A flurry of announcements from hardware vendors points to a change in how enterprises are purchasing servers, storage and networking resources for their data centers and edge deployments.\nTo entice companies to keep workloads on premises, hardware vendors including Cisco, Dell, HPE, IBM, Lenovo and others are offering consumption-based pricing for data-center infrastructure. These pay-per-use products are designed to shorten procurement cycles, allow customers to scale up or down with demand, and more economically link hardware spending with usage.\nHPE, for example, pledged to transform its entire portfolio to pay-per-use and as-a-service offerings by 2022, and last week, the company added to its GreenLake lineup with new data services and infrastructure. Dell, for its part, unveiled the first products in its Apex portfolio of managed storage, servers, and hyperconverged infrastructure.\nPay-per-use hardware models such as these can deliver cloud-like pricing structures and flexible capacity to on-prem environments, and the hardware can be deployed in a company\u2019s own data center, at edge locations or in colocation facilities. The model is generally targeted at enterprises that want alternatives to buying equipment outright for workloads that aren\u2019t a fit for the public cloud.\nThis approach can help enterprises simplify modernization efforts and preserve capital, says Daniel Newman, principal analyst at Futurum Research.\n\u201cWith hybrid cloud being the overwhelming approach of the enterprise, it is becoming a catalyst for big IT infrastructure providers to build cloud-like models for prem-based services. It makes sense,\u201d Newman says. \u201cHPE and Lenovo took this path early on with GreenLake and TruScale. Dell has been building Apex for some time and has a massive customer base in key areas like storage and compute to tap into. Cisco is also entering the fray with its Cisco Plus offering.\u201d\nThe concept of pay-per-use hardware isn\u2019t new, but interest is growing. Research firm IDC reports that 61% of enterprises plan to aggressively shift toward paying for infrastructure on a consumption basis. By 2024, half of data-center infrastructure will be consumed as a service, IDC predicts.\nUptake so far is strongest in storage. In 2024, half of newly deployed storage capacity will be consumed as a service, Gartner anticipates. On the server side, 5.6% of on-premises x86 server spending will be consumed as a service in 2024, Gartner predicts.\nWhy as-a-service infrastructure?\nPart of the appeal of as-a-service infrastructure is that companies are looking for more manageable hardware refreshes, says Tracy Woo, senior analyst at Forrester Research.\n\u201cHardware vendors that traditionally sold a system as a one-off are finding it hard to compete with the cloud providers. Customers don\u2019t want a data center that is difficult to refresh. Public cloud offers the ability to get the latest hardware, with much less upstart costs, and not having to worry about capex budgeting. To stay relevant in the conversation, hardware providers like HPE and Dell need a new way to provide their hardware\u2014hence, GreenLake and Apex,\u201d Woo says.\nAnd there\u2019s plenty of capacity that isn\u2019t destined for the public cloud any time soon, analysts say.\nPublic cloud has had an enormous impact on enterprise computing, but it still only represents a fraction of total workloads, says Bob O\u2019Donnell, president and chief analyst at TECHnalysis Research. The majority of workloads are still running locally in a data center or in a colocation facility. At the same time, enterprises have come to appreciate many aspects of the public-cloud model. \u201cThe flexibility of being able to ramp up as you need capacity, and then ramp down when you don\u2019t, for certain businesses is fantastic,\u201d O\u2019Donnell says. \u201cThat business model is appealing in a lot of ways.\u201d\nA consumption-based hardware approach can offer the best of both worlds. It combines elements of the public cloud-computing model with the realities of workloads that need to stay local, whether for regulatory reasons or because of the effort and skills required to rearchitect applications for the cloud, for example, O\u2019Donnell says.\nThe as-a-service approach provides more agility to turn things on or off, or make a left-hand turn, while also allowing enterprises to maintain control over their data and use on-prem infrastructure for their latency-prone workloads, says Patrick Moorhead, president and principal analyst at Moor Insights & Strategy.