Enabling the data center

of the future
The case for on-premise XaaS data centers
Enabling the data center of the future | The case for on-premise XaaS data centers

Introduction

The accelerated and sustained growth of cloud services adoption The rationale for hosting workloads on-premise may extend beyond
has transformed the world of information technology (IT) over the finances. In this paper, we will explore additional circumstances
past decade, repositioning enterprise IT—previously considered where it makes sense to keep or move workloads on-premise.
a pure support function—as a primary player in operations, new For these cases, we will discuss how to operationalize on-premise
product development, and in-house innovation hubs. Indeed, the hosting to realize a competitive advantage, answering the question:
benefits that the cloud has helped catalyze across end users Should all our workloads live in the cloud or should we repatriate
and small and medium enterprises are numerous, substantial, any workloads to our on-premise data center?
and undeniable.
The answer is twofold: 1) Strategically balancing a multi-cloud
IT, once a major upfront investment and a barrier to entry for small strategy that intelligently identifies workloads, operations, and
companies, is now available at any scale and is fully self-service. As teams that are operating in the cloud, in a hybrid cloud/on-premise
a result, a small player can leverage cloud services to spend more arrangement, or in on-premise data centers; and 2) using
time on mission-critical activities dedicated to growing the company everything-as-a-service (XaaS) to enable the on-premise data
without having to invest in additional talent and one-time capital center of the future.
expenditures (CapEx) for back-end operation assets. The cloud’s
flexibility and scalability provide an ideal platform to quickly ideate, Gartner projects that by 2025, 85% of infrastructure strategies
build, validate, and grow a product. In light of the benefits associated will integrate on-premise, colocation, cloud and edge delivery
with the cloud, one could argue that starting in the cloud should be options, compared to 20% in 2020.2 The issue is not cloud versus
a baseline strategy for most companies. an on-premise data center, but rather balancing flexibility and
realizing optimization at scale.
Still, the cloud’s flexibility and scalability may come at a cost
premium, and a company operating at scale may find their costs of
operating within the cloud to exceed that of maintaining an
on-premise (on-prem) operation: Equivalent workloads running
in the cloud under public prices are estimated to be 10–12 times
more expensive than the cost of running on-premise data centers—
even accounting for talent, hardware, and real estate costs.1 This
observation should not come as a surprise since publicly listed
cloud prices often incorporate negotiated discounting, especially
in business-to-business (B2B) environments. As a result, a
company’s cloud cost can still be more expensive than running
data center operations on-premise, in spite of discount pricing
from committed capacity agreements.

2
 Enabling the data center of the future | The case for on-premise XaaS data centers

Amid cloud’s growth, is the on-premise

data center going away?

Trends show that cloud providers may continue to gain market share According to Dell Technologies, “The inevitable result of this on-prem
in generalized storage and compute in coming years. In fact, both vs. cloud debate is that organizations frequently adopt multiple
Statista and Synergy Research Group cited 2019 as the first year that public cloud solutions and while also investing in private cloud
cloud infrastructure services (CIS) spend surpassed traditional data infrastructure, creating a multi-cloud environment that is intended
center hardware and software spend (figures 1 and 2).3 However, to meet all needs.”4 Similarly, Amazon Web Services (AWS) extended
CIS growth has not come at the expense of data center hardware its reach in 2019 by offering products for infrastructure on-premise
and software spend, which has roughly remained constant at a as part of its AWS Outposts product. AWS Outposts’ list of benefits
compound annual growth rate (CAGR) of roughly 2.2% over the past cites enabling a consistent hybrid experience by running a host of
10 years. AWS storage and compute services on-premise, recognizing use
cases and industries that will remain on-premise in the future.5
Data supports the idea that enterprises are spending more on
IT infrastructure services today than before, but will this trend Table 1 presents six key factors that make workloads strong
continue? How will enterprises manage their on-premise data center candidates for staying (or repatriating to) on-premise operations
operations in the years to come, and will there eventually be a to achieve an optimal operational mix.
decline in future data center spend?
Importantly, IT managers should apply rigor beyond analyzing if an
We see hybrid cloud/on-premise and distributed cloud becoming the application, workload, or dataset can be migrated to the cloud to
future of IT, and many organizations in the technology industry echo understand if it should be migrated from a business value point
this hypothesis. of view.

