2025 United States Data Center Market Outlook PDF Free Download
1 / 20/20
100%
2025
United States
Data Center
Market Outlook
Introduction
Infrastructure, both digital and traditional, has experienced cyclical and
structural transformations throughout U.S. history.
As we examine the commercial real estate landscape in 2025, digital
infrastructure stands out as the only segment in a boom – and a structural
boom at that – driven primarily by investment in AI-related data center
development. The level of spending in the past two years on the facilities
that support digital-age innovation is unprecedented, and is redirecting
commercial real estate capital, physical structures and powered land into
the data center sector. This trend is particularly evident in industrial real
estate, which shares more overlap with the data center segment than any
other sector.
Despite the ongoing generational growth, data center development is
struggling to keep pace with advancements in AI requirements and rising
demand due to one tangible constraint in particular — power.
This dynamic is both extending the new paradigm of infrastructure scaling
and spurring the exploration of new markets and solutions to meet the
demand. Adding further complexity, a key question has arisen: How will
these massive investments in AI demonstrate their value in the years to
come, especially as international competitors develop alternative AI models
that may be more cost-eective and energy-ecient?
Newmark’s U.S. 2025 data center sector report oers an in-depth analysis
of the national data center market, examining key trends, transactions and
emerging markets, and providing an outlook for what’s to come.
4NEWMARK
National
Overview
The rise of AI requires immense
computational power and density
that has well surpassed the existing
U.S. data center inventory, primarily
composed of retail/wholesale1 data
centers focused on cloud computing,
high performance computing (HPC)
and data storage, with lower power
capacity than what AI requires. Typical
data center power requirements only
recently started to surpass 50 MW.
Last year saw multiple transactions for
data center campuses between 400
MW and 900 MW, with one contract
in excess of 1 GW. Expectations are
for 1 GW+ requirements to increase in
number during this year and beyond.
The demand for new infrastructure has
propelled data center development
to new heights and markets. While AI
is a major catalyst for development,
not all new growth is driven by it.
There is also a notable increase in
edge deployments of smaller-scale
infrastructure in demographically
dense locations to extend the reach of
computing capabilities. Nonetheless,
AI remains the primary growth driver,
fueled by extraordinary investment
by hyperscalers. Microsoft invested
an estimated $40 billion in AI data
center capacity in 2024 and has
announced plans to spend $80 billion
on building AI data centers in FY 2025.
Source: New Street Research, August 2024.
Operating costs include cash operating expenses, software, depreciation and electricity.
AI data center spend reflects a portion of total hyperscaler capex.
2024 AI Data Center Spend 2024 Data Center Operating Costs
GPUs and
Other Chips
Other AI
Spend
Total AI
CapEx
Training
and R&D Inference
Total Op.
Costs
Microsoft $20 $20 $40 $3 $3 $6
Meta $11 $12 $23 $2 $2 $4
Google $14 $15 $29 $3 $1 $4
Amazon $8 $8 $16 $2 $1 $3
Tier 2 Cloud $26 $26 $52 $8 $3 $11
Enterprises &
Government $26 $26 $52 $8 $2 $10
Total $105 $105 $210 $27 $12 $39
Hyperscaler AI Data Center and Operating Costs Spending, 2024
$ in Billions
1 Wholesale here refers to a data center dedicated to one user, vs Retail where multiple customers use a shared facility.
52025 United States Data Center Market Outlook
Meta announced that its capital
expenditures would grow to as much
as $65 billion this year, compared to
roughly $38 to $40 billion in 2024.
Google’s parent Alphabet expects
capital expenditures of $75 billion
in 2025, 43% higher than 2024.
And, not to be outdone, Amazon
said it would spend $100 billion on
capex this year, largely driven by
investments in AI.
This sustained investment
across the industry is reflected
by the remarkable expansion in
construction activity. By the end of
2024, the data center development
pipeline reached nearly 50 MSF,
eectively doubling the volume from
five years ago. Capital deployment
in data center construction reached
an all-time high of $31.5 billion
annually in 2024, with no signs of
plateauing this year. While analyses
vary on adoption scenarios, some
forecasts, including an October
2024 analysis by McKinsey, suggest
that up to 70% of total data center
demand will be AI-driven by 2030,
up from under 50% currently. This
trend will require significant further
expansion of digital infrastructure
for training, inference and other
use cases in the coming years.
When AI models are being trained,
typical data center site selection
factors, such as latency, are less
critical compared to power access,
vastly broadening the geographical
scope for data center development.
