CoreWeave launches Forge to unify AI development lifecycle

CoreWeave has announced the launch of CoreWeave Forge, a development layer designed to unify the AI lifecycle for teams building models and agents. The platform aims to reduce fragmentation by connecting training, observation, curation, and evaluation within a single environment. It integrates services from Weights & Biases Models, OpenPipe, and the open-source marimo notebook project. Companies including Canva and MasterClass are already utilizing the tool.

This launch coincides with a broader industry shift toward inference-driven workloads. Analysts note that while AI infrastructure was previously focused on training, the demand for inference is growing rapidly, with projections suggesting a shift in workload distribution from a 50/50 split to 10/90 by next year. This transition is driving a move toward system-level optimization, where efficiency is measured by the ability to produce as many tokens per watt as possible across silicon, networking, storage, and software.