Dollar strength pushes Yen to 40-year low despite intervention warnings
The biggest infrastructure build since the railroads is happening right now — and most investors are looking at the wrong stocks.
While Wall Street obsesses over Nvidia’s (NASDAQ:NVDA) chip margins and debates whether Big Tech’s $600 billion AI spending spree will pay off, a new category of company is quietly becoming indispensable. Neoclouds, the purpose-built AI cloud providers are signing multi-billion-dollar contracts with the very hyperscalers that were supposed to build everything themselves. And the math is starting to get very interesting.
The $49 Billion Backlog No One Is Talking About
Nebius Group (NASDAQ:NBIS) just closed a $4.34 billion convertible debt offering on Monday, capping a three-week stretch that fundamentally transformed the company. The Amsterdam-based neocloud, spun out of Yandex’s international operations in 2024, now sits on approximately $49 billion in contracted backlog from three of the world’s most aggressive AI spenders.
The sequence went like this: On March 11, Nvidia invested $2 billion directly in Nebius, acquiring an 8.3% stake. Five days later, Meta Platforms (NASDAQ:META) signed a $27 billion, five-year AI infrastructure agreement — one of the largest hyperscaler cloud contracts ever awarded. And underneath all of that sits a Microsoft deal worth between $17.3 billion and $19.4 billion, inked last September.
Jensen Huang said at GTC 2026 that Nvidia is "scaling its cloud solutions together with NBIS, considering it a close partner for future AI demand." That isn’t a press release pleasantry. When the CEO of the world’s most important chip company singles out your firm in front of every institutional buyer on the planet, it rewrites your customer acquisition economics overnight.
The convertible offering itself was upsized from the original $3.75 billion target to $4.34 billion, which is split across $2.58 billion in 1.25% notes due 2031 and $1.75 billion in 2.625% notes due 2033. The oversubscription tells you demand wasn’t a problem.
Why Hyperscalers Are Outsourcing What They Said They’d Build
Here’s the thing most people miss about the neocloud thesis. Meta is guiding for $115 billion to $135 billion in capital expenditure this year — a 73% increase from 2025. Microsoft (NASDAQ:MSFT) is spending in a similar range. These companies aren’t outsourcing to Nebius because they can’t build data centers. They’re outsourcing because they can’t build them fast enough.
The Meta contract has two tiers worth understanding. Tier one is $12 billion in dedicated GPU capacity powered by Nvidia’s next-generation Vera Rubin processors, with deployment starting early 2027. That’s contracted, committed revenue. Tier two is a $15 billion backstop agreement where Meta has the right to purchase any remaining capacity in Nebius clusters that isn’t sold to other customers. That backstop effectively eliminates demand risk for Nebius’s entire capacity pipeline. Build it, and Meta fills whatever isn’t already committed to Microsoft.
Nebius’s management expects 60% of its growth to be financed through customer prepayments, primarily from Meta and Microsoft — with the remaining 40% coming from equity and debt. CFO Blackwell said the company is "well-funded" to meet its 2026 capital spending target of $16 billion to $20 billion.
For context, Nebius has generated only $530 million in trailing twelve-month revenue. That’s the kind of revenue-to-backlog ratio that either signals a generational opportunity or a capital execution disaster. The Nvidia stamp and Meta backstop argue strongly for the former.
How to Play It
1. Nebius Group (NBIS)
The primary bet on the neocloud category. At around $115, the stock trades at roughly 3.4x forward revenue — a premium to CoreWeave’s 1.9x but justified by the contracted backlog, the Nvidia co-branding, and the Meta backstop structure. Wall Street consensus is "Strong Buy" across nine analysts, with an average price target of $169. DA Davidson and BWS Financial both lifted their targets to $200 after the Meta deal. That’s 47% to 74% upside from current levels. The biggest risk is execution — converting $49 billion in contracts into functioning data centers running Vera Rubin chips at scale is not trivial. But Nvidia’s $2 billion bet says they believe Nebius can do it. The next earnings report is April 29.