When an AI model is put into
operation (inferencing), latency
becomes crucial, limiting location
options. The precise balance of new
infrastructure needed to support
AI growth through 2030 remains
uncertain, but it is clear that the
expansion rate of new data center
capacity will be significant. At least
twice the data center capacity built
since 2000 will need to be built in
less than a quarter of the time2.
2 McKinsey, October 2024.
Source: Newmark Research, CoStar
U.S. Data Center
Development Pipeline
50
45
40
35
30
25
20
15
10
5
0
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
MSF
Source: U.S. Census
U.S. Private Data Center
Construction Spending
35,000
30,000
25,000
20,000
15,000
10,000
5,000
0
2014
2015
2016
2017
2018
2019
2020
2021
2022
2023
2024
$M
Nov 2022
Launch of ChatGPT
Power:
The Primary
Limitation
on Growth
The data center development
pipeline — already at record heights
— would be substantially higher if
not for the singular issue of power
constraints. Analysis from S&P
Global Market Intelligence indicates
that the power demand from both
existing and planned datacenters in
the U.S. is expected to total about
30.7 GW once all data centers are
operational. However, utilities are
set to supply only 20.6 GW of that
needed capacity.
Further exacerbating this challenge
is that even where power capacity
is readily available, accessing it can
often take years due to the lack
of equipment to build supporting
transmission infrastructure.
Solving for the massive disparity
in power supply and demand is
driving some data center users
and developers to adopt a “Bring
Your Own Power” (BYOP) strategy.
Hyperscalers and developers are
signing large-scale power purchase
agreements with geothermal and
nuclear projects, exploring natural
gas, hydrogen, solar, wind and
battery solutions, and investing in
emerging technologies like small
nuclear reactors (SMRs). Post-2030,
more data center power demand
could be ‘behind the meter’, tied
to microgrids or SMRs, assuming
regulatory hurdles are overcome—
and could significantly reduce
the need for long-distance power
transmission and associated losses.
Growing scrutiny of traditional
transmission methods, heightened
by recent devastating wildfires in
the western U.S., underscores the
urgency in seeking cleaner and
safer generation and alternative
transmission solutions within the
data center industry and beyond3.
Nevertheless, purchasing power
directly from a microgrid or single-
source provider is not a decision
data center operators take lightly,
due to potential environmental
and reliability issues. Managing a
microgrid demands expertise largely
beyond what most data center
operators possess and is typically
better handled by a large utility with
extensive specialized resources.
Similarly, managing a nuclear facility
presents significant challenges,
even for traditional power utilities,
highlighting the complexity of
alternative power solutions.
6NEWMARK
3 Western Fire Chiefs Association, June 2024
The data center development pipeline
would be substantially higher if not for
the singular issue of power constraints.
72025 United States Data Center Market Outlook
Sources: S&P Global Market Intelligence; 451 Research; S&P Global Commodity Insights
Power Demand from Operational and Planned Data Centers
8NEWMARK
Emerging
Markets
The need for vast capacity of computing
power along with the land and power to
support it is driving capital deployment
into markets that have not previously
seen such investment. Traditionally,
data centers were valued on low
power cost and ease of construction.
However, with many power utilities in
traditional data center markets tapped
out and some communities actively
discouraging further development,
growth is moving to regions that have
not previously been in consideration.
‘Emerging markets’ like Pennsylvania,
the Carolinas and West Texas are
increasingly locations of choice for
data center developments focused on
latency-agnostic use cases such as
training or crypto mining operations.
Northern Virginia, Dallas, Chicago,
Phoenix, and Northern California remain
primary data center markets, but data
center development is now occurring in
at least 23 states nationwide.
Emerging Market Spotlight:
West Texas
The West Texas region, rich in energy and sparse in
population, is the least densely-populated area in the
state and would not appear on any lists of major traditional
commercial real estate markets. Yet, it has garnered
remarkable and growing interest within the data center
sector due to its abundant and inexpensive power supply.
