Are Large-Cap Companies Shifting to Data Centers?
Discover why large-cap companies are focusing on data centers and the key strategies driving this investment in 2023.
The math is forcing their hand. Large-cap companies that once treated computing infrastructure as a cost center—something to outsource, lease, or minimize—are now racing to own it outright. Servers, chips, and the data centers that house them have moved from line items to strategic assets, and the companies that figure this out fastest are pulling ahead.
This isn't a trend driven by preference; it's driven by dependency.
Why Data Center Investment Has Become Non-Negotiable
When your core product runs on AI inference, real-time analytics, or large-scale cloud services, compute capacity isn't just supporting the business—it *is* the business. Large-cap companies across finance, healthcare, logistics, and technology are discovering that relying on third-party infrastructure means handing control of their competitive advantage to someone else.
The demand side of this equation is relentless. Global data creation is projected to hit 120 zettabytes by 2023, and enterprise workloads are shifting from periodic batch processing to always-on, latency-sensitive operations. Every autonomous system, every personalized recommendation engine, and every fraud detection model requires physical infrastructure somewhere. You can't run a frontier AI model on goodwill and bandwidth alone.
What's changed in the last few years isn't the existence of data centers—it's the *urgency* of owning or controlling them. The semiconductor shortages of 2021-2022 were a wake-up call. Companies that had leaned entirely on third-party cloud providers found themselves throttled during peak demand, paying spot pricing that bore no resemblance to their financial projections. That lesson stuck.
The Financial Case for Building Rather Than Renting
On the surface, data center construction looks expensive—because it is. A hyperscale facility can run $1 billion or more to build, and that's before you factor in the ongoing power, cooling, and staffing costs. But the rent-versus-own calculus looks very different over a 15- or 20-year horizon.
Colocation and cloud contracts carry premium pricing baked in—the provider's margin, redundancy overhead, and the flexibility tax you pay for not committing to long-term capacity. For companies running consistent, predictable workloads at scale, that flexibility premium is essentially waste. Owning the infrastructure converts a variable operating expense into a depreciable capital asset—and for companies with access to low-cost capital, that's a meaningful financial restructuring, not just an operational one.
There's also the tax and depreciation angle that rarely gets discussed in mainstream coverage. Data center assets qualify for accelerated depreciation under current U.S. tax law, and companies structured to take advantage of this can significantly reduce their effective tax burden in the early years of a facility's life. Real estate investment trust (REIT) structures have made data centers an attractive vehicle for institutional capital as well—REITs like Equinix and Digital Realty have demonstrated that data center assets can generate stable, long-duration cash flows that look more like utility infrastructure than speculative tech bets.
For large-cap companies evaluating data center investment, the question isn't really "can we afford to build?" It's "can we afford to keep renting at scale?"
What Separates a Smart Data Center Build from an Expensive Mistake
Not all data center construction is created equal. The companies winning in this space are making deliberate choices about technology architecture and site selection—and the ones struggling are the ones that treated the facility like a real estate play rather than an engineering challenge.
Power Is the Constraint Everything Else Flows From
A modern hyperscale data center can consume 100 megawatts or more—enough to power a small city. Securing that power reliably, at a predictable cost, is often harder than building the facility itself. The most sophisticated operators are co-locating data centers with renewable energy generation, signing long-term power purchase agreements, or targeting regions with surplus grid capacity specifically to lock in energy costs that won't crater their operating margins a decade from now.
Texas, the Pacific Northwest, and the Carolinas have emerged as competitive data center markets partly because of their energy profiles. Proximity to renewable generation—wind in Texas, hydro in the Northwest—makes long-term power costs more predictable and increasingly helps companies hit sustainability commitments that their investors and regulators expect.
Cooling technology is the other lever. Traditional air cooling is hitting physical limits at the chip power densities that modern AI accelerators require. Liquid cooling—whether direct-to-chip or immersion-based—is moving from experimental to standard in new builds. Companies that design for liquid cooling from the ground up will have meaningfully lower PUE (Power Usage Effectiveness) ratios, which translates directly to operating cost advantages over the life of the facility.
Sustainability Isn't Just PR
ESG commitments are forcing this issue whether companies want it or not. Institutional investors increasingly require Scope 2 emissions disclosures, and a data center running on carbon-heavy grid power is a liability on that front. But beyond compliance, water usage is emerging as the next major pressure point—traditional cooling systems can consume millions of gallons annually, and in water-stressed regions, that's becoming a genuine operational and regulatory risk.
The Operational Reality Large-Cap Companies Underestimate
Building the facility is the part that gets announced in press releases. Operating it profitably over 20 years is where the real work lives.
Data center operations demand a workforce that sits at an unusual intersection of electrical engineering, network operations, physical security, and facilities management. This talent is scarce and expensive. Companies that have historically focused on software development or financial services are often caught off guard by the operational complexity—and the cost structure—of running critical physical infrastructure.
Redundancy design is another area where inexperience shows. The industry standard Tier III classification requires 99.982% uptime, which sounds impressive until you realize that 0.018% downtime is still roughly 1.6 hours per year—potentially catastrophic for certain applications. Getting to Tier IV (99.9999% uptime) requires fully redundant power and cooling with no single points of failure, which adds significant capital cost upfront. Getting that design decision wrong at the build stage means either over-engineering expensively or underbuilding and facing costly retrofits.
Cost management in data centers is also a continuous discipline, not a one-time optimization. Power costs alone can represent 40-60% of operating expenses. Companies that negotiate long-term energy contracts, invest in AI-driven cooling optimization, and regularly refresh hardware to improve performance-per-watt are compressing those costs systematically. Companies that treat operations as a fixed cost environment will find themselves increasingly uncompetitive.
Where This Goes From Here
The next phase of data center development isn't just more of the same at larger scale. Several structural shifts are worth watching.
Edge computing is distributing the architecture. Rather than consolidating everything into massive hyperscale campuses, companies are deploying smaller, purpose-built facilities closer to end users—reducing latency for applications that can't afford the round trip to a regional hub. This creates a tiered infrastructure model where large-cap companies may need both hyperscale capacity and a distributed edge footprint.
Nuclear power is entering the conversation seriously for the first time in decades. Microsoft's agreement to restart Three Mile Island's Unit 1 reactor to power its data centers is not a quirky headline—it's a signal that the power demands of AI-scale computing are pushing operators toward energy sources that can deliver consistent, carbon-free baseload power at scale that renewables alone can't guarantee.
Liquid cooling and custom silicon are converging. Companies like Google, Amazon, and Microsoft have already moved to designing custom AI chips optimized for their specific workloads. As large-cap companies in adjacent sectors build out their own data center capacity, the question of whether to use commodity hardware or invest in custom silicon will increasingly define the efficiency ceiling they can reach.
The companies positioning themselves well right now are treating data center investment as infrastructure strategy—long-duration, capital-intensive, and central to their competitive moat—rather than an IT procurement decision. That framing shift, more than any specific technology choice, is what separates the operators who will own this space from the ones still writing checks to someone else for the privilege of access.
Explore more about InfraSale Marketplace and how to optimize your data center investments.