Chris Sharp, Chief Technology Officer, Digital Realty

Building on our prior blog, AI Innovation Won’t Slow Down: How Multi-Model AI is Powering Speed to Market, this article continues our thought leadership series on AI.

Digital Realty is closely tracking AI evolution across hardware, software, and networking to provide enterprises with meaningful insights that help scale private AI efficiently. As AI adoption grows, enterprises must focus on security, optimization, and interconnection to integrate private data with public AI models successfully.

This blog will cover three key areas enterprises must address to enable private AI at scale:

Enabling private data for public AI models: How enterprises can efficiently integrate private datasets with public AI models.
Securing private data in AI workflows: Key security frameworks and compliance considerations for protecting enterprise data.
Optimizing AI performance with data centers and interconnection fabrics: Why workload placement, data centers, and interconnection fabrics are critical to AI success.

Enabling private data for public AI models

Integrating private datasets with public AI models requires a balance between accessibility and security. Enterprises should ensure structured data access and controlled AI model interactions to maximize performance while maintaining privacy. Key steps include:

Identity and authentication controls to regulate access to sensitive data
API-driven pipelines that securely connect internal databases with public AI models
Retrieval-Augmented Generation (RAG) to dynamically retrieve relevant private data without exposing it to public AI training

A hybrid approach utilizing frontier models like GPT, Claude, and Gemini, along with industry-specific AI, enables wide-ranging AI reasoning capabilities from generalized models and enhanced accuracy from custom-trained industry AI solutions. This method improves performance through strategic workload distribution across cloud, on-premises, and edge environments.

Example: Industry-Specific AI in Financial Services

Large financial institutions, such as JPMorgan Chase, leverage private AI models to optimize risk assessment, fraud detection, and customer service. By integrating proprietary financial datasets with AI, these firms enhance decision-making while ensuring compliance with strict regulatory requirements. Similar approaches can be adopted across other industries, such as healthcare and manufacturing, to drive specialized AI applications that go beyond the capabilities of general-purpose models.

Securing private data in AI workflows

Keeping private data secure in AI workflows requires a zero-trust model, which assumes no user or system should be inherently trusted, regardless of their location or credentials. Enterprises must adopt a verify-everything approach to prevent unauthorized access and data breaches. By embedding zero-trust principles such as end-to-end encryption into AI workflows, enterprises can safeguard sensitive data, reduce the risk of exposure, and maintain compliance while scaling AI adoption.

Enterprises must implement AI safety guardrails and model lifecycle security to ensure models operate as intended and remain secure from development through deployment and ongoing operation. Key components include:

Input validation – Preventing AI models from processing malicious or biased data inputs that could lead to incorrect or unethical decisions
Output monitoring – Tracking AI responses to detect anomalies, biases, or potential misuse
Secure training environments – Protecting model training data from poisoning attacks that could introduce hidden biases or vulnerabilities

Optimizing AI traffic is just as critical—AI service fabrics help manage inference requests efficiently, enforcing security without performance trade-offs. Following standards like Mitre Atlas and NIST AI Risk Management ensures strong defenses, especially as agentic AI workflows dynamically split tasks across multiple models.

Optimizing AI performance with data centers and interconnection fabrics

Getting AI to perform at its best isn’t just about having powerful models—it’s about placing workloads in the right environment, ensuring fast connections, and keeping costs under control. Enterprises need a mix of cloud, edge, and private data center deployments to strike the right balance between performance, security, and efficiency.

Strategic workload placement enhances AI performance while managing costs:

AI Workload Type	Optimal Deployment	Key Benefit
Model training	Cloud AI clusters	Scalable compute power
Batch inferencing	Private data centers	Security and cost-efficiency
Real-time inferencing	Edge AI deployments	Low-latency response
RAG query processing	AI exchanges	Secure public-private data access

AI models don’t operate in isolation—they rely on vast amounts of data moving seamlessly between locations. High-speed, low-latency connectivity is essential to keep AI responsive and efficient

As AI adoption accelerates, inference workloads require low-latency, high-bandwidth environments near dense metropolitan areas. Digital Realty’s core market infrastructure bridges cloud, edge, and enterprise, ensuring AI models run efficiently in proximity to data sources and users.

The future of AI is interconnected

The next wave of AI innovation isn’t just about bigger models—it’s about smarter infrastructure. Enterprises must focus on seamless data integration, security, and optimized performance to fully unlock AI’s potential.

At Digital Realty, we’re not just supporting AI adoption—we’re shaping the future of AI-powered business by delivering secure, high-performance, and interconnected infrastructure. Whether you’re scaling private AI, securing your data, or optimizing your workloads, we have the solutions to help.

Engage with Digital Realty today to get started on building your AI-ready infrastructure.

Data Center Frontier Special Report: Boston Data Center Market