The Quantum Cloud Bet Is Faltering—Enterprises Want Their Own Machines

Quantum computing may finally be approaching the least glamorous maturity signal imaginable: serious buyers are asking where the machine goes, how it plugs into existing infrastructure, and whether owning one makes more sense than renting time in the cloud. In a podcast interview with Yuval Boger, Bob Sorensen, senior vice president of research and chief quantum analyst at Hyperion Research, argued that interest is shifting back toward on-premises quantum systems after years of cloud-first orthodoxy.

That is a more interesting claim than another round of qubit-count chest inflation because Sorensen comes from the high-performance computing world, not the incense-burning wing of quantum commentary. Hyperion Research tracks advanced-computing markets, and Sorensen’s argument throughout the interview is essentially that quantum is being judged less like a science project and more like an accelerator: buyers care about latency, hybrid workflow integration, economics, staffing, and whether the thing solves a real bottleneck at all. Annoyingly for several years of industry branding, that framing is probably right.

According to the podcast transcript published by Boger, Sorensen said on-prem demand is rising for a few concrete reasons. Repeated cloud access can become expensive. Hybrid classical-quantum loops do not love network latency. And some organizations want hands-on access because they are trying to build internal expertise rather than outsource the entire learning curve. Sorensen said about one-third of organizations looking at quantum today are doing so in part because they lack in-house expertise and want to develop it. That is not proof of a mass procurement wave, but it is a more tangible adoption signal than the usual ritual chants about “the utility era.”

The stronger point is not really about location. It is about what kind of machine quantum is becoming. Sorensen’s view in the 2024 interview is that quantum should sit inside a broader advanced-computing environment, closer to a specialist accelerator than a general-purpose replacement for classical systems. That lines up with a position he made explicitly in an earlier 2021 interview with Classiq, where he said the industry might have saved itself some confusion by calling these systems quantum accelerators from the start. In that earlier conversation, he also argued that many users would be satisfied with practical 10x to 50x gains on the right workloads rather than some grand metaphysical victory condition. Quantum could use more of that tone and less transcendence.

Trade coverage around the June 2024 interview mostly reinforced that same reading. The Quantum Insider and Quantum Computing Report both republished or summarized the discussion, emphasizing the same themes: on-prem systems, enterprise return on investment, Quantum AI, and lessons borrowed from the HPC market. Neither adds much original reporting, but the repetition is useful in one narrow sense. It shows which parts of Sorensen’s argument the trade press thought were worth amplifying.

That still leaves a caveat large enough to park a dilution refrigerator in. This is one analyst’s interpretation, delivered in a podcast, not a fresh Hyperion survey deck, procurement filing, or earnings disclosure. Sorensen may be accurately reading the direction of travel, but the interview does not by itself establish that enterprise quantum buying has turned decisively toward on-prem deployments. It establishes that a veteran market watcher thinks the questions are changing. In quantum, that distinction matters. The field has a talent for announcing the future several quarters before the invoices arrive.

Still, the thesis is plausible because it matches what a maturing compute market usually looks like. Once customers stop asking only whether a machine is miraculous and start asking how it integrates, who operates it, and what it costs relative to classical alternatives, the conversation has moved out of pure demo culture. If Sorensen is right, the next real quantum milestone will not be a bigger number in a lab graphic. It will be more buyers treating these systems like infrastructure: expensive, specialized, awkward to install, and potentially useful. Which, for quantum, counts as progress.