A leader in renewable energy and natural gas generation,
the region does not yet have the infrastructure to send
power to more populous areas in significant amounts, in
a sense ‘trapping’ the power where it is and driving prices
down. In January 2025, the new federal administration
announced a $500 billion initiative, “Project Stargate,”
aimed at building domestic AI infrastructure, with the
first phase to begin in Abilene, Texas. This continues
to position the West Texas city as an emerging hub for
data center development. Additionally, the region has
seen impressive growth from cryptocurrency mining
data center development, with 9.1 GW of total potential
crypto mining capacity over at least 17 announcements
in the past three years (1.7 GW of which is currently
operational). Some of these mining operators are now
exploring using their infrastructure and capacity for AI
and high-performance computing workloads, given the
urgency of that demand. These facilities, designed for
high power and cooling demands, are well-suited to
support AI training with minimal modifications.
92025 United States Data Center Market Outlook
Source: Cantor Fitzgerald Research, Newmark Research. Represents announcements made from Riot Platforms, Cipher Mining, IREN, Bitdeer, Core Scientific, MARA
Holdings, Hut 8 and Galaxy.
Cryptocurrency Mining Data Announcements, 2021 - 2024
10 NEWMARK
Data Centers and the
Traditional Industrial Market
Data centers and traditional industrial real estate share more fundamental structural and operational
similarities than any other sectors of real estate. At first glance, the relationship is as basic as dirt:
both typically seek similar land sites for development. However, there’s more than competition for
land happening here, as demand for one often begets demand for the other. For instance, Amazon
needs data centers to support the algorithms and computing power that predict customer orders, and
strategically positions logistics facilities accordingly. The interplay between these two sectors along
with their shifting market conditions manifests in numerous ways, which we’ll explore below.
DEVELOPMENT PIPELINES
Between 2021 and 2022 major
development growth occurred across
numerous commercial real estate sectors,
spurred by demand and low interest rates.
The industrial sector was particularly active,
with the U.S. pipeline swelling to a record
high of nearly 750 MSF in 2022. What goes
up, however, often comes down. Higher
interest rates and normalizing demand
have curtailed new starts and caused a
historically sharp drop in the industrial
pipeline from its peak in 2022 to 2024.
Over the past two years, industrial capital
has been ever more competitive within
the data center space; as demand waned
for new logistics groundbreakings and
boomed for new data center development,
major industrial investors such as Prologis
and NorthPoint expanded their data center
development. In addition, prime industrial-
zoned sites, traditionally earmarked for
logistics facilities, are increasingly being
acquired for data center development.
These dynamics will result in less available
land for constructing traditional logistics
product in the future.
Source: Newmark Research, MSCI, January 2025
*Development site acquisition volumes of $10M or more in the stated period; acquisitions
by hyperscalers and pure-play data center developers represent the share “targeting data
center development,” which may materially understate the full scope of industrial-zoned sites
acquired and set for data center development.
Share of Industrial Site Acquisition Volumes*
Set for Data Center Development
30%
25%
20%
15%
10%
5%
0%
2017 - 2019 Last 24 Months
7%
24%
112025 United States Data Center Market Outlook
“M(AI)D IN THE USA”
Supply chain volatility, taris, geopolitical concerns, corporate derisking strategies and federal incentivization of critical-
industry repatriation are among the long list of factors driving U.S. manufacturing occupancy growth through reshoring,
foreign direct investment (FDI) and domestic expansion. At the forefront of this trend is the semiconductor industry. Two
companies, TSMC and Samsung, dominate the global supply, producing close to 100% of advanced chips, with TSMC
currently being the exclusive supplier of advanced chips to NVIDIA. While most of this production has traditionally taken
place in Taiwan, the landscape is shifting as both companies, along with many others, have invested tens of billions of
dollars to establish fabrication plants in the U.S. This acceleration in semiconductor manufacturing is benefitting the
industrial and data center markets alike by securing and shortening supply chains, safeguarding intellectual property,
creating jobs, bolstering economic growth, and increasing industrial occupancy, among other positive impacts.
Source: Source: FRED, Index 100=2017
150
140
130
120
110
100
90
80
70
60
11/2017
3/2018
7/2018
11/2018
3/2019
7/2019
11/2019
3/2020
7/2020
11/2020
3/2021
7/2021
11/2021
3/2022
7/2022
11/2022
3/2023
7/2023
11/2023
3/2024
7/2024
11/2024
U.S. Industrial Production Index:
All Manufacturing (——) vs. High Tech (Semiconductors) (——)
LOGISTICS LEASING AND OCCUPANCY
The ripple eects of data center and semiconductor fab construction are impacting the logistics sector.