Figure 1. Global cloud and data center spending 2020 Figure 2. Enterprise spending on cloud data centers

3
Enabling the data center of the future | The case for on-premise XaaS data centers

Table 1. Key factors and illustrations for candidates of on-premise workloads

Key factors Description

Assets are subject to Application, workload, or database has jurisdictional privacy, security, data residency, or other regulatory and
external regulatory compliance requirements that will pose a significant challenge, create risk, or require on-premise residency.
and/or compliance
requirements Example: Since the enforcement of the General Data Protection Regulation (GDPR) in May 2018, “Software-as-a-service (SaaS)
providers that collect data on EU residents need to have much stronger controls around the collection, tracking, sharing and
protection of that user data.” As the collection of that user data grows, some assert that “the cost of a hosted, multi-tenant
delivery model [may increase] relative to a single-tenant, ‘on-premises’ delivery model that limits data collection.”6

Application, workload, Application, workload, or database requires a specific hardware asset or configuration that is difficult to find in the
or database requires public cloud market. The combination of technical specs requires customization best suited for in-house and on-
specialty hardware premise execution.
and/or special
configuration Example: Machine learning for video-graphic content and engineering/architecture simulation modeling are recommended to
run with at least 16GB of RAM (up to 72GB in extreme cases) with very few virtual CPUs. Such configurations are best created
in-house given that a hyperscaler will simply offer its biggest instance type available, which is not a cost-effective option.

Application, workload, Application and workloads require an exceptionally low latency to deliver on mission. The company’s latency needs
or database requires require maintaining a private network with customized configuration of hardware assets.
exceptionally low
latency Example: Online gaming and e-sports experiences require the lowest latency possible, while rendering detailed graphics,
security, multiplayer live-play, and streaming to external audience platforms. While cloud providers vouch to reduce latency to
as much as 20% of standard acceptable lag, on-premise solutions are still preferred to support the highest quality gameplay.7

Time, effort, and Workload requires a predictable and numerous amount of resources, and its availability is mission-critical without a
complexity to refactor forecastable need for legacy functionality changes.
and migrate an
existing application Example: Health insurance companies adjudicated up to 19.4 million in-network claims in 2022 according to cms.gov.8
and copious number Business systems are often built on legacy mainframe architecture, and the code reflects decades of business rules in
of resources adjudication engines on-premise. Mainframe continues to provide secure, fast, and reliable execution for legacy back-office
systems unlikely to change.

Capacity for application Application and workloads are containerized and run on horizontally scalable server clusters that can be adjusted
or workload needs to for capacity by adding or removing servers.
easily be adjusted by
adding or removing Example: Platform9 offers a managed service for maintaining on-premise deployments of Kubernetes clusters, giving
servers on-premise in-house developers a cloudlike experience and the ability to develop highly scalable applications in a private cloud based
on the capacity of the compute cluster available. By adding hardware to a data center, the capacity of these applications is
increased proportionately.

Operating an Internet IoT products incur high and growing egress costs, as they transfer terabytes of data through their digital solutions
of Things (IoT) solution each day.
at scale generates high
egress costs Example: Considering that certain cloud providers charge up to $0.12/GB for egress fees, extensive IoT solutions, which involve
transferring terabytes of data daily, are a growing cost to take into consideration.

