[Blog] Closing the Secure Control Gap in AI Datacenters with FPGAs

The global datacenter market is expanding at an incredibly rapid pace, projected to surpass over $900 billion USD by 2033. This unprecedented growth is driven by demand for more artificial intelligence (AI) computing capacity, which in turn requires more powerful and complex servers.

These servers have historically operated in streamlined, cohesive, tightly integrated architectures. However, increased demand has given way to more distributed, heterogeneous datacenter infrastructure that’s harder for operators to manage and control.

In a recent webinar hosted by Embedded Computing Design, Lattice explored this growing complexity, the expanding risk landscape threatening AI server networks, and how field programmable gate arrays (FPGAs) can be used to provide a foundation for control and trust that datacenters need to innovate safely.

Datacenter Disaggregation Drives Complexity and Risk
The shift from cohesive server systems to disaggregated AI infrastructure is evident in different ways throughout the modern datacenter. On the one hand, it involves integrating components from multiple vendors in the same server racks, including different CPUs, GPUs, NICs, DPUs, accelerators, and more. It also includes supporting hybrid models that leverage both cloud and on-premises infrastructure, fragmenting management layers across system hardware and software.

The results of this disaggregation can be boiled down to three core operational pain points:

1. The explosion of complexity. The proliferation of hardware components introduces new firmware stacks, validation requirements, and failure modes with each added vendor or subsystem.
2. Eroding trust in system data. With so many disparate data sources in the system, it becomes increasingly difficult to verify whether reported metrics (temperature, performance, etc.) are accurate or have been compromised.
3. Increasingly high-risk operations. Previously routine actions like hardware provisioning or configuration now carry a significant downside, as a single misstep or failure can disrupt and damage high-value AI infrastructure.

These pain points combine to create an environment with fragmented visibility, uncertain dependencies, more complex management tasks, and a growing inability to exert consistent, deterministic control.

Security Challenges Create Added Pressure
Compounding these operational challenges is a rapidly intensifying security mandate that many organizations have yet to fully grasp. While traditional cryptographic models may still be mathematically sound today, that does not mean they’ll be able to defend against the threats of quantum computing. The long lifecycle of AI infrastructure introduces a new dimension of risk to server systems, as the sensitive information and system credentials transmitted today may still hold value years from now.

This provides ample motive for “harvest now, decrypt later” attacks, in which hackers capture encrypted data today with the intent of decrypting it in the future with the support of quantum capabilities. In response, datacenter and telecommunications industry bodies and governments have already begun formalizing the transition to post-quantum cryptography (PQC) through various standards and mandates, including multiple NIST PQC encryption standards, the United States’ CNSA 2.0, the DTMF’s SPDM version 1.4, and most recently, Executive Order 14409, which mandates ML-KEM key establishment for U.S. Federal High Value Assets by December 31, 2030.

For datacenter operators, these PQC security concerns and related standards and regulations add extra layers of complexity on top of already fragmented server environments. Not only do systems need to unify control, but they must do so in a manner that’s capable of adapting to evolving cryptographic standards.

How FPGAs Unify Control, Trust, and Security
Addressing these control and security issues requires a reimagined approach to system architecture. FPGAs are emerging as hardware that can support both real-time control and long-term security flexibility, all without taking over system operations from existing compute and networking components. These act as companion chips operating alongside the main compute engines in servers, taking on the control challenges, acting as an anchor of system-wide trust, and enabling existing hardware to operate at full capacity.

From an operational perspective, FPGAs help restore deterministic control in disaggregated server environments. Their hardware-based architecture enables:

• Parallel telemetry, monitoring input from hundreds of sensors simultaneously.
• Real-time processing so systems can analyze and act on data in a timely manner at the edge.
• Deterministic response with predictable, microsecond, bounded-latency control loops.
• Sideband operation to exhibit control outside the host/data path and ensure critical actions are not delayed by system workload activity.
• Separation between hardware control planes and software-based management.

At the same time, FPGAs provide the flexibility needed to meet evolving security demands like the transition to PQC. Core security capabilities include:

• Establishing a hardware root of trust (HRoT) that acts as a verifiable foundation for secure boot, firmware validation, and overall system integrity.
• Trusted telemetry and attestation to validate device identity across distributed components.
• Crypto agility, or the ability to update cryptographic algorithms in the field as standards evolve without replacing hardware.
• Support for hybrid cryptography, including both classical algorithms and newer PQC models.
• Lifecycle security management, with support for secure firmware updates, authentication, and key management.

Achieving Datacenter Resilience
By combining deterministic control with flexible, hardware-based security, FPGAs provide a unified solution for both the immediate challenges of datacenter server disaggregation and the longer-term requirements of post-quantum resilience.

To learn more about building a resilient, adaptable, hardware-based foundation for disaggregated datacenter servers, watch the full webinar here. To dive deeper into the role of FPGAs as key control and security companions in modern infrastructure, contact our team today.