Colocation vs. Cloud for Financial Services

About 10 years ago, financial institutions began to shift away from owned-and-operated data centers and toward the cloud. At the time, analysts from Deutsche Bank said that some IT executives from big global banks intended to go from zero use of IaaS cloud to about 30% in just three years.[1] The prevailing belief was that adoption would continue to accelerate from there, and that public cloud could become the default home for financial workloads.

This prediction didn’t fully materialize.

Financial institutions learned that some workloads thrive in the public cloud, while others require greater control, predictability and proximity. What’s often described as “cloud repatriation” is better understood as something more strategic and more mature: cloud rebalancing.

Financial institutions may be rethinking the way they use public cloud, but they’re not leaving it. Instead, they’re aligning different workloads to different environments to get the best possible balance of cost-efficiency, performance benefits, and risk mitigation.

Now, a new wave of innovation is putting that approach to the test.

As financial institutions scale production AI and begin supporting regulated digital assets including stablecoins, infrastructure decisions are becoming more complex and consequential. These workloads introduce new demands around cost-efficiency, data control, latency and regulatory compliance that a public cloud-only model wasn’t designed to meet.

The result is a shift toward hybrid architecture: combining colocation and public cloud to balance flexibility with control and innovation with stability.

AI: From experimentation to production

Financial institutions typically didn’t start their AI journey in colocation. Many started in the public cloud, which was the right choice for their needs at the time.

The public cloud offered on-demand access to high-performing infrastructure, enabling teams to experiment quickly, iterate models and prove value without up-front investment. This enabled development and testing teams to build and test models in real-world situations and begin to understand the potential benefits that AI can bring to their business.

Moving from experimentation to production is an entirely different challenge.

Production AI for financial services introduces sustained, high-density workloads that require continuous compute and create increased scrutiny around where data resides and how it’s processed. Also, financial institutions likely need to make significant investments in the modern infrastructure necessary to power their AI workloads, including GPU hardware.

These pressures are amplified by ever-stricter regulatory requirements, particularly around operational resilience, and the need to manage sensitive data and intellectual property securely. These companies’ infosec teams are working overtime to balance innovative new strategies against infrastructure cost, stability and scalability.

This is where over-relying on public cloud services can become problematic.

To scale AI effectively, financial institutions must balance three critical requirements:

• Cost-efficiency: If financial institutions default to public cloud for all AI workloads, they’ll miss out on cost-savings opportunities. Cloud uses a pay-per-token model, which can be fine in some cases, but often gets very expensive for consistently demanding production workloads.
• Control and compliance: When financial institutions put data in the public cloud, it means letting that data leave their perimeter. There is now a shared ownership model in place, which reduces control by definition. This can create risk and compliance issues. Also, they can’t ensure proximity between data sources and processing locations, which is essential for low-latency AI inference.
• Convenience: Historically, deploying a new colocation point of presence would require lengthy design and supply-chain cycles, managed by in-house experts that not all organizations have. However, colocation providers increasingly offer fit-out, remote hands and managed services. This allows customers to tap into institutional capabilities that make deployment easier.

No single infrastructure model can meet all these requirements on its own. That’s why financial institutions are adopting a hybrid approach: using public cloud for speed and flexibility, and colocation for control, performance and cost-efficiency.

While the benefits of hybrid infrastructure are now widely accepted, there’s a big difference between a basic hybrid multicloud model, where workloads are siloed into different environments, and truly integrated hybrid multicloud, where workloads can move easily between environments whenever the need arises. To move AI from proof of concept to production at scale, financial institutions need the latter.

Digital assets: From startup mentality to institutional stability

The same infrastructure dynamics shaping AI are now emerging in the world of digital assets, especially as financial institutions move from experimentation to adopting regulated assets.

For years, crypto largely operated outside the traditional financial system. It was the realm of startups that prioritized rapid innovation over control and compliance.

New regulations, like the GENIUS Act and CLARITY Act in the U.S. and the MiCA regulation in the EU, are now enabling traditional financial services institutions to hold, issue and trade assets like stablecoins for the first time.