As major specialized facilities are established and expanded, there are increasing contractor-related
requirements to store equipment and supplies for build-out. Additionally, warehousing and distribution
requirements for specialized materials and components are rising in a maturing semiconductor
manufacturing ecosystem. The outsized development of data centers drove some of the largest
warehouse leases in 2024, with hyperscalers such as Google and service providers such as Integral
Mission Critical inking numerous warehouse leases exceeding one million square feet in markets
where data center development is active.
Capital Markets, M&A
and Financing Landscape
The data center capital markets and merger/acquisition landscapes, like the development segment, were highly dynamic
in 2024. Among numerous notable M&A deals was Blackstone’s acquisition of APAC data center firm AirTruck for $14.9
billion, making it the second largest-ever deal in the space after KKR/GIP’s acquisition of CyrusOne in 2021 for $15.45
billion. Nearly $5 billion in volume transacted last year, up 88.5% from 2023 in dollar volume as the number of assets
that traded hands was largely stable, but the price per pound substantially higher. M&A is expected to remain elevated,
with an estimated $29 billion in M&A already agreed upon but still in progress, according to Synergy Research Group.
As data center operators pursue capital recycling programs (sale of stabilized, yielding assets to support development
opportunities) and merchant developers begin to deliver projects, the volume of stabilized, single-asset or portfolio
transactions is expected to increase materially, both in terms of number of transactions and overall volume. There is a
significant amount of capital formation required to support the continued execution of stabilized transactions given the
balance of transaction scale, cost of capital, and capitalization rate expectations. The timing of this capital formation will
directly impact the stabilized transaction market in 2025.
12 NEWMARK
4 Synergy
Forecasts indicate that by 2030, hyperscaler-owned
facilities will comprise over 40% of data center
capacity, up from below 25% currently4.
132025 United States Data Center Market Outlook
Source: NMRK Research, JP Morgan, RBC, Wells Fargo
(1) Includes adjustments for booked-but-not-billed (“BBnB”) revenue, known-churn, and other one-time adjustments
Select Global Data Center Mergers & Acquisitions, 2023-2024
DateDate TargetTarget Buyer(s)Buyer(s) TypeType TransactionTransaction MarketMarket
Transaction Transaction
Size / EVSize / EV
Contracted Contracted
EBITDA(1)EBITDA(1) EV/EBITDAEV/EBITDA
Oct '24Oct '24 DatabankDatabank AustralianSuperAustralianSuper WholesaleWholesale MinorityMinority North AmericaNorth America $2,000$2,000 NDND ~22.0x~22.0x
Oct '24Oct '24 Global Switch HMC Capital Wholesale Sale / Majority APAC $1,400 $54$54 ~26.0x~26.0x
Sep '24Sep '24 Airtrunk Blackstone Hyperscale Sale / Majority APAC $14,906 $573$573 ~26.0x~26.0x
Sep '24Sep '24 EdgeConnex Sixth Street Hyperscale Minority North America $1,000 NDND ~22.0x~22.0x
Jan '24Jan '24 Vantage (NA) Silverlake Hyperscale Minority North America $6,400 NDND ~21.0x~21.0x
Sep '23Sep '23 Vantage (EMEA) AustralianSuper Hyperscale Minority EMEA $1,600 NDND ~20.0x~20.0x
Sep '23Sep '23 Serverfarm Manulife Wholesale Sale / Majority North America, EMEA $1,500 $61 24.4x
Jun '23Jun '23 Compass Brookfield, Ontario Teachers Hyperscale Sale / Majority North America, EMEA $5,500 $229$229 24.0x24.0x
May '23May '23 ODATA "Aligned Data Centers /
SDC Capital Partners" Hyperscale Sale / Majority Latin America $1,860 $76$76 24.4x24.4x
Average:Average: ~23.3x~23.3x
DateDate TargetTarget Buyer(s)Buyer(s) TypeType Deal TypeDeal Type MarketMarket
Cap Cap
RateRate
Dec '24Dec '24 JDM PartnersJDM Partners HMC CapitalHMC Capital Powered ShellPowered Shell PortfolioPortfolio "Richardson, TX "Richardson, TX
/ Olathe, KS"/ Olathe, KS" 6.25%6.25%
