In the US, data centers surpassed offices in construction spending for the first time in 2025

Spending on data center construction in the USA by the end of 2025 has for the first time exceeded investments in office buildings. This is not just unusual commercial real estate statistics, but a sign that cloud and artificial intelligence infrastructure has become one of the main directions of capital investment.

Market Turning Point

For a long time, office towers were a symbol of commercial real estate growth: large amounts of money went into them, quarters were built around them, and demand seemed almost guaranteed. Now the balance is changing. If investors and developers previously focused on employee workplaces, they are now increasingly betting on computing power, cooling, power supply, and network connectivity. The market is literally being rebuilt for the digital economy, where the key asset is not square meters for desks, but server racks.

This shift is important not only for developers. When data center spending outpaces offices, it shows where demand is shifting from corporations, technology platforms, and financial markets. Business needs capacity for data storage, launching cloud services, training models, and real-time AI inference. Against this backdrop, the traditional office is no longer an absolute priority: companies are reducing floor space, rethinking their presence format, and being more cautious about entering new construction projects.

Why Data Centers Are Growing

The main growth driver is the simultaneous expansion of cloud infrastructure and the AI boom. The more actively companies deploy generative models, corporate assistants, internal data search, and AI services for customers, the more computing resources they need. This applies not only to big tech. Banks, retail, logistics, media, and industry are also increasing cloud consumption, which means they indirectly support demand for new data centers.

• Growth in cloud services for business and consumers
• Demand for training and deploying AI models
• Need for new data storage and processing capacity
• Investments in facilities near cheap energy and network infrastructure

A separate factor is the logic of building such facilities itself. A modern data center is no longer just a technical room, but a complex engineering facility with high demands on power grids, backup systems, and industrial cooling. Therefore, the increase in spending reflects not only the number of projects, but also their complexity. The denser server clusters become and the higher the AI workload requirements are, the more expensive each new facility becomes. As a result, capital expenditures scale rapidly even with a limited number of major projects.

What's Pressuring Offices

The office segment, on the other hand, is experiencing a prolonged correction. After many companies transitioned to remote and hybrid work, the need for large headquarters declined. Business is optimizing floor space, rethinking long-term lease agreements, and increasingly questioning how many workplaces a team really needs if some employees are in the office only two or three times a week. For new projects, this means less predictable demand and more cautious investment decisions.

There is a broader effect as well. Offices remain a necessary format for some industries, but they no longer look like the universal bet they once were. Developers have to think about repurposing, mixed-use buildings, or projects in niches with clearer returns. Against this backdrop, data centers look more attractive: they have long-term demand from cloud providers and companies that are building AI infrastructure as the foundational layer of their business.

What This Means

For the technology market, this is a signal that AI is changing not just software, but the physical economy. Money is going into server capacity, energy, and engineering infrastructure, not additional office floors. If the trend continues, competition will shift not only over talent and models, but also over land, electricity, and the ability to quickly build new data centers.