Akamai introduced two new products for its cloud platform on Tuesday, October 27, 2026, that together turn Akamai Cloud into a full stack for running artificial intelligence in production: Akamai Inference, a managed service for serving open-weight and custom models, and Akamai Agent Sandbox, an isolated execution environment for the code AI agents write and run. Both products are available in beta, and Akamai says they sit on infrastructure that has already won multiyear cloud contracts collectively worth more than $14 billion.
Managed Inference Without Running The Stack
Akamai Inference gives enterprises managed access to open-weight and custom models through OpenAI-compatible APIs, so existing applications can move over without rewrites. Customers can choose serverless, pay-per-use inference or dedicated deployments, depending on how predictable their traffic is and how much control they want over the hardware underneath.
At launch, the catalog includes 12 open-weight models ranging from 8 billion to 2.8 trillion parameters, including NVIDIA Nemotron 3 Ultra, OpenAI gpt-oss-120b, and Google Gemma 4, along with embedding models for retrieval-augmented generation. That technique lets a model pull in a company’s own documents before answering, which matters more than raw model size for most enterprise deployments.
Akamai Cloud supplies managed databases and object storage, and connects to private and public cloud platforms, so models can draw on enterprise data. Akamai Inference then handles orchestration, model execution, and authentication, while requests run across the company’s global network, which Akamai says reduces latency and keeps performance consistent. The pitch is aimed squarely at teams that want to ship an AI feature this quarter without hiring a team to operate model-serving infrastructure.
A Boundary Around Agent-Written Code
The second product addresses the riskier half of the agentic shift. AI agents increasingly do real work, from taking actions on a user’s behalf to writing and running their own code, and that code is exactly what enterprises are least comfortable letting loose inside their systems.
Akamai Agent Sandbox runs agents on Akamai Functions, while the code they write runs in its own isolated environment, sequestered from the agent’s runtime. Delivered as an open source component of the Spin framework, it uses WebAssembly isolation to set clear limits on compute, memory, outbound connections, and system capabilities, so a runaway or compromised agent script cannot simply reach whatever it wants.
Akamai is not betting on a single isolation model. The company says developers can choose the architecture that fits their workload, from microVMs with snapshot and fork capability to policy-enforced containers to WebAssembly-isolated functions, with NVIDIA-accelerated and dedicated CPU infrastructure, managed Kubernetes, and serverless execution across edge and core locations. It also plans to add capabilities that govern what agents can reach and what can reach them, which is the control plane enterprises will actually audit.
More GPU Muscle Underneath
Alongside the two products, Akamai expanded its dedicated GPU fleet to include NVIDIA Blackwell Ultra clusters, and said it is preparing its infrastructure to support the NVIDIA Vera Rubin platform next year. The fleet also includes NVIDIA RTX PRO 6000 Blackwell Server Edition GPUs for low-latency inference.
Each eight-GPU Blackwell Ultra node carries 288 GB of HBM3e memory per GPU, 2.3 TB in total, enough to hold frontier-class models. Akamai says the architecture targets the workloads driving that memory demand, including test-time scaling, chain-of-thought reasoning, and long-context agentic work, and that GPU clusters run without rate limits or shared-tenancy queueing. That control over model versions and deprecation timelines is a real selling point for enterprises tired of a provider swapping a model underneath them.
Adam Karon, chief operating officer and general manager of Akamai Cloud, said customers had pulled the company further than its original thesis, which was that inference would represent the majority of enterprise AI focus and investment. Jon Alexander, senior vice president of cloud technology, added that every team running agents needs a boundary around the code those agents write, and that Akamai Cloud now supports managed, dedicated, and edge deployment styles on one platform.
Why It Matters
This matters because the bottleneck in enterprise AI has shifted from getting a model to getting it into production safely. Inference is where the money goes once the demos end, and agent code execution is where the risk concentrates once agents start touching real systems. Akamai is packaging both problems into one platform, with a global network of more than 4,350 edge locations as its distribution advantage over smaller neoclouds. The open question is whether beta products built on a content-delivery heritage can win trust against hyperscalers and specialist inference providers, and the same identity question is now facing every agent platform, while enterprises are wiring agents directly into their workflows and cloud providers are opening up to rival chips to meet demand.
Background
Akamai, long known for content delivery and security, has been building Akamai Cloud into a general-purpose cloud platform, and the company says its AI infrastructure strategy rests on the idea that inference, not training, will dominate enterprise spending. Both new products are in beta, and Akamai has not disclosed pricing in its announcement.
Source: Akamai launch announcement
0 Comments