The deal whispers started last night. TPG, the private equity giant, is circling Netrality Data Centers. Price tag: $3 billion. But don't confuse this for a routine infrastructure acquisition. This is the real estate endgame for AI’s compute hunger, and I’ve seen this pattern before.
Tracing the AI infrastructure endgame back to the genesis block of crypto mining. In 2017, I watched EOS block producers hoard tokens before the mainnet launch. Today, it’s data center capacity. The same playbook—accumulate scarce, physical assets before the demand curve steepens. The difference? This time the numbers are bigger, the timeline longer, and the players include pension funds.
Context: Why Now? Netrality operates over 10 interconnection hubs across secondary U.S. markets—St. Louis, Kansas City, Philadelphia. These aren’t the hyperscale campuses in Northern Virginia or Silicon Valley. They’re carrier hotels, dense with fiber and power. For AI workloads that don’t require ultra-low latency—batch inference, training on public datasets—these locations offer a 20% discount on electricity. The average price per kilowatt-hour in those metros hovers around $0.07, versus $0.10 in Ashburn.
TPG’s move isn’t isolated. Blackstone bought QTS for $10 billion in 2021. KKR took CyrusOne private for $15 billion. The EBITDA multiples have ballooned from 20x to 30x+ in three years. The market is pricing in a permanent AI-induced demand shift. But here’s the catch: most of those deals were for Tier 1 markets. Netrality is Tier 2. That’s where the contrarian play lies.
Core: The $3B Math Let’s break the valuation. Roughly $3 billion for a portfolio of data centers. Industry benchmarks value IT load at $6-8 million per megawatt for purchased assets. That implies Netrality controls 400-500 MW of critical IT capacity. Assuming an average load of 40 kW per rack (standard for high-density), that’s 10,000-12,500 racks. Each rack can host 40-60 H100 GPUs. So we’re talking 400,000 to 750,000 GPUs worth of space. That’s enough to train GPT-5 three times over.
But the hidden variable is power. Secondary markets have cheaper electricity but lower grid headroom. Netrality’s facilities likely draw 20-30 MW each. Scaling to 50 MW requires substation upgrades. TPG will need to invest another $500 million to $1 billion in power infrastructure over five years. That’s not in the $3 billion headline.
I’ve run this math before. During the 2020 Curve Wars, I calculated the liquidity crisis probability by modeling withdrawal speeds. Same principle here: TPG is buying an option on future power availability, not just concrete and cabling.
Chasing the alpha while the market sleeps – the acquisition was quietly negotiated for months. The public market hasn’t priced in the spillover effects. Digital Realty (DLR) trades at 22x EBITDA. Equinix (EQIX) at 25x. If this deal closes, expect those multiples to re-rate upward by 10-15% as investors realize that infrastructure scarcity isn’t limited to hyperscale.
Contrarian: The Overhyped Risk But let’s pause. Is AI demand infinite? Not even close. The enterprise adoption curve is S-shaped, and we’re still in the early vertical segment. A 500 MW portfolio requires 400,000 GPUs running at 100% utilization to break even on the investment. Today, global GPU utilization for AI training is around 60-70%. The gap between capacity and usage is widening. If AI hits a winter in 2026, these data centers could sit half-empty.
Speed over precision when the chart breaks – I learned in 2018 that when EOS collapsed, the block producers who overleveraged on hardware got wrecked. TPG is not overleveraged—they’re using 40-50% equity—but debt costs are rising. The 10-year Treasury is at 4.5%. If rates stay high, the financing costs eat into the yield. The projected unlevered IRR of 12-15% might shrink to 8%.
Reading the room in the order book silence – look at the data center REIT order books. They’re quiet. No massive buying or selling. Institutional money is waiting. TPG’s acquisition could be the catalyst. But that also means the market is already expecting this, so the upside might be priced in.
My Takeaway From the sprint to the sprawl of DeFi – I remember when Axie Infinity’s economy imploded. The warning signs were there: reward inflation outpacing user growth. Data centers face a similar dynamic. The construction pipeline is enormous. Over 2 GW of capacity is under development in the U.S. alone. By 2027, supply could outstrip demand. TPG is betting on scarcity, but the real alpha is in the exit strategy.
Forward-looking thought: TPG will likely merge Netrality with its existing platform (DataBank) and take the combined entity public via a REIT within three years. The $3 billion acquisition is a construction of a narrative, not just a portfolio. Watch the Form 13F filings for other PEs. KKR and Blackstone will follow with secondary market deals. The signal is clear: AI infrastructure is the new crypto mining – capital will rush to the bottleneck.
Tags: ["TPG", "Netrality Data Centers", "AI Infrastructure", "Data Center M&A", "Private Equity", "Institutional Investment"]