2. CoreWeave (NASDAQ:CRWV)
The other neocloud, and the direct comparison. At around $84, CoreWeave carries a larger market cap (~$43 billion) despite a smaller contracted backlog. The company plans $30 to $35 billion in 2026 capex and sits on a $66.8 billion total backlog with $28 billion to be recognized within 24 months. CoreWeave trades cheaper on a forward revenue basis (1.9x vs. Nebius’s 3.4x), but that discount exists for a reason — higher debt levels, a class action lawsuit, and what Bernstein flagged as potential cannibalization risk from Big Tech customers building their own infrastructure. The analyst consensus is "Strong Buy" with an average target of $121, implying 44% upside. My read: CoreWeave is the higher-beta, higher-risk version of the same trade. If the neocloud thesis works, both win. If it doesn’t, CoreWeave’s balance sheet makes it more vulnerable.
3. Equinix (NASDAQ:EQIX)
The landlord play. At around $960, Equinix is the world’s largest data center operator with 270 properties across 36 countries. While neoclouds like Nebius build custom AI facilities, they still need physical locations, and Equinix controls the prime real estate. The company’s Q4 monthly recurring revenue grew 10% year-over-year, and Reflexivity CEO Jan Szilagyi called it "the landlord for the AI revolution" in February. Barclays just raised its target to $1,020 from $870. The 21-analyst consensus is "Buy" with an average target near $1,000. Equinix also pays a 2.1% dividend, making it the defensive way to play the AI infrastructure buildout. Lower upside ceiling, much lower execution risk.
4. Digital Realty (NYSE:DLR)
Equinix’s smaller counterpart, and arguably the better value at current levels. At around $175, DLR operates 310 data centers globally and recently expanded into Portugal through a Lisbon acquisition. Total bookings hit $1.2 billion in Q4, and management guided for 2026 Core FFO growth of 8% with revenue growth exceeding 10%. The analyst consensus target is $194, implying roughly 11% upside, with Bernstein initiating at Outperform and BMO maintaining Buy. DLR yields 2.7%, giving income investors a meaningful dividend while they wait for the AI infrastructure cycle to play out. At just 50x trailing earnings versus Equinix’s richer multiple, DLR offers more margin of safety.
5. Global X Data Center & Digital Infrastructure ETF (VPN)
For investors who want the entire data center ecosystem without single-stock risk. VPN holds major data center REITs, neocloud infrastructure plays, and related technology companies, which is a diversified exposure to the physical layer of AI without forcing you to pick which company executes best.
The Bear Case You Can’t Ignore
Not everything about the neocloud model is bulletproof. CoreWeave’s ~$100 billion cumulative funding gap through 2030, flagged by multiple analysts, illustrates the core risk: these companies are deploying capital at a pace that would make an oil major blush, and they’re doing it based on the assumption that AI demand continues accelerating. If enterprise AI adoption plateaus, the revenue needed to service this debt evaporates.
There’s also a structural question about what happens when hyperscalers finish their build-out. Meta, Microsoft, and Google are all constructing massive proprietary data centers simultaneously. The neocloud value proposition depends on the gap between demand and supply persisting. Bernstein’s analysts explicitly warned that CoreWeave’s business "could be cannibalized" as Big Tech customers eventually bring capacity in-house.
And Nebius specifically carries country risk. The company was born from Yandex’s restructuring, and while it’s now Amsterdam-headquartered and Nasdaq-listed, its operational roots trace back to Russia. That overhang isn’t priced at zero.
But here’s why the bull case still dominates: the AI compute shortage isn’t closing. It’s widening. Nvidia’s GTC 2026 keynote confirmed $1 trillion in combined Blackwell and Vera Rubin orders through 2027. Someone has to deploy those chips. The hyperscalers literally can’t build fast enough. And as long as that’s true, the neoclouds are the bottleneck — and the bottleneck gets paid.
What to Watch
Three catalysts in the next 30 days. First, Nebius Q4 2025 earnings on April 29 — the first report that will begin to reflect the Meta and Microsoft deal momentum in forward guidance. Revenue run-rate acceleration and updated 2026 capex targets will be the key metrics.
Second, CoreWeave earnings on May 20 — the market will compare CoreWeave’s execution and funding trajectory directly against Nebius’s. Any signs of widening or narrowing in the execution gap between the two neoclouds will move both stocks.
Third, the broader oil/Iran situation — Brent crude’s trajectory directly impacts global capex willingness. If oil stays above $100 and recession fears intensify, hyperscaler spending guidance becomes the canary in the coal mine. Watch Meta’s next capital update closely.
The neocloud category didn’t exist two years ago. It now holds more contracted revenue than most S&P 500 companies generate. Whether that converts into sustainable profit is the trillion-dollar question — but the checks from Nvidia, Meta, and Microsoft say the smart money has already decided.