4
 Enabling the data center of the future | The case for on-premise XaaS data centers

The role of XaaS in the

data center of the future
As organizations continue to move into the hybrid and distributed In an XaaS model for product-centric businesses, a customer
cloud world of IT, on-premise data centers will need the ability to consumes hardware assets and maintenance as a service (aaS),
respond quickly to changing needs to overcome the general view giving IT control and greater flexibility over what they consume
that on-premise operations tend to sacrifice flexibility for control. and pay for. XaaS offerings for product-centric businesses typically
Companies should consider shifting to as-a-service (aaS) business bundle hardware with software and support services to give users
models to create more flexibility within the on-premise data center optimal management of their assets and the ability to rightsize their
and address pain points experienced with traditional one-time on-premise hardware asset inventory according to their needs at any
hardware asset purchases. According to Deloitte’s 2021 edition of point in the growth journey. This typically removes the need for bulk
its Everything-as-a-Service (XaaS) Study, more than 50% of companies investments in hardware assets, instead enabling ongoing increment
classified as leading adopters of as-a-service models reported and asset mix adjustments according to business needs. In this
improved decision-making, better collaboration, accelerated way, the hardware asset provider builds a relationship driven by the
experimentation, and an ability to reinvent business processes as customer’s success and growth.
organizational benefits of adopting XaaS offers in their
business operations.9

5
Enabling the data center of the future | The case for on-premise XaaS data centers

In table 2, Deloitte research shows that XaaS can

solve the following pain points that may arise when
product-centric businesses run on-premise data
centers under a traditional one-time purchase for
hardware assets.

Table 2. XaaS addresses one-time purchase data center pain points

Pain points One-time purchase model XaaS purchase model

Reduced time to value Hardware assets purchased in one-time transactions In an XaaS model, customer does not purchase the asset;
due to complex, time- are typically classified as a CapEx. CapEx contracts instead, they incur much lower monthly fees that are
consuming CapEx typically involve high dollar amounts and negotiations typically classified as recurring operating expenses (OpEx).
purchase cycle can be complex and time-consuming, requiring months As a result, it is usually easier to obtain corporate approval,
of internal approval and putting the onus on IT to making XaaS better suited for incremental purchases
demonstrate the business value of the asset purchase. and subscription changes in response to evolving
customer needs.

Hardware asset Customer must trigger a new CapEx purchase cycle Since the customer does not own the asset, they do not
obsolescence due after a typical hardware life cycle of two to five years.10 have to worry about asset obsolescence, as the provider is
to lengthy intervals tasked with managing the asset’s useful life.
between hardware
asset renewals

Significant upfront High upfront investment limits the customer’s capital A monthly recurring fee model allows customers to
investment required allocation opportunities. It also deters customers with consume hardware assets without a significant upfront
cash budget limitations. investment. (Most companies do require a minimum
commitment for a specified term period.)

Limited ability to Customer must trigger a new CapEx purchase cycle Customer can adjust their hardware asset mix through
support wide variety of to change data center hardware asset mix. Also, they upgrades, add-ons, or renewals to accommodate a variety
workloads due to fixed cannot return or exchange hardware assets with a one- of workloads.
hardware asset mix time purchase model.

Internal teams may be Customer must trigger a new CapEx purchase cycle to Customer may adjust total hardware asset count at any time
constrained by fixed increase a single data center’s capacity. under the same contract:
capacity and run-time
• Add/remove hardware to scale within the terms of the
to accommodate
contract considering minimum commitments
respective workloads
• Ability to rightsize hardware assets

Physical asset Customer must trigger warranty for replacement of Ongoing service for replacement of faulty equipment is the
ownership is an faulty equipment. Warranty time span is limited. responsibility of the provider and typically included in a
ongoing responsibility service offering, for example:
to maintain
• Rack fans

• Faulty drives

• Cables

6
 Enabling the data center of the future | The case for on-premise XaaS data centers

By solving these and other customer pain points, XaaS has the These solutions revive concepts proved in the mid-2000s, allowing
potential to provide the following benefits: customers to purchase a “data center in a box,” a fully integrated,
plug-and-play data center inside a shipping container designed to
• Asset flexibility
expand onsite data center operations. Other promising models
Customer may adjust on-premise hardware asset mix
include partnerships between colocation data center providers
incrementally at any point in time.
and server management software providers. Such market solutions
• Scalability on demand further support the idea that there will continue to be an appetite for
Customer may adjust capacity and pay only for the assets on- on-premise or colocated data center operations, particularly those
premise at any point in time. that allow companies to rightsize IT and that demonstrate a real
commitment for providers to grow with their customers.
• Purchasing experience
Customer pays for IT hardware assets and maintenance as a
It is important that the data center of tomorrow position itself
recurring OpEx expense.
as a scalable and flexible resource aligned with the hybrid cloud
• Ongoing support architecture of the future. By consuming hardware assets under
Customer selects the desired level of support to help optimally an XaaS model, customers can position their on-premise data
manage assets. centers as internal, operationally optimized resources that deliver a
consistent cloudlike experience.
We acknowledge that despite the added flexibility that comes
with consuming hardware assets as a service, not all customers
have the physical space, energy capacity, or cooling systems to
adjust capacity on-premise. Our research found that hardware
providers are responding by offering modular data center solutions
designed to address customers with limited space, edge computing
needs, and even modular cryptocurrency mining operations.

