Investment Strategy

Cutting Through the Noise: The Long-Term Case for Data Centers

May 1, 2025

By Mike Forman

These days, headlines about data centers are everywhere—DeepSeek, Stargate, and reports of leasing slowdowns, to name a few.

For investors, it’s essential to look beyond short-term volatility and focus on long-term fundamentals. I believe digitalization of the economy, including the AI revolution, is a megatrend that supports strong long-term demand for data centers.

SECULAR DEMAND GROWTH

For the past 30 years, the economy has been rapidly digitalizing, and we’re still in the early innings.

The share of GDP attributed to the information technology sector has more than tripled over the last 25 years but remains at just 8%. [ 1 ] As technology usage grows, the increased computing and storage requirements fuel demand for data centers. Said differently, the cloud does not live in the clouds! We’ve seen a long trend of robust growth in data center demand as cloud computing and content creation have boomed, and, more recently, as AI has begun to take off.

Digitalization is driving the economy [ 2 ]

Data Created and Stored
(Zettabytes)

Fig1_Data_Created_and_Stored_v1



IT Sector Contribution to Economy
(% of GDP)

Fig2_IT_Sector_Contribution_to_Economy

AI adoption is still in its infancy. Early adopters of AI, such as software engineers, showcase its potential — over 25% of all new code at Google is now generated by AI [ 3 ] and engineers using AI are 26% more productive. [ 4 ] Yet more broadly, AI has barely begun to be integrated into our lives, the technology we interact with, and the economy. Today, just 1-5% of work hours are assisted by AI. [ 5 ] This is changing rapidly, particularly as AI models become more capable and reliable. For example, this year, we are beginning to see the first examples of AI “agents,” which upgrade capabilities from do-it-yourself tools (e.g., generating vacation itineraries) to white-glove services (e.g., actually booking your trip).

The potential for AI is enormous. I expect it will become embedded into everything — enterprise processes, scientific discovery, medical advances, manufacturing, and robotics. Each of these use cases will require significant increases in computing power compared to today.

Stay in the loop with Blackstone insights

COMPUTING EFFICIENCY GAINS

AI is not only becoming more powerful but also more efficient, as illustrated by DeepSeek earlier this year.

For example, the cost to access equivalent-intelligence AI models has declined by 99% over the past two years. [ 6 ] These efficiency gains have introduced volatility in public markets, but they also highlight the rapid pace of technological advancement.
 
Counterintuitively, efficiency gains have been the propellant of digitalization. Compute has been getting exponentially more efficient in terms of both cost and energy — computations per dollar of IT investment and per watt have been doubling every two to three years for many decades. [ 7 ] On top of this, innovation in software has doubled the number of tasks which can be completed per unit of compute every year. [ 8 ] Despite these gains, IT spend and compute power usage have not gone down, but rather have increased dramatically, as these efficiency gains create new capabilities and unlock new use cases.

Efficiency gains have spurred massive investment in computing [ 9 ]

Compute Efficiency
(Billions of Calculations per $)

Fig3_Compute_Efficiency_v1



Hyperscaler Capex
($ Billions)

Fig4_Hyperscaler_Capex_v1

We are seeing a similar trend within the AI space. As capabilities improve and costs come down, usage is exploding — usage of leading AI platforms is up 4x year-over-year, and OpenAI now has over 600 million monthly active users. [ 10 ] Revenues are also growing rapidly, with 2025F revenues for AI model providers up 16x since 2023 and 3x year-over-year to $14 billion.
 
This phenomenon, known as Jevon’s Paradox, explains why efficiency gains often lead to greater resource consumption. It’s a dynamic we’ve seen throughout history, from lightbulbs to transportation. Digitalization is the megatrend of our lifetime.

AI revenues have increased rapidly as cost to access models has declined [ 11 ]

Cost to Use AI Models
($ per million tokens)

Fig5_Cost_to_Use_AI_Model_v1



AI Model Revenue
($ Billions)

Fig6_AI_Model_Revenue_v1

Downside Protected Growth Opportunity

Technological innovation will continue at a rapid pace, and it remains to be seen where exactly value will ultimately accrue in the AI ecosystem.

However, with data centers, we are investing in the “picks and shovels” of the digital economy, and are well positioned if we believe that long-term demand for compute is going up.

Capital markets volatility may lead to periods of slower growth, and we have seen this happen before. Our current leasing pipeline is as large as it’s ever been, but we continue to be disciplined in our approach and structure our investments to minimize downside in the event demand slows. For example, we typically only build or lend to data centers after securing pre-leases with investment-grade counterparties on long-duration contracts. This means we are protected against short-term shifts in demand and can play the long game.

Supply Constraints: A Competitive Advantage

As data center owners, we benefit from significant barriers to new supply.

Vacancy rates across the U.S. are below 2%, [ 12 ] tighter than any other sector we invest in. This does not appear to be changing anytime soon — there is almost no speculative construction in the industry given the capital intensity of each project. As we are seeing real time in our businesses, it’s becoming harder every day to build new data centers.

