Three years. $11 billion. Now $50 billion on the table.
Switch, the Las Vegas-headquartered data center operator that DigitalBridge's consortium took private in 2022, is reportedly targeting a near-$50 billion valuation in its upcoming IPO. Same buildings. Same power meters. Same concrete floors that didn't care about "AI" in 2022 and still don't today. The only thing that changed is the story.
The backdoor was open, but the key was volatility.
I've been trading through narrative shifts long enough to recognize the shape of this setup. A physical asset — boring, necessary, revenue-predictable — gets relabeled as "AI infrastructure," and suddenly the market is asked to pay growth multiples for something that behaves like a REIT with extra steps. The label isn't a technology upgrade. It's a pricing upgrade.
The mechanics first. Switch is not an AI company. It's a multi-tenant colocation provider — a landlord that secures land, locks down power capacity, and leases finished data center space to hyperscalers and AI labs. It doesn't train models. It doesn't write code. Its facilities span Nevada, Michigan, Georgia, and Texas, with the Nevada footprint benefiting from renewable generation access and some of the more competitive electricity rates in the American West. The business model runs on long-term wholesale contracts, typically stretching five to fifteen years. Revenue visibility is high. Optionality is low. It's the AI trade's "sell water to gold miners" play.
The round-trip: 2022 private acquisition at roughly $11 billion including debt. 2025 IPO target near $50 billion. That's a 4.5x enterprise value expansion in about three years — a move that can't be explained by organic growth alone. It's a narrative compounding. Same playbook as every private equity cycle: buy at the bottom of one story, dress the asset up, sell at the top of the next one.
The sell-side lineup is maximum institutional conviction. Bank of America, Citi, Goldman Sachs, JPMorgan, and Morgan Stanley are all underwriting. Ben Horowitz, a16z's most recognizable partner, is taking a board seat and leading what appears to be a fresh funding round ahead of the listing. Confidential paperwork filed in August. The street expects a November debut.
November is not a coincidence. That calendar slot clears the U.S. election, lands ahead of year-end allocation flows, and sits in the window when institutional CIOs deploy fresh budgets. The positioning is surgical. That's what five banks and one celebrity venture capitalist actually buy you: timing discipline.
This is not an isolated deal either. Since the start of the year, a wave of data center operators and related equipment and service companies have tapped the U.S. IPO market. Switch sits at the top of that wave — a signal that the AI compute story has officially moved from private ledgers to public pricing.
Now let's pressure-test the $50 billion.
At that enterprise value, Switch would need somewhere in the range of $4-6 billion in revenue to justify itself as a growth asset. At the sector's standard 35-45% EBITDA margins, that puts EBITDA around $1.5-2.5 billion — call it $2 billion in the center. The implied EV/EBITDA lands near 25x.
Here's the comp check: Equinix and Digital Realty, the two largest public data center operators, trade in the 12-20x band. Traditional data center assets get utility-like multiples because their growth is visible, slow, and boring. Switch is asking the public market to pay a 30-40% premium to that band, purely on the AI label and the quality of its tenant list. Twice the multiple of your comps, for an asset class whose competitive advantage is power access, not technology. That's the puzzle at the heart of this deal.
Let me count what's real. Transformer lead times now exceed two years. Grid interconnection windows in many U.S. markets stretch past planning horizons. Approved land with contracted power is genuinely scarce, and Switch's geographic portfolio represents a real diversification of political and grid risk. That's structural scarcity — the kind of moat you can see in aerial photos: empty land, substations, and cooling infrastructure waiting for tenants.
Now the signals. Five banks underwriting means the S-1 backlog numbers have survived forensic scrutiny. Construction-in-progress, signed capacity, EBITDA margin trajectory — the diligence at this tier is brutal. And Ben Horowitz's board seat isn't just money. It's a deal-flow engine. Every portfolio company in a16z's AI network becomes a warm lead for Switch's leasing desk. The a16z portfolio is a customer acquisition channel disguised as an investment relation.
This is the moment where my own history starts screaming at me. In 2017, EOS was "the future of everything" at $10. I went in with $15,000 of savings, conviction thick and diligence thin, and watched my portfolio bleed 70% when the music stopped. That lesson — hype is not utility — is the most expensive tuition I've ever paid. And this deal, for all its institutional polish, is asking the same question: are you paying for the building, or for the story around it?
The institutional convergence angle is impossible to miss. Post-ETF, I moved a significant chunk of my own exposure from DeFi yield chasing into regulated custodial vehicles. Same convergence, new asset class: the infrastructure stays physical, but the pricing becomes a public market function. That's exactly where DePIN and tokenized compute projects want to go — and Switch's IPO just handed them a valuation anchor. If a concrete landlord in Nevada can hold a 25x multiple, every tokenized data center project suddenly has a comp that justifies the same narrative premium. The blockchain layer is watching this deal as closely as the REIT layer.
The contrarian read isn't that Switch is a bad business. It's that the market is drawing the wrong conclusion from the right data.
AI data center operators have terrible customer diversification by design. One hyperscaler or top-tier AI lab can take 20-30% of a facility's capacity on a single agreement. Those contracts make the income statement look pristine and the negotiating position fragile. The contract is law, but the whale is truth. When one tenant holds a third of your revenue, you don't set prices. You set the table.
Then there's the self-build problem. Microsoft, Google, and Amazon are simultaneously Switch's best customers and its most dangerous competitors. They rent facilities where it's cheaper and build where it's strategic. That's not a moat — that's a lease where a competitor owns the building next door.
The balance sheet compounds the question. The 2022 buyout loaded Switch with debt, and the data center CAPEX treadmill never stops. Net debt isn't disclosed yet, but the gap between enterprise value and equity value will be visible the moment the S-1 drops. If the real leverage sits in the $20-25 billion range, the "AI pure-play" equity story is half-consumed by carrying costs before the first institutional buyer enters.
The ESG angle cuts both ways too. Nevada's renewable adjacency is a great pitch-deck story — right up until a drought year degrades hydro output. The green halo that gets a stock into ESG mandates is the same flashlight that exposes weak spots in an earnings call. Every narrative has a mirror, and this one reflects a highly leveraged, customer-concentrated real estate vehicle wearing an AI costume.
The S-1 is the only document that matters in the next thirty days. Read it for customer concentration, net debt, and the megawatt backlog. Everything else is theater.
Honor the 25x entrance premium. Size the position as a trade, not a conviction. Greed has a timer, and it always expires — and the AI infrastructure narrative is now on the clock in the most public way possible.
The question that should keep every AI-bull awake: when the capex cycle turns — and it always turns — who wants the building next when the tenants stop growing?
