CP-67 and CMS: IBM’s Investment in Virtual Machines Before Cloud Computing
IBM's Cambridge Scientific Center funded CP-40, CP-67, and CMS as research systems, creating the virtual-machine model that later became VM/370 and eventually modern mainframe virtualization.
Virtual machines began as a research investment inside IBM rather than a mainstream product plan
In the mid-1960s IBM’s Cambridge Scientific Center explored ways to give individual users interactive computing on System/360 hardware. The project produced CP-40 and CMS as experimental systems, with CP acting as a control program that could create multiple virtual machines and CMS providing an interactive user environment.[1] The work did not begin as the centerpiece of IBM’s operating-system strategy. It was a research response to time-sharing needs and the limitations of sharing expensive computers among many users.
The economic problem was utilization and access
Mainframes were costly capital assets. Virtualization promised to let more users and operating environments share one physical system while preserving isolation.
Cambridge invested in hardware modification to prove the concept
The CP-40 project used a specially modified System/360 Model 40 with experimental dynamic-address-translation hardware. IBM’s retrospective explains that Cambridge spent 1965 and 1966 developing the virtual-machine concept and placed the modified machine into internal use in early 1967.[2] This was classic corporate R&D: researchers altered both hardware and software because the commercial product line did not yet provide the capabilities needed to test the idea.
Prototype hardware bought knowledge before a product existed
The return on the modified Model 40 was not unit sales. It was evidence that a general-purpose computer could efficiently host isolated virtual computers.
CP-67 transferred the idea onto the commercial System/360 Model 67
When the System/360 Model 67 with dynamic address translation became available, the Cambridge team reworked the system as CP-67. IBM’s historical timeline records CP-67 as a research system made available to selected time-sharing customers in 1967.[1] The move reduced the gap between laboratory prototype and customer use. A research asset was becoming a field-tested technology without yet being the corporation’s primary operating-system product.
Customer use acted as market discovery
Selected users could reveal which features mattered, how the software behaved under real workloads, and whether virtual machines solved practical organizational problems.
CMS made the virtual-machine idea useful to ordinary interactive users
Infrastructure gained value when users could work productively inside it
A hypervisor alone is infrastructure. CMS gave each user an interactive environment with files, commands, and development tools. IBM’s later history emphasizes that the combination of CP and CMS became the basis of the VM family.[3] The investment therefore joined infrastructure and user experience. Each virtual machine felt like a personal computer long before personal computers were affordable, while the expensive underlying mainframe remained shared.
Source availability created a community around IBM’s research system
Early CP/CMS software circulated to customers with source, and users contributed modifications. IBM’s historical review notes that a user community formed around the system and its source code.[4] This reduced the burden on IBM alone and accelerated learning. The research investment effectively recruited customers into product development, an early example of community-driven enterprise software evolution.
VM/370 converted the research option into a supported product
IBM formally announced VM/370 in 1972 as System/370 hardware gained virtual-memory support.[5] The years of CP-40 and CP-67 work meant IBM did not start from zero when market conditions finally aligned. The company had architecture, software, operational experience, and customer feedback ready to commercialize. This is the core financial value of sustained R&D: it creates strategic options that can be exercised quickly when technology and demand catch up.
Virtualization improved the economics of expensive computing assets
Running multiple isolated environments on one mainframe raised utilization and simplified experimentation, migration, development, and support. A customer could test a new operating system without dedicating another machine or risking the production environment. Those benefits later became fundamental to data-center economics. The underlying investment logic is the same one that drives cloud virtualization today: software can make fixed hardware capital serve more workloads with greater flexibility.
Why CP-67 and CMS belong in the history of profound software investments
CP-67 and CMS belong in investment history because IBM funded a research path whose commercial importance emerged years later. Cambridge’s experimental systems created technical knowledge and a user community before virtualization became an obvious product category.[2][4]
The long-term return is visible in VM/370, z/VM, mainframe consolidation, and the broader idea that one physical computer can safely host many logical machines. Modern cloud computing treats virtualization as ordinary infrastructure, but IBM had to finance the concept when the economic payoff was still uncertain. The investment demonstrates why research organizations matter: they let large companies create options that conventional product groups may not yet be ready to own.
Another return came from compatibility. A virtual machine could preserve an older operating environment while the physical mainframe evolved, reducing the cost of migration and testing. That option had strategic value for both IBM and customers because software investments often outlived the hardware generation on which they were first written.
Works Cited
- 01IBM — z/VM History Timeline ibm.com
- 02
- 03IBM — VM 50th Anniversary ibm.com
- 04
- 05IBM — VM/370 Announcement History vm.ibm.com
CodeHistory is a living archive. Citations document the evidence used for this edition; later evidence may refine the account.
Submit a research lead