Guest Post by Kyle Okamoto, President of Axe Compute .
When people talk about quantum readiness, the conversation often moves quickly to the quantum computer itself. How many qubits will we need? When will fault-tolerant systems arrive? Which industries will find commercial applications first?
Those questions matter, but I think there is another one worth asking: what will all of that quantum hardware actually connect to?
The architecture taking shape today suggests that quantum processors will operate alongside classical infrastructure rather than in isolation. CPUs, GPUs and eventually QPUs will handle different parts of a workload, which means the infrastructure around quantum computing could matter almost as much as the processor itself. For most enterprises, quantum readiness is therefore not a decision to buy quantum capacity today. It is the ability to evaluate, simulate, orchestrate and access specialized compute as the technology evolves.
We can already see this thinking in platforms such as NVIDIA’s CUDA-Q, which is designed to orchestrate workloads across CPUs, GPUs and QPUs. GPUs can also simulate quantum systems, giving researchers and developers a way to test algorithms before suitable quantum hardware is available.That matters because much of the work in a hybrid quantum-classical workflow will remain classical: preparing data, coordinating execution, running optimization loops and processing results.
That is where I think the quantum conversation gets particularly interesting for the AI infrastructure industry.Over the past few years, AI has exposed how difficult specialized computing becomes at scale. Deploying thousands of advanced GPUs involves far more than sourcing the chips. Power, cooling, high-speed networking, geography, latency, financing and orchestration all become part of the equation.
At Axe Compute, we have seen that demand develop quickly. In 2026, we have signed more than $3 billion in contracted value, including a five-year, $1.5 billion agreement involving more than 9,200 NVIDIA Blackwell B300 GPUs in a dedicated U.S. cluster. Our experience deploying dedicated AI infrastructure has reinforced that specialized compute is never just about the processor. At scale, the real work includes power, cooling, networking, data locality, security, orchestration and capital.
We are building that infrastructure for AI. Axe is not a quantum computing company, and a GPU cluster obviously does not become a quantum computer. I think the more useful connection is what the industry is learning in the process. Quantum systems will have materially different physical and operational requirements, but the broader operational lesson carries forward: enterprises need a practical way to integrate specialized resources without rebuilding their technology environment every time the hardware changes
We are figuring out how to deploy highly specialized compute at scale, give enterprises dedicated access to it, finance enormous infrastructure requirements and build environments that can accommodate hardware whose capabilities can change remarkably quickly.
Some of those lessons could matter when quantum processors start finding their way into commercial computing environments.
Quantum computers are highly specialized machines. Even if QPUs eventually solve commercially useful problems that are impractical for classical systems, there will still be an enormous amount of work happening around them. In practice, the challenge will not only be whether an organization can access a QPU. It will be whether it can move data efficiently, schedule workloads across different resources, integrate quantum tools into existing development environments and determine which portion of a workflow actually benefits from quantum execution.
Classical infrastructure will prepare data, coordinate workflows, run simulations, process results and handle everything that does not need a quantum processor.
That is why I think quantum readiness is partly an infrastructure question.
Companies can also start exploring this before mature quantum hardware is widely available. GPU-based quantum simulation already allows researchers to develop and test quantum algorithms using classical infrastructure. CUDA-Q, for example, lets developers work with simulated quantum systems and physical quantum processors within the same broader environment.
That progression matters. An organization does not necessarily need to own a quantum computer to start thinking seriously about how quantum computing could eventually fit into its technology stack.
The neocloud model has grown alongside AI partly because specialized computing has become extremely expensive and complicated to build internally.
At Axe, customers can specify the GPU architecture, geography and infrastructure configuration they need, while we design, deploy and operate the environment. Our more than $3 billion in signed 2026 contracts is expected to represent an annualized run rate of over $696 million once those deployments are fully online.
I think the broader principle is relevant to quantum. Enterprises will want access to the right computing resources without necessarily owning every layer of the infrastructure themselves.
Exactly how neoclouds participate in that market is still an open question. Quantum hardware has very different requirements from GPU infrastructure, and I would be skeptical of anyone claiming today to know exactly what commercial quantum infrastructure will look like a decade from now.
What does seem clear is that computing is getting more heterogeneous. CPUs now work alongside GPUs and other accelerators, and QPUs could eventually become another specialized resource within that environment.
For enterprises, I think this changes how we should talk about quantum readiness.
The question is not simply when to start buying quantum capacity. Companies should be asking whether their architecture can accommodate another major change in compute.
Can workloads move between different processor types? Can teams experiment with quantum algorithms through simulation today? Can an organization access specialized hardware without rebuilding its entire infrastructure every time computing architecture changes?
AI has already shown us how quickly these questions can go from technical considerations to business constraints. Access to GPUs, power and data center capacity became strategic issues much faster than many companies expected.
Quantum will develop on its own timeline, and serious technical challenges remain. That uncertainty is exactly why flexibility matters.
The companies best prepared for quantum may be the ones that spend the years beforehand getting comfortable with heterogeneous compute, building infrastructure that can evolve and learning how to treat specialized computing capacity as something they can access and orchestrate rather than something they have to own.
A lot of that work is already happening. Today, we happen to be doing it for AI.