7
Enabling the data center of the future | The case for on-premise XaaS data centers

Beyond the data center

of tomorrow

This paper does not intend to narrow the relevancy of XaaS business Companies will need to manage fleets of these devices that require
models to on-premise data center operations. The flexibility and ownership and consumption models that fit a particular business
scalability that XaaS models provide also apply to many other need. XaaS models are positioned well to support both device fleet
promising business operations technologies. For example, the management and edge computing capabilities that allow changing
number of connected devices is forecasted to be nearly double needs for scaling and hardware flexibility.
from 15.14 billion in 2023 to 29.4 billion by 2030.11

Contacts

Jordan Cole Veronica Pulido

Everything-as-a-Service Practice Lead Everything-as-a-Service Leader
Deloitte Consulting LLP Deloitte Consulting LLP
[email protected] [email protected]

Contributors

Jorge Reyes
Malavika Sharma
Rohan Talukdar
Julia Castanet

8
Endnotes

1. Sarah Wang and Martin Casado, “The cost of cloud, a trillion dollar paradox,” Andreessen Horowitz, May 27, 2021.

2. David Cappuccio and Roger Williams, “The everywhere enterprise: A Gartner Q&A with David Cappuccio,” Gartner, October 6, 2020.

3. Lionel Sujay Vailshery, “Global cloud and data center spending 2021,” May 15, 2023; Synergy Research Group, “2020 – The year that cloud service revenues finally
dwarfed enterprise spending on data centers,” March 18, 2021.

4. Dell Technologies, “Hybrid cloud management,” accessed May 22, 2023.

5. Amazon Web Services, “AWS Outposts Family overview,” accessed May 22, 2023.

6. Taylor Wakefield, “GDPR – SaaS reconsiders on-premises software,” Teleport, March 26, 2018.

7. Liz Alton, “An introductory guide to esports technology,” Connected, September 27, 2019.

8. Centers for Medicare & Medicaid Services, “Health Insurance Exchange Public Use Files (Exchange PUFs),” accessed May 22, 2023.

9. Susanne Hupfer et al., Gaining and sustaining a competitive edge with cloud and as-a-service IT, Deloitte Insights, 2021.

10. Petroc Taylor, “Data center server replacement frequency 2020,” March 2021.

11. Lionel Sujay Vailshery, “Number of IoT connected devices worldwide 2019–2021, with forecasts to 2030,” November 22, 2022.

About this publication

This publication contains general information only and Deloitte is not, by means of
this publication, rendering accounting, business, financial, investment, legal, tax, or
other professional advice or services. This publication is not a substitute for such
professional advice or services, nor should it be used as a basis for any decision
or action that may affect your business. Before making any decision or taking any
action that may affect your business, you should consult a qualified professional
adviser. Deloitte shall not be responsible for any loss sustained by any person who
relies on this publication.

About Deloitte
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private
company limited by guarantee (“DTTL”), its network of member firms, and their
related entities. DTTL and each of its member firms are legally separate and
independent entities. DTTL (also referred to as “Deloitte Global”) does not provide
services to clients. In the United States, Deloitte refers to one or more of the US
member firms of DTTL, their related entities that operate using the “Deloitte” name
in the United States, and their respective affiliates. Certain services may not be
available to attest clients under the rules and regulations of public accounting.
Please see www.deloitte.com/about to learn more about our global network of
member firms.

Copyright © 2023 Deloitte Development LLC. All rights reserved