智能体AI推动企业转向自有本地算力
SiliconANGLE AI报道称,智能体AI工作负载正促使部分企业选择拥有本地算力,而非完全租用;主要考虑包括成本、安全、延迟和数据控制。Oxide Computer Co.联合创始人兼CTO Bryan Cantrill称,开放权重模型和CPU密集型工作负载正推动部署靠近企业数据中心,而Upbound的Modelplane则采用跨云推理路线。报道未量化市场采用情况,主要依据来自Oxide高管访谈,缺少独立市场数据。
模型部署位置正成为智能体AI基础设施和数据治理的重要决策,各厂商也在推动不同的推理路由与工作负载隔离方案。
Agentic AI is reviving an argument over renting versus owning compute, and the model plane — the layer that decides where AI models run — sits at the center of it. Open-source efforts such as Upbound Inc.’s Modelplane route inference across clouds, but many enterprises want the models on hardware they control.
That shift favors companies that built early for on-premises cloud infrastructure. Oxide Computer Co., which raised $445 million in October to expand rack production, built its business on the idea that enterprises at scale will want to own their computers, according to Bryan Cantrill (pictured), co-founder and chief technology officer at Oxide.
“Cloud computing is great, but the model shouldn’t be exclusively to rent it. You get to a certain scale, as it turns out, you want to own your own computer,” Cantrill said.
Cantrill spoke with theCUBE Research’s John Furrier at SC26, during an exclusive broadcast on theCUBE, SiliconANGLE Media’s livestreaming studio. They discussed renting versus owning compute, why agentic workloads are moving on-prem and multi-tenancy for isolating those workloads.
Agentic workloads move on-prem
Agentic workloads are driving an explosion in compute demand, and much of it is CPU-heavy work that enterprises would rather own than rent. Economics remains the main reason to bring it in-house, but where models live is a close second, Cantrill explained. That makes the model plane as much a data governance decision as an infrastructure one.