The Data Center Is Having Its IPO Moment — and the Numbers Are Staggering
The AI economy has a physical layer, and that physical layer needs to be financed. Data centers — the buildings full of servers, cooling systems, fiber optic cables, and power distribution equipment that make cloud computing and AI inference physically possible — have become the most strategically important infrastructure asset class on the planet. And in 2026, they are going public at a scale that is beginning to rival the technology companies that depend on them.
The flagship deal is Switch, the Las Vegas-headquartered data center operator that developed a reputation for building some of the most energy-efficient and densely provisioned facilities in the industry. Switch is targeting a valuation of approximately $80 billion in a listing that would make it one of the largest US infrastructure IPOs ever recorded — larger than most utility companies, larger than most REITs, and comparable in scale to mid-tier semiconductor manufacturers.
Switch is not alone. A cohort of data center operators, AI infrastructure financing vehicles, and hyperscale facility developers are moving toward public markets simultaneously, creating a wave of AI infrastructure capital formation that is happening largely beneath the surface of the model-focused AI narrative.
Why Data Center Operators Are Going Public Now
The timing of the data center IPO wave is not coincidental. It reflects the convergence of three forces that have made public market capital both available and necessary for infrastructure operators.
First, demand has become structurally predictable. For most of the 2010s, data center demand was tied to enterprise cloud migration — a lumpy, multi-year transition that was difficult to forecast at the facility level. AI inference demand is different. The hyperscalers — Microsoft Azure, Amazon Web Services, Google Cloud — have made multi-year commitments to AI compute procurement that give data center operators visibility into future occupancy rates that is unusual in capital-intensive industries. That predictability makes the business model legible to public market investors in a way it was not before.
Second, the capital requirement has outgrown private financing. Building a hyperscale AI data center is not a $50 million project. A single campus purpose-built for high-density GPU compute — with the power substations, cooling infrastructure, fiber connectivity, and physical security that enterprise customers require — can cost $500 million to $2 billion, and the demand pipeline justifies building dozens of them simultaneously. Private equity and sovereign wealth funds have been willing to write large checks for data center development, but the scale of the current build cycle has made public equity markets a necessary additional source of capital.
Third, the asset class has proven its resilience. Data centers emerged from the COVID pandemic and the subsequent interest rate cycle in better financial shape than almost any other commercial real estate category. Occupancy rates stayed high, lease durations are typically 10–20 years, and the tenant base — dominated by hyperscalers with investment-grade credit — means default risk is negligible. That combination of growth, stability, and credit quality is exactly what institutional equity investors in infrastructure seek.
Switch's Competitive Position
Switch's path to an $80 billion valuation rests on a differentiated competitive position that it has built over two decades. The company's PRIME campus in Las Vegas and its newer facilities in Reno, Atlanta, and Grand Rapids were designed from the ground up for extreme power density — capable of supporting the 30–60 kW per rack power loads that GPU-intensive AI workloads require, compared to the 5–10 kW per rack that was the industry standard when most enterprise data centers were built.
That power density advantage matters enormously in the AI era. A data center that cannot accommodate GPU density is essentially stranded from the most valuable workloads in the market. Switch's legacy facilities are GPU-ready; many older colocation facilities are not, and retrofitting them is expensive and technically constrained by building geometry and power infrastructure.
Switch has also been strategic about geography, locating facilities near major renewable energy sources — Nevada's solar resources, Michigan's hydroelectric access — in anticipation of the energy-intensive nature of AI compute. As data center power consumption has become a political and regulatory issue, operators with credible clean energy stories command premium pricing from hyperscaler tenants with net-zero commitments.
The Broader IPO Wave: What Else Is Coming
Switch is the most prominent name in the current data center IPO cohort, but it is not the only one. Several other infrastructure operators and financing vehicles are in various stages of public market preparation:
- Multiple AI infrastructure REITs — structured to hold data center real estate and pass through income to shareholders — have been formed in the past 18 months and are exploring public listings. The REIT structure is attractive for capital intensive infrastructure because it allows favorable tax treatment of income distributions.
- Several hyperscale development companies — firms that build data center capacity on behalf of hyperscaler tenants under long-term contracts before handing over operations — are weighing whether to access public capital markets or continue to operate through private equity-backed structures.
- In Europe, where AI infrastructure buildout is running significantly behind North America and Asia, there is active government interest in creating publicly financed data center development vehicles — a form of infrastructure policy that has precedents in broadband deployment and rail infrastructure.
The Valuation Question: Is $80 Billion for Switch Reasonable?
At $80 billion, Switch would trade at a multiple of its EBITDA that is toward the premium end of the data center sector but not unprecedented for operators with strong growth profiles. For comparison, Equinix — the publicly traded data center operator that is the most direct comparable — has historically traded at 20–30× EBITDA. Digital Realty, another publicly traded peer, trades at a somewhat lower multiple reflecting its more diversified, less GPU-intensive tenant mix.
The bull case for a premium valuation on Switch rests on GPU density specialization and long-duration hyperscaler leases that provide cash flow visibility unusual for most growth-stage infrastructure operators. The bear case centers on concentration risk: Switch's largest tenants represent a significant fraction of its revenue, and if a major hyperscaler decided to pursue a more aggressive in-house data center strategy, the demand signal could shift.
This mirrors the dynamics investors have been watching in semiconductor infrastructure listings like SK Hynix's $28 billion Nasdaq debut — where the asset class premium is real, but concentration in AI-dependent demand creates tail risk that standard infrastructure valuation models were not built to assess.
What the Data Center IPO Wave Means for AI Investors
The emergence of publicly traded AI infrastructure plays creates an interesting portfolio construction option for investors who want exposure to the AI economy without betting on specific model winners. The "picks and shovels" argument — that the companies supplying infrastructure to a gold rush often outperform the prospectors themselves — has a long history in technology investing, from Cisco's rise during the internet era to TSMC's semiconductor dominance during the smartphone era.
Data center operators occupy a structurally attractive position in the AI value chain: they are upstream of every model company, every cloud provider, and every AI application — and they collect rent regardless of which model or platform wins market share at the application layer. Their exposure to the AI economy is diversified in a way that individual model companies are not.
As TSMC's 36% revenue surge and Korea's semiconductor volatility have both demonstrated, the AI infrastructure trade can generate extraordinary returns — but it can also be volatile when AI spending cycles contract. Data center operators, with their long-duration leases and high switching costs, may offer a more stable expression of the same underlying thesis.
Switch's IPO will be the first major test of whether public markets will value that stability at a premium. The answer will shape how the rest of the data center IPO wave is priced — and how institutional investors think about the physical infrastructure layer of the AI economy for years to come.