\n\u201cI think the best workloads to consider are those already in the public cloud that don\u2019t require a tremendous amount of \u2018burst\u2019 capabilities. Anything that has a lot of data would benefit from going on-prem as this is where the data is generated,\u201d Moorhead says.\nWoo agrees. \u201cGood fits are items that are very data-heavy, where the cost of ingress and egress will ramp up quickly and where latency and fast processing are an issue. These are workloads in AI and ML or in high-performance computing,\u201d Woo says. \u201cOther areas these are useful are in remote locations where public cloud coverage isn\u2019t possible, like on an oil rig in the middle of the ocean. For these scenarios, companies will look at these sorts of on-premises solutions.\u201d\nAnalysts caution, however, that making a switch to this model won\u2019t be without challenges, and estimating usage requirements can be tricky. \u201cThis represents a fundamental shift in how IT gets paid for,\u201d O\u2019Donnell says. \u201cThe devil\u2019s in the details. Some of [the services] are truly consumption based, and some are actually more subscription based, which is slightly but importantly, different.\u201d\nFor these reasons, companies won\u2019t rush into the consumption-based model, and there will be a lot of experimentation, O\u2019Donnell says. \u201cLet\u2019s not forget, there have been plenty of horror stories of people who moved to the public cloud, thinking it was going to be great because they only paid for what they needed, and it ended up costing them way more than they expected.\u201d\nOn the positive side, \u201cWe are watching public cloud and on-prem converge quickly to serve the needs of the user \u2013 which is an encouraging sign,\u201d Newman says. \u201cI love seeing competition among the providers as it tends to foster innovation and make offerings more competitive."\nNew moves in consumption-based pricing\u00a0\nHere\u2019s a summary of the latest pay-per-use servers, storage and networking announcements:\nDell Apex\nDell unveiled the first products in its Apex portfolio, which include a storage-as-a-service offering and a bundle of on-premises compute, storage, and networking resources for private and hybrid cloud deployments. A self-service dashboard called Apex Console lets enterprises identify, subscribe to, and monitor the health and performance of Apex services; access usage and spending reports; and refine capacity as business needs change. Read more: Dell delivers lineup of on-prem, pay-per-use hardware\nHPE GreenLake\nHPE last week added to its GreenLake portfolio of on-premises, pay-per-use services. Its new data-services platform consists of three new elements: Data Services Cloud Console, a cloud console that\u2019s designed to streamline storage management, from deployment and provisioning to ongoing maintenance; cloud data services, a suite of software subscription services that automate infrastructure management; and HPE Alletra, a new infrastructure portfolio that\u2019s managed by the Data Services Cloud Console and features the All-NVMe HPE Alletra 9000 and 6000 systems. Read more: HPE kicks off software-defined storage-as-a-service\nCisco Plus\nCisco in March advanced its network-as-a-service plans with the debut of Cisco Plus and a commitment to deliver the majority of its portfolio as-a-service over time. The first available solution will be Cisco Plus Hybrid Cloud, which includes the company\u2019s data-center compute, networking, and storage portfolio in addition to third-party software and storage components all controlled by Cisco\u2019s Intersight cloud-management package. The second Cisco Plus service will feature Cisco\u2019s secure access service edge (SASE) components, including its SD-WAN and cloud-based Umbrella security software. Read more: Cisco takes its first steps toward network-as-a-service\nIBM Tailored Fit Pricing for IBM Z\nTwo years ago, IBM announced a cloud-like pricing model for its mainframe software, a move designed to deliver more flexible, consumption-based pricing for workloads on IBM z\/OS. This month, it added a complementary flexible pricing model for its mainframe hardware: Tailored Fit Pricing for IBM Z-Hardware Consumption Solution. Read more: IBM moves toward consumption-based mainframe pricing\nLenovo TruScale\nLenovo in April announced a collaboration with Nutanix to deliver a hosted desktop as-a-service offering. Enterprises can choose from a range of Lenovo client devices, such as thin clients and PCs, along with Citrix and other virtual desktop environments, and ThinkAgile HX Series hyperconverged infrastructure powered by Nutanix. Monthly pricing covers the client devices, data-center systems, and software and services, all managed by Lenovo.