Let’s Talk About What’s Next
Contact us
Connect with Canopy
Whether you’re exploring compute solutions, potential partnerships, or simply want to learn more about Canopy and what we’re building, we’d love to hear from you.
Reach out to our team and let’s start a conversation about how we can work together.
FAQ
The Questions Behind Every Compute Deal
What is GPU Compute Procurement?
GPU compute procurement is the process of sourcing, comparing and contracting GPU and cloud capacity for training and inference workloads. It covers working out what the workload needs, benchmarking providers, modelling total cost over the term and negotiating price, commitments and exit terms. Canopy runs this process on the buyer's side across hyperscalers, neoclouds and specialist GPU providers.
What is a Neocloud?
A neocloud is a cloud provider built around GPU compute for AI rather than general-purpose infrastructure. CoreWeave, Lambda, Nebius, Vultr and Civo are examples. Neoclouds often price below hyperscalers and offer newer chip generations sooner, but they vary widely in scale, financial stability and support, so due diligence matters more than it does with AWS, Azure or Google Cloud.
Why Do GPU Prices Keep Changing?
GPU prices move because demand for AI compute is running ahead of supply and capacity is often sold before it's built. Quotes are typically held for 30 days or less, new chip generations reset pricing every two to three years and hundreds of providers have entered the market with different cost bases. That's why a price that's good today usually needs to be locked or re-quoted within weeks.
Should We Use a Hyperscaler or a Neocloud for AI Workloads?
It depends on the workload, the term and the risk you can carry. Hyperscalers offer breadth, compliance and integration with services you already use. Neoclouds tend to offer lower GPU-hour rates and faster access to new generations. Many teams run both: a hyperscaler for the platform and a neocloud for training or inference capacity. An independent comparison on your own workload is the reliable way to decide.
What is a Take-or-pay Compute Contract?
A take-or-pay contract commits you to paying for an agreed amount of compute over a fixed term whether you use it or not. It's the standard structure behind reserved and forward GPU deals and the reason committed rates are lower than on-demand. The terms to negotiate are the ramp, the right to shift or resell capacity and the exit clauses.
What is Forward Capacity in AI Compute?
Forward capacity is GPU compute reserved for delivery at a future date, often before the cluster is built. Buyers use it to lock a price and a delivery slot for next-generation chips, and suppliers use the commitment to finance the build. It's the second phase of a two-phase plan: bridge on available capacity now, roll into reserved next-gen capacity when it comes online.
How Much Can You Save on GPU Compute?
Savings depend on where you start and how you buy. Canopy typically cuts cloud and compute spend by 10 to 30% against the incumbent quote, and buyers who run a competitive process have secured GPU capacity at around 40% below prevailing market rates, as Geodesic Research did. The biggest savings come from timing, commitment structure and competition rather than discounts on list price.
What Does Canopy Do?
Canopy helps AI teams source, compare and secure GPU compute.