Speaking with TFiR, Danielle Cook of Akamai defined the current state of AI infrastructure and the three architectural shifts enterprise leaders must prepare for in 2026.

WHAT IS THE RELATIONSHIP BETWEEN AI EXPERIENCE QUALITY AND INFRASTRUCTURE IN 2026?

Cook’s first prediction closes the gap between infrastructure teams and product teams. “Speed and responsiveness — they’re not just performance metrics anymore. They’re defining what good feels like from a product point of view.” This represents a fundamental repositioning: infrastructure decisions are no longer backend concerns — they are direct drivers of product quality and customer satisfaction. As AI raises the baseline for responsiveness, organizations that fail to make the right infrastructure choices will feel it immediately in customer experience.

AI Inference Placement as a Primary Design Choice

Cook’s second prediction reframes where AI runs as a strategic, product-level decision. The model is moving from centralized compute in a few data center regions to distributed execution — closer to users and data — by default. This is not an optimization pass applied after a system is designed. It is the architecture. Latency introduced by round-trips to distant data centers becomes a product defect in an era where users expect instant, intelligent responses. “We’re making the execution distributed and much closer to the users and data, and we’re doing that by default.”

Kubernetes Achieving Product-Market Fit Through AI

Cook’s third prediction is grounded in data. Citing the recent CNCF annual survey, she argues that Kubernetes is settling into its role as the standard portable runtime for serving AI models wherever distribution is needed. “Kubernetes is where AI is running and AI inference is becoming its defining workload.” After years of broad but diffuse adoption, AI is giving Kubernetes a clear, singular identity — the portable, standard layer for running inference anywhere.

Broader Context: Challenges, Opportunities, and Akamai’s 2026 Strategy

Cook’s full TFiR interview reveals the operational complexity confronting organizations as they navigate this infrastructure shift. Centralized architectures will struggle to support real-time AI interactions at scale — echoing the early web’s latency constraints, where distance and round-trip time showed up immediately in user experience. Operationally, Cook explains that AI collapses the entire day-two stack — security, observability, GPU management, and data pipelines — into a single, unified problem. Teams cannot address each challenge in a silo; they must solve it as one.

Agentic systems introduce a further architectural challenge: continuous, machine-driven interaction loops that break traditional request-response traffic models. The patterns that worked when a human was waiting at a keyboard do not map to a world where AI agents are generating ongoing, high-frequency interactions.

On the opportunity side, Cook identifies the largest prize at the platform layer: making Kubernetes effectively invisible through opinionated platforms, internal developer platforms (IDPs), and golden paths that remove operational burden without sacrificing portability. “Your teams can just be deploying AI anywhere.” Akamai’s own execution against this vision centers on its managed Kubernetes engine (LKE), GPU-backed infrastructure, and distributed edge — designed to let developers run any AI model wherever their users are, without friction. “We want to make AI inference workloads great. We want our customers’ experiences to be amazing.”

Watch the full TFiR interview with Danielle Cook here.