To maximize sales server vendors want to make their products available to customers in multiple ways. In terms of the way they are acquired, this means outright purchase, leases and pay-per-use schema. In terms of what you buy, this means Platform as a Service (PaaS) and Infrastructure as a Service (IaaS) offerings in addition to physical servers themselves. IBM is later than its competitors in deciding to make its Power servers available as an on-premise cloud offerings – akin to HPE’s Greenlake, in the market since 2018 , AWS Outposts, which were launched at the end of 2019 and Oracle Compute Cloud@Customer launched in August 2023. In this post I’ll look at what IBM’s move and its consequences.
IBM Power servers run AIX, Linux and IBM i operating systems. I show the versions of the server currently available by location in my Figure above. Up until its latest announcement IBM had already made Power Virtual Servers available from the IBM Cloud. Partners have created their own private clouds, such as CSI and Racksquared. At the beginning of 2020 Google Cloud Services partnered with Converge to launch its new Power as a Service offering, while Microsoft Azure’s users can run Power workloads via Kyndryl’s Skytap and AWS’s can integrate with Power servers running on Connectra’s private cloud. I tend to think of these offerings tied to public clouds as hybrid solutions rather than vanilla IaaS ones.
IBM has done a lot over the years to make its Power chips and systems open and easy to adopt; forming the OpenPower Foundation in 2013, enhancing its overall status as an open source software vendor through the acquisition of Red Hat in 2018 and launching PowerVS in 2019. According to IBM PowerVS is currently available from 21 public datacenters across the world, has around 650 customers who have deployed it typically for production, High Availability/Disaster Recovery (HA/DR) and test/development workloads. It also reports that it has provided eight quarters of double-digit revenue to the company.

What has IBM announced?

IBM is currently offering 2 different pods available as a small (1 rack) or medium (2-4 rack) solution (see my Figure above for details). In particular:

• The small systems are based on either the Power S1022 or Power E1050 servers and its FS5200 storage array.
• The medium systems are based on the Power S1022, E1050 and E1080 servers and FS9500 storage array.

Each pod also includes appropriate network and IBM advanced or premium support for IBM i, AIX and Linux workloads. The machines are fully owned by IBM and do not include memory swaps within a server or upgrades across pod sizes. IBM will not allow customer equipment to be installed within the pod or the addition of third party components without its consent. Keeping the solution fully under its own control allows IBM to offer a Service Level Agreement (SLA) of 99.995% for the pods.
Once ordered IBM will activate the pod in 30 days and fully meter its usage. The contract will last for either 3 or 5 years, with a 1-year renewal. Customers won’t need to pay an up-front cost, but they will need to commit to a monthly spend, which they can go over, but not under. This flexible Committed Monthly Spend (CMS) approach avoids the high up-front hardware cost and the need to over configure a system in order to accommodate future maximum usage. If the customer needs it during the contract period, IBM will add compute/storage capacity or deploy more pods. IBM has designed the pricing so that the cumulative CMS will always be less than the purchase price of an equivalent system. IBM is offering its customers the ability to try PowerVS for free through providing a $2k cloud credit and, along with some of its business partners, to pilot their workloads for free and/or migration funding to get up and running quickly.

Typical use cases

There are a number of use cases IBM envisages at launch. In particular:

• Digital transformation and application modernization through the adoption of Red Hat OpenShift. Customers and ISVs can install OpenShift in just 2 hours.
• Infuse business applications on IBM Power servers with AI. The pods are designed for rapid integration with IBM’s watsonx SaaS, allowing them to build chatbots and other AI applications more easily.
• Building Disaster Recovery (DR) solutions through installing 2 pods in different locations synchronized through Global Replication Services, storage or hardware replication. IBM claims that a DR solution can be set up and running in hours, rather than weeks with a Recovery Point Objective (RPO) of around 10 minutes, as long as the customer’s network meets the requirements.
• Running specific tools and applications, such as Oracle databases and SAP RISE.

Undoubtedly there will launch a number of targeted campaigns to run on these systems in future. When Power 11 is launched, IBM will introduce it into the PowerVS offerings. I expect it to be adopted particularly by its business partners where the inclusion of IBM i will help them to run managed services for an important sub-set of small and medium companies; especially in EMEA and Italy in particular.
At launch IBM has worked with Precision IT to compare its Total Cost of Ownership (TCO) with its leading competitors, suggesting that HPE’s Greenlake and Oracle’s Cloud Compute @ Customer are 20% more expensive and AWS’s Outposts, 50% more. Results that should perhaps be taken with a pinch of salt, since the comparison is between a new solution with three that have been in the market for sometime. Nevertheless I believe this is an important part of the development of IBM’s hybrid cloud strategy and expect it to be successful in extending revenues from Power 10 at a time when they will fall in the run up to the introduction of Power 11. It reminds me somewhat of its introduction of PureSystem in 2012 – a solution which pre-dated the sale of System x to Lenovo in 2014. IBM has the right expertise to make this a long-term success.