These regulatory changes, coupled with the ability to tokenize real-world assets (RWAs), are leading to predictions of huge growth in the financial services industry. For instance, a report from Ripple and Boston Consulting Group forecasts that asset tokenization will expand into a $19 trillion business opportunity by 2033.[2]

Enabling digital assets at institutional scale will require a new approach to the underlying physical infrastructure:

• Traditional finance is highly centralized, often using a single database to record transactions no matter where they took place.
• With crypto, control is decentralized, which means that databases must also be decentralized. Transactions must be recorded and replicated across many distributed blockchain ledgers.

Just like with AI, financial services firms are learning that a cloud-first approach is not right for the needs of digital assets. While public cloud does offer some advantages, such as the ability to quickly deploy compute and storage in new locations, colocation can meet specific infrastructure requirements:

• Private, low-latency, deterministic networking: Ensuring efficient markets and enabling trade execution at institutional scale will require market-making arrangements with reputable liquidity providers. In turn, there must be proximity and direct connectivity between exchanges and chains and the market-making algorithms. Institutions can achieve this by partnering with a colocation provider that also offers dedicated, vendor-neutral, low-latency connectivity solutions.
• Resilience, control, security and compliance: As traditional institutions and the vendor community build out pre-trade and post-trade workflows for digital assets, they’ll need digital infrastructure that can keep those workflows protected and available. With colocation, they can build their infrastructure environment with these exact needs in mind.

What this means for financial institutions

AI and digital assets are reshaping the requirements for financial infrastructure. Decisions about where and how workloads run are now strategic imperatives, with direct implications for cost, risk, and long-term competitiveness.

Several priorities are emerging for financial institutions:

• Design for a hybrid future: Cloud-first has evolved into hybrid by design. Institutions need to deliberately align workloads to the environments that best meet requirements for cost, control and performance, rather than defaulting to a single model.
• Plan beyond experimentation: What works for pilots won’t scale to production. As AI and digital assets move into core operations, infrastructure decisions must support sustained demand, predictable costs and regulatory compliance over the long term.
• Treat infrastructure as a competitive lever: The ability to deploy workloads in the right locations, with the right connectivity and control, will directly impact speed to market, resilience, and customer experience.
• Choose partners that extend capabilities: The right colocation partner provides not just space and power, but also low-latency connectivity to cloud providers, ecosystems and expertise that accelerate execution and reduce complexity.

Equinix enables hybrid architecture at scale

Executing a hybrid strategy is about more than just combining environments. It’s about connecting them in a way that delivers control, performance, and flexibility at scale.

This is where the right infrastructure foundation becomes critical.

At Equinix, we help financial institutions operationalize hybrid infrastructure by bringing together the key elements required to support production AI and institutional-grade digital assets:

• Regulatory experience: In November 2025, Equinix was designated as a critical ICT third-party provider (CTPP) under the EU’s Digital Operational Resilience Act (DORA). The same regulator that met with financial services organizations to support their DORA compliance in January 2025 is now visiting Equinix data centers and partnering to drive alignment on operational resilience.
• Global reach and local control: Equinix operates 280 colocation data centers in 77 markets across 36 countries, so banks can deploy in all the right locations to enable low-latency AI inference and ensure control over their data sovereignty requirements.
• Ecosystem access: Equinix hosts about 3,000 cloud and IT services providers, many of which are easily accessible via Equinix Fabric®, our virtual interconnection solution. Financial services companies can also access hardware from ecosystem partners, like IBM Z mainframes, NVIDIA, Dell, HPE, and more.

More than 1,250 financial institutions have chosen Equinix as their colocation parter, and the vast majority leverage Equinix to support their hybrid architecture.

Rather than choosing between public cloud and colocation, financial institutions can use both strategically. The result is a hybrid model that supports innovation without compromising on control, enabling organizations to optimize their workloads in the right environments. Hybrid architectures have matured to the point where CIOs and CTOs are now choosing the right places for the right workloads. It’s taken a decade of learning to get to this point, but we’ve very much arrived.

Learn more about why leading financial institutions are rethinking infrastructure in the AI era: Read the white paper, “From cloud-first to private-first in financial IT infrastructure.”

[1] Steven Norton, Big Banks Starting to Embrace Public Cloud, Deutsche Bank Says, Wall Street Journal, June 9, 2016.

[2] Approaching the Tokenization Tipping Point, Ripple and BCG, April 7, 2025.