Neoclouds Split Data Center Investors: AI Boom or WeWork Sequel?

Last updated: October 15, 2026

A new class of tenant is reshaping who finances America's data center boom, and not everyone in the room is comfortable with it. According to bisnow.com, executives gathered at the Data Center Capital Markets Summit on July 16 at the New York Marriott Marquis to debate whether AI-focused cloud startups, dubbed neoclouds, are sturdy enough to anchor the long-term leases the industry runs on. The conversation cuts straight through commercial real estate, where capital allocation decisions made today will shape rent rolls and asset values for the next decade.

The stakes are high. Data center construction in the United States has ballooned into a roughly $2 trillion building program. Most of that money has historically flowed to projects pre-leased to Amazon, Microsoft, Google or Meta. Now neoclouds are taking a fast-growing slice of that demand, forcing lenders and developers to rethink how they underwrite a deal.

Key facts

• Neoclouds are AI-specialized cloud providers offering GPU computing on demand; JLL counts more than 190 operators, most of them startups.
• The segment posted 223% year-over-year revenue growth in the fourth quarter, per Synergy Research Group, cited by bisnow.com.
• Largest named players include CoreWeave, Core Scientific, Nscale and Lambda.
• Industry executives met July 16 at the New York Marriott Marquis for Bisnow's Data Center Capital Markets Summit.
• Skeptics compare neocloud lease structures to WeWork, a comparison echoed at the summit by lawyer Jeffrey Moerdler of Haynes Boone.

How the mismatch shows up in a lease

Here is the pattern lenders are losing sleep over. A neocloud signs a 15-year lease for power and floor space. It then sells that GPU capacity to end customers on agreements that can run by the hour, the day, the month, or rarely past three years. If AI demand cools, or if a customer builds its own chips and racks, the neocloud still owes the landlord for the full building.

Phillip Koblence, co-founder and COO of data center provider NYI, told attendees that underwriting these tenants is uncomfortable work. He asked how confident anyone can be that these firms will still be around to honor a long-dated lease, according to bisnow.com. That is the question every credit committee is now asking.

Moerdler went further. He said the lease-versus-revenue gap reminds him of WeWork, pointing to the failed coworking operator that collapsed after business-model doubts. He also flagged a second concern: a meaningful chunk of neocloud revenue comes from hyperscalers themselves, using these providers as a temporary bridge while they build out in-house AI capacity. If that bridge traffic fades, so does a key source of cash flow.

What this means for you

If you underwrite, lend to, or invest alongside data center developers, the trade is straightforward. The classic hyperscale deal still pencils, with investment-grade tenants, low default risk and cheap debt. Neocloud deals require deeper diligence, tighter covenants and often an investment-grade guarantor or hefty upfront collateral to clear a credit committee. Fentress Boyse of Partners Group told the summit his firm is a believer in many of these businesses, but acknowledged the financing complexity. Expect more bespoke structures, not blanket formulas.

For investors sizing the broader opportunity, two things matter. First, neocloud demand growth is real. A 223% revenue jump in a single quarter, per Synergy Research Group, is not a rounding error. Second, capital is rationed. Lenders are not closing their doors; they are repricing risk. That gap between demand and traditional financing is exactly where specialists, niche funds and private credit shops can earn their fee.

Worked example: underwriting a neocloud deal

Say you are a development director evaluating a build-to-suit data center project. Under the old playbook you would chase a hyperscaler pre-lease, secure cheap construction debt, and hand the asset to a core buyer on stabilization. Now imagine a neocloud anchor tenant instead. Your lease still runs 15 years, but the tenant's customer book is a mix of AI startups and hyperscaler overflow, with contracts as short as one quarter.

Your underwriting changes. You would model a stress case in which a large share of customer contracts lapse early, and price the risk of finding a replacement tenant in a tight market. You would push for a parent guarantee from a sponsor with serious committed capital, a meaningful debt service reserve, and a change-of-control clause that lets you reprice or find a replacement tenant if the neocloud is acquired. You would also widen your lender pool to private credit funds and infrastructure debt shops that price risk explicitly rather than running from it. Done well, the deal clears. Done lazily, you end up with a stranded shell and a workout.

What to watch next

Watch three signals. First, hyperscaler disclosure of in-house AI capacity, since every gigawatt they bring online is potential demand leaving neoclouds. Second, lease structures at the next round of financings, and whether guarantees, reserves and covenants become standard, or stay negotiated case by case. Third, the survival rate of the 190-plus neocloud operators over the next stretch; consolidation would harden the survivors into stronger tenants, while a wave of failures would validate the WeWork comparison.

The bottom line: neocloud data center tenants are not a passing fad, but they are also not the investment-grade anchor the market is used to. Smart capital will finance them. Careless capital will learn the hard way.

Key takeaways

• Hyperscale-anchored deals remain the safest path; neocloud deals require bespoke terms, guarantors and bigger reserves.
• Demand from AI startups is real and growing fast, which means opportunity, not just risk, for investors who can underwrite it.
• Closely watch hyperscaler capex disclosures and the next two quarters of neocloud tenant credit performance before sizing new exposure.