Fluidstack removes egress and ingress fees

Zero transfer fees are a wedge against the biggest hidden tax in AI infrastructure, which is moving massive datasets and model checkpoints in and out of a cluster. For a frontier training run, teams can push in training data, pull out checkpoints, copy results to another cloud, and recover from failed runs without watching a separate networking bill pile up. That makes the real cost of a Fluidstack cluster easier to predict, and makes switching onto the platform less risky.1 2 3 4

• This matters most for customers training large models, because the data exhaust is huge. A single project can repeatedly move terabytes or petabytes of embeddings, checkpoints, logs, and evaluation outputs. When transfer is free, the customer mostly pays for GPUs and storage, not for every movement around the workflow.1 2 4
• The policy also narrows one of the old hyperscaler advantages. Independent GPU clouds like CoreWeave and Lambda now market the same no ingress and egress message, which shows that transparent networking bills have become table stakes for winning serious AI training workloads.2 3 4
• For Fluidstack specifically, the pricing choice fits its broader sell. The company is renting bare metal GPU capacity with single tenant clusters, cloud like orchestration, and fast support, so removing transfer fees reinforces the pitch that customers are buying predictable dedicated compute, not a maze of metered cloud line items.1 5

The next step is for zero fee data movement to become standard across serious GPU clouds, while competition shifts to who can offer the best all in economics on reserved capacity, storage, and utilization. In that market, the providers that win will be the ones that make giant training jobs feel financially boring as well as technically reliable.2 3 4 5