Dec '24Dec '24 Prologis / SkyboxPrologis / Skybox HMC CapitalHMC Capital TurnkeyTurnkey AssetAsset Chicago, ILChicago, IL 5.93%5.93%
Dec '24Dec '24 PowerhousePowerhouse CyrusOneCyrusOne Powered ShellPowered Shell AssetAsset Ashburn, VAAshburn, VA NDND
Jul '23Jul '23 Digital RealtyDigital Realty TPG (JV)TPG (JV) TurnkeyTurnkey PortfolioPortfolio NoVANoVA 6.00%6.00%
Jul '23Jul '23 Digital RealtyDigital Realty GI Partners (JV)GI Partners (JV) TurnkeyTurnkey PortfolioPortfolio Chicago, ILChicago, IL 6.50%6.50%
Jul '23Jul '23 DirecTVDirecTV GI PartnersGI Partners Powered ShellPowered Shell Sale LeasebackSale Leaseback Los Angeles, CALos Angeles, CA 6.40%6.40%
Jun '23Jun '23 Digital RealtyDigital Realty DataBankDataBank TurnkeyTurnkey AssetAsset Dallas, TXDallas, TX 4.40%4.40%
Apr '23Apr '23 Digital RealtyDigital Realty GI PartnersGI Partners TurnkeyTurnkey AssetAsset Ashburn, VAAshburn, VA 5.50%5.50%
Apr '23Apr '23 VantageVantage "Consortium led by "Consortium led by
MEAG / Infranity"MEAG / Infranity" TurnkeyTurnkey PortfolioPortfolio EMEAEMEA 5.25%5.25%
Average:Average: 5.78%5.78%
Select Asset Sales, 2023 - 2024
Source: Newmark Research, JP Morgan equity research, RBC equity research
Source: MSCI
U.S. Data Center Sales Volume
7
6
5
4
3
2
1
0
2Q15
4Q15
2Q16
4Q16
2Q17
4Q17
2Q18
4Q18
2Q19
4Q19
2Q20
4Q20
2Q21
4Q21
2Q22
4Q22
2Q23
4Q23
2Q24
4Q24
Billions
14 NEWMARK
An illuminating way to more fully grasp the explosion in data center trades is to look at the land acquisition of the
hyperscalers. In 2024, development site purchases by hyperscalers such as Meta, Microsoft, Amazon, Alphabet and
Apple were four times higher than their average from 2010 to 2023. For the first time, hyperscalers have surpassed 10%
as a share of total development site purchases tracked by RCA, illustrating their significant influence on the market.
Hyperscaler Land Purchases Growing Exponentially
Share of Dev Site Transaction
Volume (%)
Source: Newmark Research, MSCI
14%
12%
10%
8%
6%
4%
2%
0% 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
In 2024, Northern Virginia persisted as the leading region for development site purchases by hyperscalers capturing
roughly one-third of the total volume and cementing its position as the nation’s most critical data center hub. Atlanta
was second, mostly due to EdgeconneX’s acquisition of a 63-acre plot for over $300 million, where they plan to build a
324 MW data center. Microsoft and EdgeConneX together account for 90% of hyperscaler development site purchases
in the U.S., with CyrusOne, QTS, Stack Infrastructure and Alphabet rounding out the list.
152025 United States Data Center Market Outlook
State of the
Debt Capital Markets
Data Center Lending & Pricing Update
Despite the high-interest rate environment, the data center debt capital markets sector has
shown a distinct ability to navigate the perceived headwinds with very little turbulence. To
that end, 2024 displayed the market’s deep liquidity and robust lender confidence in digital
infrastructure assets at every stage of their life cycle, from development to stabilization. With
the combination of existing lenders doubling down and new entrant lenders deploying in a
meaningful way, there is a deep and diverse array of lender types seeking opportunities to
finance data center & digital infrastructure related projects. From project finance, TMT, and
real-estate investment banks to private credit funds, and infrastructure-focused lenders, there
is broad and aggressive interest to support the sector. This availability of financing solutions
allows developers and operators to access capital for new builds, acquisitions, or refinancings,
ensuring flexibility and competitive terms. With the robust adoption of cloud, AI (training &
inference), and enterprise computing workloads, this support has been, and will continue to
be critical to the industry’s ability to keep pace with the tremendous capital demands.
The following summaries provide general feedback on data center pricing and sizing. It’s
important to note that each deal is unique, and there may be outliers.
