center Boom and Its Energy Footprint

Data centers, the physical hubs for computing power, storage, and networking, are at the epicenter of this energy consumption surge. Their role in processing the vast datasets required for AI training and operation makes them one of the primary drivers of increased electricity demand across the globe. This escalating need has not only compelled tech companies to invest heavily in their own power generation but has also put significant pressure on utility providers.

The urgency to secure reliable power sources has been amplified by directives from governmental bodies. For instance, the Trump administration had previously urged data center operators to "bring their own power," a policy that incentivized self-sufficiency in energy provision. However, the economic realities are such that utilities often pass on the costs of new power generation infrastructure to their customers, leading to potential increases in electricity rates for the general public. This dynamic, coupled with concerns about the environmental impact and land use associated with large-scale data centers, has fueled a growing wave of public opposition. This backlash is not merely a localized issue; it is a significant factor influencing where and how new data center capacity can be built.

Forecasting the Future Energy Demand

The scale of the projected energy demand from data centers is immense. While they are not the sole contributors to the growth in electricity consumption, they represent one of the fastest-expanding sectors. Projections indicate that by 2035, data centers will demand approximately 2.7 times their current electricity needs, escalating from around 40 gigawatts (GW) to an astonishing 106 GW. This forecast underscores the critical need for innovative and sustainable energy solutions.

A key factor contributing to this exponential rise in demand is the sheer size of new data center facilities. Currently, only about 10% of data centers are considered "large," with capacities of 50 megawatts (MW) or more. However, the trend is shifting dramatically. Over the next decade, the average data center is expected to exceed 100 MW, indicating a significant increase in the power footprint of individual facilities. This trend necessitates a corresponding surge in reliable and scalable power generation.

The Shifting Energy Strategy: From Renewables to Natural Gas

Historically, tech companies have demonstrated a strong preference for grid-connected data centers bolstered by Power Purchase Agreements (PPAs) for renewable energy sources such as wind, solar, and battery storage. These agreements offered a pathway to power their operations with cleaner energy while often providing cost predictability. However, the current landscape is forcing a reassessment of these strategies.

The twin pressures of escalating electricity demand, largely driven by AI workloads, and the increasing public scrutiny and opposition to new data center construction have pushed more companies to consider natural gas projects. This shift is not necessarily an ideological one but a pragmatic response to the immediate challenges of securing sufficient and reliable power in a rapidly evolving energy market.

The Turbine Bottleneck: A Supply Chain Crisis

The intensified pursuit of natural gas power plants has created an unforeseen consequence: a severe shortage of the critical component – gas turbines. These sophisticated pieces of machinery are essential for the operation of CCGT plants and constitute a significant portion of the overall construction cost, reportedly up to 30%. By the end of the current year, prices for these turbines are anticipated to have soared by a remarkable 195% compared to 2019 levels.

The manufacturing process for gas turbines is complex and does not lend itself to rapid scaling. The specialized techniques and intricate engineering required mean that production cannot simply be ramped up overnight to meet sudden spikes in demand. As a result, companies seeking to procure these turbines are facing extended waitlists, with delivery times now stretching into the early 2030s. This supply chain constraint further exacerbates the challenges faced by tech companies in building the necessary power infrastructure to support their data centers.

A Chronology of Escalating Demand and Costs

The current situation is the culmination of several years of rapid growth and evolving market dynamics:

• Early 2020s: Tech companies begin to solidify their commitment to AI development, recognizing its transformative potential. This leads to a significant increase in the demand for computing power and, consequently, electricity.
• 2023: The cost to build a new combined cycle gas turbine (CCGT) power plant hovers below $1,500 per kilowatt of generating capacity. Renewable energy PPAs remain a primary strategy for powering data centers.
• 2024: The price to build a CCGT plant surges to $2,157 per kilowatt, a 66% increase from the previous year. Construction timelines for these facilities also lengthen by approximately 23%. The demand for gas turbines escalates, leading to supply chain pressures. Public opposition to data center development begins to gain momentum in various regions.
• Late 2024: Prices for gas turbines are projected to be up 195% over 2019 levels, signaling a critical bottleneck in the energy infrastructure expansion for AI.
• 2025-2035: Data center electricity demand is forecasted to nearly triple, reaching 106 GW by 2035. The average size of new data centers is expected to more than double.

Analyzing the Implications: Economic and Environmental Trade-offs

The reliance on natural gas, even as a transitional fuel, carries significant implications. While natural gas is often considered a cleaner-burning fossil fuel compared to coal, it is still a source of greenhouse gas emissions, contributing to climate change. The long-term sustainability of powering AI infrastructure with fossil fuels is a growing concern for environmental advocates and increasingly for the companies themselves.

The economic implications are also profound. The substantial increase in construction costs for natural gas plants means that the investment required for AI infrastructure is escalating. This could impact the profitability of AI services and potentially lead to higher prices for consumers and businesses that rely on these technologies. The extended wait times for turbines also create uncertainty and risk for project planning and execution.

Furthermore, the increasing cost of building new power plants, whether gas or potentially renewable-reliant, could put upward pressure on electricity prices for all consumers. As utilities face higher capital expenditures, these costs are often passed on to the end-user, potentially exacerbating affordability concerns.

Divergent Paths: Alternatives to Natural Gas

While many tech giants are navigating the challenges of natural gas power plants, not all are solely focused on this path. Google, for example, has begun to articulate a different strategy for augmenting its power generation capacity. This approach emphasizes a synergistic combination of renewable energy sources with advanced, long-duration energy storage solutions.

Google’s strategy includes the integration of technologies like Form Energy’s iron-air batteries. These batteries are designed to provide electricity for extended periods, potentially up to 100 hours, offering a more consistent and reliable power supply than some shorter-duration storage systems. Unlike the soaring costs associated with gas turbines, solar panels and battery technologies have generally seen a continuous decline in their levelized cost of energy (LCOE) over time. This trend presents a compelling alternative for companies seeking to power their operations with cleaner and potentially more cost-effective solutions in the long run, although the scalability and integration challenges of such systems are still being addressed.

The BloombergNEF report serves as a critical wake-up call, highlighting the complex and costly reality of powering the AI era. The rapid escalation in natural gas plant construction costs, coupled with supply chain disruptions, underscores the urgent need for diversified and innovative energy solutions. As the demand for computing power continues to grow exponentially, the choices made today regarding energy infrastructure will have long-lasting economic, environmental, and societal consequences. The race to power AI is on, and it is becoming an increasingly expensive and intricate undertaking.