JLL 2025 Global Data Center Outlook

0.0(0)
Studied by 0 people
call kaiCall Kai
learnLearn
examPractice Test
spaced repetitionSpaced Repetition
heart puzzleMatch
flashcardsFlashcards
GameKnowt Play
Card Sorting

1/29

flashcard set

Earn XP

Description and Tags

Comprehensive practice flashcards covering market trends, power constraints, semiconductor advances, cooling technologies, nuclear SMRs, and capital market dynamics from the JLL 2025 Global Data Center Outlook.

30 Terms

1
New cards

What is the projected compound annual growth rate (CAGR) for the global data center market through 2027?

The baseline forecast is a 15%15\% CAGR through 2027, with upside potential reaching a 20%20\% CAGR.

2
New cards

What total value of data center assets will require development or permanent financing globally in 2025?

Roughly $170 billion\$170\text{ billion} in data center asset value will need to secure construction lending or permanent financing in 2025.

3
New cards

Which four major technology companies drive Big Tech capital expenditures predominantly focused on AI?

Alphabet, Amazon, Meta, and Microsoft.

4
New cards

What are the process node, transistor count, and rack power density of the NVIDIA H100 GPU?

It is built on a 7 nm7\,\text{nm} process node, features approximately 208 billion208\text{ billion} transistors, and has a rack density of 41 kW41\,\text{kW}.

5
New cards

What process node and expected rack density characterize NVIDIA's next-generation GB200 (Project Blackwell)?

It uses 5 nm5\,\text{nm} technology with an expected rack density of 130 kW130\,\text{kW}.

6
New cards

What rack density are future 2 nm2\,\text{nm} semiconductor chips projected to reach?

They are projected to reach an astounding 250 kW250\,\text{kW} per rack.

7
New cards

How has GPU advancement accelerated computation times for AI large language model training?

Computation tasks that previously took 32 hours32\text{ hours} to execute can now be completed in just 1 second1\text{ second} using the latest GPU technology.

8
New cards

What resource constraint do industry experts project AI developers will face within the next couple of years?

Technology companies are projected to run out of high-quality data for training AI language models.

9
New cards

How much electricity does a single 1 GW1\,\text{GW} AI training data center consume compared to U.S. households?

A 1 GW1\,\text{GW} facility consumes the same amount of electricity as 800,000 homes800{,}000\text{ homes} in the U.S. over the course of a year.

10
New cards

Why is there a geographic decoupling between AI training facilities and AI inference facilities?

AI training facilities require immense power (1 GW1\,\text{GW} or more) and are located near power sources, whereas inference facilities are positioned near population centers to ensure low latency for end users.

11
New cards

According to the IEA, what will be the global electricity demand growth rate in 2025, and what share will data centers represent?

Global electricity demand will grow by around 4%4\% in 2025, with data centers accounting for about 2%2\% of total global electricity consumption.

12
New cards

How does the projected growth in data center electricity demand through 2030 compare to electric vehicles and air conditioning?

The increase in data center electricity demand through 2030 is projected to be less than one-third of the increased electricity needed for both electric vehicles and air conditioning.

13
New cards

What proportions of total electricity demand do data centers represent in Singapore, Ireland, and Virginia?

Data centers account for around 7%7\% of total electricity consumption in Singapore, 21%21\% in Ireland, and approximately 26%26\% in Virginia.

14
New cards

What is the global data center energy demand projected to reach by 2029?

Global data center energy demand is forecast to double over the next five years to reach 100 GW100\,\text{GW}.

15
New cards

How long can power transmission line extensions delay new data center projects, and what is the primary cause?

They can delay developments by 4 years4\text{ years} or more, primarily due to securing easements and regulatory approvals rather than equipment procurement.

16
New cards

How has power infrastructure scarcity changed the primary site selection criteria for data centers?

Land is now evaluated primarily based on available power capacity and proximity to transmission lines, rather than pricing or total acreage.

17
New cards

What is the electrical power output range of small modular reactors (SMRs)?

SMRs provide between 1.5 to 300 MW1.5\text{ to }300\,\text{MW} of power.

18
New cards

When is commercial deployment of SMRs expected in the United States, and which developers have secured high-profile contracts?

Commercial deployment is not expected until 20302030 at the earliest; early contract winners include Oklo, Kairos, X-energy, and Holtec.

19
New cards
<p>What planned small modular reactor (SMR) powerhouse design is depicted in this architectural rendering?</p>

What planned small modular reactor (SMR) powerhouse design is depicted in this architectural rendering?

The Oklo Aurora powerhouse (designed by Gensler).

20
New cards

What phased geographic path will small modular reactor (SMR) deployments follow beginning in the early 2030s?

Initial deployments will occur at existing nuclear facilities, followed by decommissioned power plants or remote rural areas, with metropolitan deployments not expected for some time.

21
New cards

What is Power Usage Effectiveness (PUE) and how has the worldwide average changed from 2007 to 2024?

PUE is the ratio of total facility energy to IT equipment energy (where 1.01.0 is perfectly efficient); the worldwide average improved from 2.502.50 in 2007 to 1.651.65 in 2013, 1.581.58 in 2018, and 1.561.56 in 2024.

22
New cards

What typical hybrid cooling combination is utilized in high-density data centers today?

A combination of 70%70\% liquid cooling and 30%30\% air cooling.

23
New cards

What are the two primary forms of liquid cooling retrofits and new installations implemented today?

Rear door heat exchangers (RDHx) and direct-to-chip (DTC) cooling technologies.

24
New cards

At what rack density threshold does immersion cooling typically become necessary, and what is its current adoption rate?

Immersion cooling becomes common as GPU densities push above 150 kW150\,\text{kW} per rack; it is currently implemented in less than 10%10\% of data centers where average density is 12 kW12\,\text{kW}.

25
New cards

What major structural design challenge is introduced by large-scale immersion cooling tanks?

The largest cooling baths can weigh up to 4 metric tons4\text{ metric tons} when filled with fluid and IT equipment, requiring significantly reinforced flooring.

26
New cards

How much data center capacity is forecast to break ground and reach completion globally in 2025?

An estimated 10 GW10\,\text{GW} is projected to break ground globally, while 7 GW7\,\text{GW} is expected to reach completion.

27
New cards

What typical leverage metrics are used for data center development and permanent financing?

Development financing is typically arranged at 65% to 80%65\%\text{ to }80\% loan-to-cost, while permanent financing is arranged at 65% to 75%65\%\text{ to }75\% loan-to-value.

28
New cards

Why is the pace of data center megamergers expected to slow in favor of joint ventures in 2025?

Major operators have already established global platforms and prefer organic growth, bolt-on acquisitions, and regional joint ventures to navigate local regulatory and political landscapes without paying large M&A premiums.

29
New cards

How does the historical annual investment sales volume of data centers compare to office properties?

Data center investment sales have averaged $7 billion\$7\text{ billion} annually since 2020 (excluding entity trades and recaps), compared to $241 billion\$241\text{ billion} annually for office properties.

30
New cards

Why will existing data centers avoid becoming obsolete within 10 years despite rapidly advancing AI chip technology?

Internal IT hardware is upgraded continuously, traditional non-AI workloads will still represent roughly 50%50\% of capacity in 2030, and most long-term asset value lies in secured power connections and underlying infrastructure.