Land & Pre-Development Financing
A trend that started towards the end of 2023 and continued to gain momentum throughout
2024 is the willingness of lenders to take on exposure during the land acquisition and pre-
development phases of construction. For both single site financings and large revolving
facilities, there is a growing and competitive pool of lenders looking to deploy capital at the
earliest stage of a data center development’s life cycle. This financing, although at a premium
to development spreads, has proven to be extremely accretive to sponsors/operators who
are not capitalized to deploy and hold large quantums of balance sheet capital for land/pre-
development work. Although the pool of lenders playing in the space is still well below the
appetite for vertical development financings, there is a strong and growing mix of lenders who
are convicted on the opportunity. These groups see this not only as an attractive risk adjusted
return, but also as a very unique, and less competitive, way to gain exposure into the data
center space.
Please see table below for broader pricing guidelines:
LAND / PRE-DEVELOPMENT FINANCINGLAND / PRE-DEVELOPMENT FINANCING
LenderLender Te rmTe r m Fixed/ FloatingFixed/ Floating CharacteristicsCharacteristics
Key Sizing Key Sizing
ParametersParameters
Leverage Leverage
ConstraintsConstraints RecourseRecourse SpreadsSpreads
CRE Banks 2-5 Yrs Float Low
Leverage LTC / LTV LTC: 50-55%
LTV: <50% Yes 4.25% - 5.25%
Private Credit 2-5 Yrs Float/Fixed Low-Max
Leverage LTC / LTV LTC: 50-75%
LTV: <55% No 4.75% - 7.00%
Hedge Funds 2-5 Yrs Float/Fixed Max
Leverage LTC / LTV LTC: 65-75%
LTV: <55% No 6.25% - 7.00%
KEY FINANCING
THEMES
Deep and Growing
Lender Pool
fueled by new entrants
providing an abundance of
available debt capital.
Flexible Financing
Structures
resulting in hybrid
executions purpose-
built to accommodate
both developer/operator
and capital investment
requirements.
Robust Appetite for
Securitizations
incentivizing permanent
takeouts and enabling
lenders to recycle capital.
16 NEWMARK
$2.3 Billion
Hyperscale
BTS
Abilene, TX
$900 Million
Hyperscale
BTS
Atlanta, GA
$600 Million
AI GPU
Provider BTS
Richmond, VA
$500 Million
Delayed Draw
Construction Facility
Various, US
$475 Million
Permanent
Financing
Northeast, US
$450 Million
(CAD) AI GPU
Provider BTS
Toronto, ON
$200 Million
Powered Land &
Pre-Dev Facility
Various, US
$150 Million
Permanent
Financing
Various, US
SELECT DATA CENTER SECTOR FINANCING TRANSACTIONSSELECT DATA CENTER SECTOR FINANCING TRANSACTIONS
Development Financing
The nexus of all data center financing activity continues to be on both investment grade and credit-tenant anchored build-to-suit
development opportunities. There was record setting transaction volume in 2024, a pace that is not slowing down with over $25
billion of development financings underwritten for Q1 2025. Executions for each transaction comes with a number of nuances
including Sponsorship, timeline to close, and most importantly customer credit. For IG hyperscale build-to-suits there is a very deep
pool of lenders led by Asian and European project finance banks who utilize cheaper costs of capital and highly ecient syndication
markets to maximize pricing power. This pricing power has resulted in numerous mandates for these lenders. With that said, towards
the end of 2024, there was a noticeable movement from more traditional CRE investment banks who situationally brought pricing
in-line with their project finance bank counterparts. Although more situational, in these instances these lenders can come with very
appealing executions given their ability to hold larger quantums which can reduce structuring costs and minimize execution risk.
Over the past 12 months non-IG BTS executions have become a major focal point across the market. With AI customers such as
CoreWeave, Lambda Labs, OpenAI, and Crusoe rapidly expanding capacity requirements, there has been a material increase in
developers/operators signing long-term lease agreements with these credits. The result of this has been a substantial demand for
debt capital in order to fully capitalize the developments. There still is a limited bank liquidity for deals backed by non-IG customers,
especially at leverage levels where IG transactions execute. This perceived gap in the market is where unique pockets of capital can
be found to eectuate these developments. From both real estate and infrastructure private credit to alternative insurance capital,
there has been a meaningful increase in appetite to lend on these opportunities. Despite these lenders not generally being able to
compete on pricing on IG opportunities, they are often able to provide flexible and borrower friendly structures and commit to larger
underwritten amounts. Although this may not fully bridge the gap in economics compared to where banks would be on a 75%-80%
LTC ask, it does provide groups more accretive executions based on the respective business plans, as well as providing higher
certainty of executions which is extremely meaningful for more nascent Sponsorships or less familiar credits.
