Guests: Ari Weil | Matt Butcher
Companies: Akamai | Fermyon
Show Name: An Eye on AI
Topic: Edge Computing, Kubernetes

Traditional serverless and container platforms weren’t designed for the demands of edge computing, especially when AI workloads enter the picture. Matt Butcher, now VP of Product at Akamai and former CEO of Fermyon, explains why legacy architectures hit their limits at the edge and how WebAssembly is unlocking capabilities that weren’t possible before.

The Edge Exposes Serverless Limitations

Serverless computing revolutionized how developers build applications by abstracting away infrastructure management. But as workloads move closer to end users, the model breaks down. Cold start times, resource constraints, and geographic distribution challenges create friction that traditional platforms can’t solve efficiently.

Matt Butcher identifies the core problem: “There are limits of traditional serverless or container platforms, especially at the edge.” When you’re trying to deliver real-time AI experiences or latency-sensitive applications across thousands of global locations, the old patterns simply don’t scale.

Deploy Once, Run Everywhere: 4,000+ Locations

Akamai’s acquisition of Fermyon brings WebAssembly-based serverless to a global edge network spanning more than 4,000 locations. Butcher highlights the immediate unlock: “This enables a developer to deploy once and have an application distributed globally. That’s one of those exciting things that unlocks right away.”

The managed container services platform allows applications to run from core regions all the way out to edge points of presence. Developers no longer need to manually orchestrate multi-region deployments or worry about geographic replication. The platform handles distribution automatically while maintaining performance.

AI Inferencing Moves Beyond Chat

Large language models captured the industry’s imagination, but Butcher sees a bigger opportunity. “There’s always been sort of some ambiguity about how developers were going to engage with AI, particularly on the inferencing side, to be able to build real world applications and to get past that early paradigm where LLMs are used for chat.”

Running AI inference at the edge—optimized for sub-millisecond performance—unlocks entirely new application categories. Imagine personalization engines that respond instantly, fraud detection systems operating in real-time, or conversational interfaces that don’t require round-trips to centralized data centers. These use cases demand compute close to users, and WebAssembly makes it practical.

WebAssembly Enables Intelligent Code Placement

The longer-term vision is even more transformative. WebAssembly’s portability and component model allow applications to intelligently locate code based on where it makes the most sense. “How do we get this piece of code to execute very close to the database, where there’s all this data, and this other piece of the application to execute as close to the user as possible?” Butcher asks.

This represents a fundamental shift in how developers architect applications. Instead of monolithic deployments or manually sharded microservices, applications can dynamically optimize themselves—placing data-intensive operations near storage and user-facing operations near end users. WebAssembly’s unique component model makes this level of granularity possible without sacrificing security or performance.

A New Class of Edge-Native Applications

Butcher’s conclusion points to what’s coming: “Once we can run all of this inside of the Akamai network, and again, take advantage of this wealth of services that are already there, I think we’re going to open up the opportunity for developers to actually write a new class of edge-native applications.”

These aren’t just faster versions of existing apps. They’re fundamentally different architectures that combine global distribution, AI capabilities, and intelligent workload placement. For enterprises exploring real-time personalization, AI-powered automation, or next-generation user experiences, this represents a significant capability expansion.