A bullish market or a period of volatility can quickly push trading activity into overdrive. More trades, more data, and more recalculations keep models running continuously, ingesting market signals and updating positions in real time. The supporting infrastructure is under more sustained pressure than ever.

These environments are feeling that strain acutely as artificial intelligence (AI) moves from pilot to operational deployment across financial services (FS). Trading, fraud detection, and risk management are all demanding more compute power—often at the same time. Digital Realty's 2026 Global Data Insights Survey shows that enterprises are moving decisively from AI planning to execution. A surprising 84% are actively executing a formal AI strategy or evaluating a partner to help execute one, while 79% of AI deployments are scheduled for 2026.

As those workloads scale, the infrastructure challenges become harder to ignore. Dense AI compute environments place additional demands on power delivery and cooling infrastructure. If organizations don't address these requirements, thermal constraints and infrastructure instability can compromise processing performance, service availability, and the predictable response times latency-sensitive financial applications require. That pressure will only intensify. Gartner expects generative AI adoption to be near-universal across banks by 2026 and warns that power shortages could constrain 40% of AI data centers by 2027.

Once AI supports live trading or intraday risk models, the question changes from compute availability to whether infrastructure can reliably power and cool higher-density workloads where they are needed. That is the gap many organizations are now encountering. Digital Realty's 2026 Global Data Insights Survey found that 40% of respondents cited a lack of specialized infrastructure as the top challenge in adopting a formal AI strategy, up from 9% in 2024.

According to Digital Realty’s 2026 Global Data Insights Survey, making incremental upgrades to existing infrastructure will not meet the demands of high-density AI in financial services. In fact, 49% of respondents identified ensuring infrastructure can handle AI power and energy demands as a critical requirement for a successful AI strategy.

Why financial services feel the pressure sooner

The FS industry has long relied on high-performance infrastructure, and that demand is only growing. Forrester predicts the sector will spend nearly half a trillion dollars on technology in 2026, as firms invest more heavily in the platforms, infrastructure, and partner ecosystems needed to operationalize AI at scale.

Financial services increasingly rely on a diverse range of AI workloads, each with distinct infrastructure requirements. High-frequency trading and real-time fraud detection demand predictable, low-latency processing, while AI model training, large-scale risk simulations, and backtesting prioritize compute capacity, high-bandwidth interconnectivity, and efficient power and cooling. As these workloads scale, financial institutions must balance performance, resilience, and infrastructure density according to their specific operational requirements. Traditional enterprise environments weren't designed to sustain these demands. Most were built around modest rack densities—often in the 5–10 kW range according to Uptime Intelligence—whereas modern AI workloads regularly exceed that by an order of magnitude. As graphics processing unit (GPU)-intensive, high-performance compute environments scale up, rack densities rise, and heat output increases, placing greater pressure on power delivery and cooling systems.

In financial services specifically, the consequences of getting this wrong are sharper than in most other sectors. Latency can affect trading performance and profitability. Infrastructure instability carries regulatory risk. Data residency rules limit how much workloads can be centralized. That means the industry is managing three pressures at once: denser AI workloads, higher power and cooling demands, and less flexibility in where those workloads can run.

High-density capability is therefore needed, not just at scale, but in the right locations. Digital Realty's 2026 Global Data Insights Survey found that 88% of respondents are following a distributed data approach, while 92% now tie their data location strategy to their AI plans. The research also found that 98% expect real-time AI workloads that require immediate data processing and response within the next 12 months. Many legacy bank-owned data centers struggle to support that model. They can run pilots in smaller environments, but scaling across the business is more complex and often more expensive when infrastructure isn't designed for higher-density workloads.

What AI-ready infrastructure needs

Making incremental upgrades to existing infrastructure won't meet the demands of high-density AI in financial services. As McKinsey notes, next-generation data centers are being redesigned around power and cooling as core components, rather than secondary considerations.

At Digital Realty, our approach reflects that reality in the way we build. We design our High-Density Colocation environments from the outset to support sustained AI workloads, with clearly defined power capacity per rack, scalable distribution architecture, and resilient power and cooling frameworks that deliver the reliability financial services environments demand. Starting at 30 kW per cabinet and scaling to 150 kW, our High-Density Colocation offers liquid cooling options, including Air-Assisted Liquid Cooling (AALC) and Direct Liquid Cooling (DLC)—purpose-built for environments where thermal management is critical.

As density rises, cooling must be factored into infrastructure design from the start, as traditional air-based systems become increasingly difficult to scale under sustained thermal load. That is why hybrid and liquid cooling are becoming central to AI-ready environments that meet demanding performance needs without sacrificing high levels of energy and water efficiency. At our Cloud House facility in London, for example, a pioneering closed-loop dock-and-river water cooling system supports energy-intensive workloads, dissipating heat more efficiently and reducing reliance on traditional air-based methods, all while maintaining the performance standards financial operations require.

Where AI infrastructure is deployed is as important as how it is designed. Financial institutions must consider not only proximity to exchanges and financial hubs for latency-sensitive applications, but also the location of enterprise data, access to cloud platforms, connectivity ecosystems, and regulatory requirements. While AI training workloads may prioritize power availability and compute scalability, real-time inference and transaction processing can benefit from deployment closer to data sources and critical applications. A distributed infrastructure strategy lets organizations place workloads where they can balance performance, connectivity, compliance, and cost. Our financial services solutions are designed with exactly this in mind—improving latency, performance, and customer outcomes by deploying infrastructure in the right locations, particularly in major financial hubs for time-sensitive transactions.

Repeatability matters too, as financial institutions deploy AI across regions. Through PlatformDIGITAL®, Digital Realty provides a globally consistent approach to deploying and scaling digital infrastructure. By combining standardized design principles with site-specific power, cooling, and connectivity capabilities, financial institutions can simplify multi-market deployments while adapting to local infrastructure availability, regulatory requirements, and workload demands. That, in turn, helps financial services firms improve execution speed, manage risk more effectively, and strengthen their competitive position.

Infrastructure is becoming a competitive decision

AI is raising the performance demands placed on financial services infrastructure. What were once isolated, high-intensity workloads are now part of standard day-to-day operations, placing sustained pressure on how infrastructure is designed and scaled.

That has implications beyond capacity alone. Performance and location correlate when infrastructure sits much closer to business, especially where latency, regulation, and data access affect how firms deploy workloads. Our blog on optimizing data infrastructure for AI in financial services explores this further, as does our recent colocation and financial services innovation piece—both of which offer practical perspectives on how firms are approaching this transition.

As AI becomes more deeply embedded in core financial services operations, the immediate challenge is to assess whether existing environments can support the next phase of demand. Power, cooling, and proximity now directly affect how quickly and effectively AI can scale. Firms that address these questions early will be better positioned to expand without costly trade-offs. You can also explore our guide to data center cooling innovations for a deeper look at where cooling technology is heading.

Let's talk about your AI infrastructure requirements

We work with financial services firms worldwide to navigate the move from AI pilots to enterprise-wide deployment, helping you decide where to start, what infrastructure you need, and how to deploy at scale without compromising performance or compliance.

If you'd like to explore how Digital Realty and Options can support your specific environment, our solutions architects are ready to help. Whether you're evaluating high-density colocation, hybrid cooling options, or a distributed infrastructure strategy across financial hubs, we can help you build a foundation designed for what's next.

Explore our financial services solutions →

AI Power & Cooling for Financial Services