Summary:
- Cloud giants are reevaluating their strategies due to GPU scarcity, grid connection delays, and AI’s increased demand for compute power.
- Hyperscalers are expanding their data center capacity to meet growing demands, with a focus on power availability and efficiency.
- The industry is shifting towards low-carbon energy sources and secondary markets to address power constraints and optimize operations.
Article:
In the year 2025, major cloud companies found themselves at a crossroads, rethinking their approaches in response to challenges such as GPU shortages, delays in grid connections, and the surging demand for computing power driven by artificial intelligence. These factors forced them to reconsider their energy, network, and risk management strategies, with electricity emerging as a critical limiting factor.
Hyperscalers, the giants of the cloud industry, embarked on a mission to expand their data center capacity on an unprecedented scale. This involved not only building new facilities but also ensuring that these expansions were powered cleanly and reliably, securely connected across regions, and resilient in the face of concentrated capacity and growing risks. The decisions made by these companies were increasingly influenced by factors like grid access, local regulations, permitting timelines, and the speed at which they could bring new capacity online.
A key trend that emerged was the divide between owned campuses, particularly in the US, and leased capacity from colocation providers in other regions like EMEA, APAC, and China. The mix of build and lease strategies was driven by the need to overcome challenges related to electricity scarcity, slow grid connections, and permitting obstacles, highlighting the importance of time-to-power in decision-making processes.
As the demand for cloud infrastructure continued to rise, enterprise spending in this sector reached new heights in the second quarter of 2025. The dominance of industry leaders like AWS, Azure, and Google Cloud was challenged by the emergence of "neoclouds," privately financed specialists focusing on building ultra-dense campuses in power-rich secondary markets. These neoclouds gained momentum as they were better equipped to meet the escalating demand for AI-related services, surpassing the capabilities of the Big Three in sourcing electricity and accelerators.
In response to the growing pressure to meet the demands of AI workloads, cloud providers have focused on refining their architectures, improving efficiency, and aggressively expanding their infrastructure. However, the pace at which new capacity can be added ultimately depends on factors like available chips, grid interconnects, and permitting processes, underscoring the critical role of accelerators and power supply in shaping the industry’s growth trajectory.
The proliferation of AI-driven data centers has raised concerns about heightened credit and overbuild risks, prompting developers to seek out power-rich remote sites, forge direct utility partnerships, and implement on-site generation to mitigate financial exposure. The industry’s shift towards AI factory campuses with surging rack densities has underscored the importance of disciplined power sourcing and risk-sharing finance to prevent the data center boom from tipping into overbuild and credit stress.
Looking ahead, cloud companies are investing in low-carbon energy sources and securing long-term solar procurement deals to ensure a dependable and cost-effective power supply. This shift towards direct power strategies reflects a broader trend in the industry, where the focus is no longer just on purchasing electricity from the grid but on securing megawatts to support sustainable growth.
As the cloud industry continues to evolve, new data center capacity is being directed towards power-rich secondary markets where land is more plentiful and interconnects are faster. This trend is driven by the rising intensity of AI compute workloads, which require ample power resources to operate efficiently. The concentration of global hyperscale capacity in a few key metro areas highlights the importance of reliable and affordable power sources, as well as permissive siting policies, in determining the location of data center facilities.
In light of recent incidents like the AWS outage in Northern Virginia, cloud providers are under increasing pressure to adopt resilient architectures and diversify their supply chains to mitigate the risks associated with centralized infrastructure. The episode served as a wake-up call for the industry, highlighting the need for improved risk management strategies and supply-chain diversification to ensure uninterrupted service delivery.
Investment in subsea cables is expected to remain robust in the coming years, driven by the growing demand for cloud services, AI applications, and content delivery. Hyperscalers are co-investing in high-capacity routes to bolster their infrastructure, while also addressing security concerns related to physical sabotage, cyber threats, and potential vulnerabilities posed by quantum computing advancements. As traffic volumes increase and geopolitical risks escalate, subsea cables are becoming strategic lifelines that require enhanced protection and resilience measures.
In conclusion, the year 2025 was defined by a renewed focus on power availability as the primary driver of decision-making in the cloud industry. Looking ahead to 2026, industry leaders are expected to prioritize securing on-demand power, expanding into power-rich secondary markets, and investing in subsea infrastructure to diversify their operations and ensure the scalability of their services. The future of the cloud industry lies in striking a delicate balance between growth and sustainability, with a strong emphasis on optimizing operations to meet the evolving needs of a digital-first world.