Data center vacancy is lower than other asset classes [ 13 ]

Fig7_Data_Center_Vacancy_v1

Power is a major bottleneck to delivering new capacity. After 25 years of flat electricity demand, the US is experiencing a surge in demand from data centers, electric vehicles, electrified heating systems, and re-shoring of manufacturing. [ 14 ] The power grid has not been able to keep up, and wait times for new grid connections have increased to 7-10 years in key data center markets. [ 15 ] Blackstone is investing heavily in new generation and transmission infrastructure, but upgrading and modernizing the grid takes decades—not months or years. We have been able to procure power at scale by leveraging our 20+ years of experience investing in energy markets and the highly capable teams at our portfolio companies, but most new entrants to the space lack this expertise.
 
Beyond power, developers continue to experience supply chain constraints, labor shortages, and increased regulatory and permitting challenges. Amidst this backdrop, operational execution is crucial, and we are wary of the many “fly-by-night” operators who seem to have emerged overnight. When investors ask us what we worry about the most, operational execution is at the top of the list. We are focused on investing behind best-in-class platforms, including our portfolio companies QTS, AirTrunk and Lumina, and our joint venture partner Digital Realty. Many of these platforms have proven track records spanning multiple decades.
 
Capital has also become a more important differentiator, as these campuses often require billions of dollars to build. Blackstone’s financial resources combined with high-quality operators is a winning formula.

WHAT’S NEXT FOR BLACKSTONE?

Today, Blackstone manages an $85 billion global data center platform [ 16 ] with a powered land bank that can support over $125 billion of future growth.

We have built the largest and fastest growing data center business in the world in partnership with incredibly talented teams at our portfolio companies. For years now, we have been planting the seeds for future growth and are positioning ourselves for continued leadership over the long-term.
 
Beyond our equity investments, we have also built a leading digital infrastructure lending platform, and are focused on investing at scale across the entire capital stack. Our conviction and access to capital enable us to unlock unique opportunities and provide differentiated solutions across the digital infrastructure ecosystem.

Stay up-to-date

Sign up for our latest insights and firm announcements.

Please fill out this field.
Please fill out this field.
Please fill out this field.
Please fill out this field.
Please fill out this field.

You have successfully submitted your information

Bureau of Economic Analysis, as of March 2025.
IDC, as of May 2024; Bureau of Economic Analysis, as of March 2025.
Google Q3 2024 Earnings Call, as of October 2024.
Cui, Zheyuan and Demirer, Mert and Jaffe, Sonia and Musolff, Leon and Peng, Sida and Salz, Tobias, The Effects of Generative AI on High-Skilled Work: Evidence from Three Field Experiments with Software Developers (February 10, 2025). Available at SSRN: https://ssrn.com/abstract=4945566 or http://dx.doi.org/10.2139/ssrn.4945566.
Federal Reserve Bank of St. Louis, as of February 2025.
OpenAI, Wayback Machine, as of April 2025.
Greenmantle, as of July 2024.
Danny Hernandez, Tom B. Brown, Measuring the Algorithmic Efficiency of Neural Networks. Available at: https://doi.org/10.48550/arXiv.2005.04305.
Hobbahn et al. “Trends in Machine Learning Hardware”, Epoch AI, Morgan Stanley Equity Research, Public Reporting as of July 2024.
Similarweb, Reuters, as of April 2025.
OpenAI, Wayback Machine, as of April 2025. Revenue data includes OpenAI, Anthropic, Character.ai based on availability (comprises 90% of total web visits for LLMs). CNBC, New York Times, The Information as of Q4’24. “$ per million tokens” represents the cost to access one million tokens through OpenAI’s API.
DCH, as of December 2024.
As of December 31, 2024. Office, retail: CoStar. Retail reflects all retail. Logistics: CBRE. Multifamily: RealPage Market Analytics. Data centers: datacenterHawk.
GridStrategies, as of December 2024.
Blackstone proprietary data.
Represents total enterprise value, including future buildout of pre-leased capacity.

Important Disclosures

This commentary does not constitute an offer to sell any securities or the solicitation of an offer to purchase any securities. This commentary discusses broad market, industry or sector trends, or other general economic, market or political conditions and has not been provided in a fiduciary capacity under ERISA and should not be construed as research, investment advice, or any investment recommendation. Past performance does not predict future returns.

The views expressed in this commentary are the personal views of the authors and do not necessarily reflect the views of Blackstone. The views expressed reflect the current views of the authors as of the date hereof, and neither the authors nor Blackstone undertake any responsibility to advise you of any changes in the views expressed herein.

Blackstone and others associated with it may have positions in and effect transactions in securities of companies mentioned or indirectly referenced in this commentary and may also perform or seek to perform services for those companies. Blackstone and others associated with it may also offer strategies to third parties for compensation within those asset classes mentioned or described in this commentary.
 
Investment concepts mentioned in this commentary may be unsuitable for investors depending on their specific investment objectives and financial position. Tax considerations, margin requirements, commissions and other transaction costs may significantly affect the economic consequences of any transaction concepts referenced in this commentary and should be reviewed carefully with one’s investment and tax advisors. All information in this commentary is believed to be reliable as of the date on which this commentary was issued and has been obtained from public sources believed to be reliable. No representation or warranty, either express or implied, is provided in relation to the accuracy or completeness of the information contained herein.