DEVELOPMENT FINANCING - HYPERSCALEDEVELOPMENT FINANCING - HYPERSCALE
LenderLender Te rmTe r m
Fixed/ Fixed/
FloatingFloating CharacteristicsCharacteristics
Key Sizing Key Sizing
ParametersParameters
Cash Flow Cash Flow
ConstraintsConstraints
Leverage Leverage
ConstraintsConstraints SpreadsSpreads
Project Finance Banks 3-5 Yrs Float Max Leverage DSCR / LTC DSCR: Sculpted -
1.05x - 1.25x 90% LTC 2.25 - 2.75%
CRE Banks 3-5 Yrs Float Moderate Leverage DY / LTC DY: 9%+ 80% LTC 2.50 - 3.50%
Private Credit 3-5 Yrs Float Max Leverage DY / EBITDA
Mult. / LTC
DY: 8-9%
EBITDA Mult.: 11x-12x 75-90% LTC 3.00 - 4.00%
172025 United States Data Center Market Outlook
DEVELOPMENT FINANCING - NON-IGDEVELOPMENT FINANCING - NON-IG
LenderLender Te rmTe r m
Fixed/ Fixed/
FloatingFloating CharacteristicsCharacteristics
Key Sizing Key Sizing
ParametersParameters
Cash Flow Cash Flow
ConstraintsConstraints
Leverage Leverage
ConstraintsConstraints SpreadsSpreads
Project Finance Banks* 3-5 Yrs Float Moderate Leverage DSCR / LTC DSCR: 1.25x+ 70-80% LTC 3.25 - 3.50%
CRE Banks* 3-5 Yrs Float Moderate Leverage DY / LTC DY: 10%+ 65-75% LTC 3.50 - 4.00%
Private Credit 3-5 Yrs Float Max Leverage DY / EBITDA
Mult. / LTC
DY: 9-10%
EBITDA Mult.: 11x-12x 70-85% LTC 4.00 - 5.00%
Permanent Financing
With the cost of development for data center projects increasing rapidly to keep up with the capacity requirements of hyperscale
and AI customers alike, there is a tremendous need for debt capital to support both the acquisition financing and refinancing of
these large scale developments. As a result, expect CMBS and ABS to continue to play an important role in supporting the access
to accretive and ecient permanent debt financing at scale. Both ABS and CMBS are forms of securitized credit, which involve a
stream of cash flows from underlying secured collateral. The major dierences reside in structuring.
ABS transactions are structured as master trusts so the issuer can add additional collateral in the form of data center or MW capacity,
to the trust and issue additional note series over time. The focus is primarily on debt service coverage ratio performance triggers
with sizing being a result of duration and credit of contractual cash flows. In the CMBS market the collateral pool is established at
issuance. Focus is on collateral value, which incorporates cash flow analysis but also includes emphasis on basis and residual values.
Given these variances, CMBS deals have priced tighter than ABS deals due in part to dierences in ratings and structures.
BROADER CMBS SASB PRICING GUIDELINES
SASB FixedSASB Fixed SASB FloatingSASB Floating
Loan-To-Value: 50 - 55% 60 - 65% 50 - 55% 60 - 65%
Term: 5-10 Years 5-10 Years 5-10 Years 5-10 Years
Total Spread: 1.25% - 1.50% 1.50% - 1.75% 1.40% - 1.60% 1.60% - 1.85%
AAA Spread: 1.00% - 1.25% 1.00% - 1.25% 1.20% - 1.35% 1.20% - 1.35%
Debt Yield: 10%+ 8.5%+ 10%+ 8.5%+
Permanent Financing Trends to Watch
Private Placement: With the size of mandates continuing to scale, there are new and creative structures for borrowers to access
once they complete development. One structure anticipated to gain momentum in 2025 and beyond is a debt private placement.
These debt placements are a form of financing where institutional investors, primarily insurance companies, pension funds, and
asset managers, purchase securities directly from the issuer rather than through a public oering. The private placement debt market
allows companies to raise capital without the need to meet the legal and financial requirements of the public market which can be
very attractive given the evolving customer credit and regulatory markets. This execution can allow for a more bespoke financing,
which can result in lower upfront fees, more leverage compared to a typical bank loan, increased flexibility, and certainty of execution
and pricing compared to public bond issuances. Although not an everyday occurrence, this market will likely gain considerable market
share in the coming years and is certainly a trend to watch.
Outlook
Without a doubt, 2025 is poised for greater volumes of
data center development and capital investment. Yet, in
this “new to us” structural boom in data center demand,
concerns circle around the vast amounts of spending and
energy needs required to support this growth.
The revelation of DeepSeek’s cost-ecient AI model in
January 2025 heightened discussions around falling behind
in an international AI race. Hyperscalers are not likely to
slow their capex plans in light of this cheaper model, but as
Meta noted on a January 2025 earnings call, one company’s
advances end up benefitting others and ”…investing very
heavily in CapEx and infra(structure) is going to be a
strategic advantage over time5.”
Even with current development at all-time highs,
hyperscalers are concerned about the ability to construct
enough infrastructure to support the existing AI use-case
environment, let alone anticipated future innovation and
adoption. AI is advancing so swiftly that performance
evaluation benchmarks are becoming obsolete, creating
challenges in measuring progress. Meanwhile, innovations
in training techniques and components are reducing training
times and cost, as seen with the DeepSeek breakthrough,
potentially shortening timelines to inference use cases
and leading to quicker monetization of AI than previously
forecasted.
These rapid transformations raise other important questions
regarding infrastructure lifecycle and pace of development
growth, if companies decide to scale back as the AI market
matures. How the new crop of AI training data centers,
particularly in tertiary markets, adapt in the AI evolution is an
answer unfolding in real time. Although those new facilities’
greatest asset for adaptability may be their access to large
volumes of power, a resource in increasingly limited supply.
18
19
Policies enacted by the new U.S. administration stand to
influence the data center landscape this year and beyond.
The current administration is clearly supportive of bolstering
energy resources and building out AI infrastructure (Project
Stargate and similar announcements); on the other hand,
policies around immigration have the potential to hamper
timely development, with the construction industry already
running at a nearly 500,000 job deficit . Data center
development will be competing for a finite skilled labor
pool with other major construction projects – including
manufacturing facilities and infrastructure projects.
New regions will continue to emerge in the data center
landscape in the ever-important search for power. Many
of the new regions such as West Texas being explored
for data center potential have been passed over by
traditional industrial site selection projects in recent years,
in some cases trapping them in a cycle of infrastructure
underinvestment that compounds the lack of interest from
industrial companies. Hyperscaler pursuit of locations
in these regions open up a number of new possibilities,
investing in power infrastructure on land that had previously
been underserved, and turning nearby underpowered
agricultural land into viable industrial sites. However, it won’t
all be smooth sailing – there is often a learning curve for
utilities that have not provided a huge amount of power to a
single user, including internal policies regarding service and
rates that will need to be adjusted. The complexity of these
requirements, regardless of location, demands extensive
due diligence from all stakeholders – a critical step that, if
managed well, could help regions capitalize on this unique
generational opportunity in data center demand.
5 Yahoo Finance, January 2025
6 Associated Builders and Contractors
Newmark has implemented a proprietary database and our tracking methodology has been revised. With this expansion and refinement in our data, there may be adjustments in historical statistics
including availability, asking rents, absorption and eective rents. Newmark Research Reports are available at nmrk.com/insights
All information contained in this publication is derived from sources that are deemed to be reliable. However, Newmark has not verified any such information, and the same constitutes the statements
and representations only of the source thereof, and not of Newmark. Any recipient of this publication should independently verify such information and all other information that may be material
to any decision that recipient may make in response to this publication, and should consult with professionals of the recipient’s choice with regard to all aspects of that decision, including its legal,
financial and tax aspects and implications. Any recipient of this publication may not, without the prior written approval of Newmark, distribute, disseminate, publish, transmit, copy, broadcast, upload,
download or in any other way reproduce this publication or any of the information it contains. This document is intended for informational purposes only and none of the content is intended to advise
or otherwise recommend a specific strategy. It is not to be relied upon in any way to predict market movement, investment in securities, transactions, investment strategies or any other matter.
Research Contacts:
Lisa DeNight
Managing Director,
National Industrial Research
lisa.denight@nmrk.com
Joseph Biasi
Head of Commercial
Capital Markets Research
joseph.biasi@nmrk.com
Kevin Nesburg
Head of GIS
kevin.nesburg@nmrk.com
David Bitner
Executive Managing Director,
Global Head of Research
david.bitner@nmrk.com
nmrk.com
Business Contacts:
Brent Mayo
Head of Data Center Capital Markets
brent.mayo@nmrk.com
Clint Frease
Vice Chairman
clint.frease@nmrk.com
