If you own or follow neoclouds & AI plays like $TeraWulf Inc.(WULF)$ , $IREN,$NEBIUS(NBIS)$ , $Cipher Mining Inc.(CIFR)$ , & $ALPD, you need to read this 🧵👇
📉 The Market's Blind Spot:
Wall Street is currently punishing the entire AI infrastructure sector as if every company operates the exact same way. They are confusing a CoreWeave ($CRVW) financing problem with an AI demand problem.
Let’s look at the actual facts:
⚡ Compute Demand is Roaring:
• Rising Prices: Blackwell Ultra (B300) GPU rental prices hit a new all-time high after rising for 17 straight days (Source: Matthew Sigel). Older H100s and A100s also command strong rates.
• Massive Hyperscaler Capex: Meta, Amazon, Microsoft, and Google are projected to spend roughly $750 billion on capex this year—up over 80% from last year's ~$410B (Source: Unusual Whales).
• The Real Bottleneck: Power availability, permits, land, and cooling systems remain heavily constrained. AI infrastructure doesn't magically appear overnight.
So, why are stocks getting crushed? It all comes down to financing risk, not a lack of demand.
💥 Where the Real Trouble Lies: $CoreWeave, Inc.(CRWV)$
Discussed in All-In Podcast (Ben Horowitz, Chamath Palihapitiya, David Sacks), strong demand doesn't guarantee a great business model if your financing structure is fragile:
• Debt & Obligations: Estimated between $35B to $50B, with another ~$35B potentially needed.
• Crushing Borrowing Costs: Bond yields climbing toward 15% (up from ~8.5%).
• Fixed-Price Trap: Long-term fixed contracts can leave them underpaid while financing costs skyrocket.
Lesson: You can have booming demand, but if your leverage and capital costs turn against you, the equity holders suffer.
🏗️ Why Powered Infrastructure Plays ($WULF, etc.) are Different:
Not all neoclouds or data center providers take on the same risk as pure GPU-rental models.
Take TeraWulf ($WULF) as an example:
• CEO Paul Prager highlights that their model centers on developing and leasing powered real estate backed by long-term agreements and high-grade counterparties (like Google), rather than taking massive, debt-financed GPU ownership risks.
• Companies like TeraWulf, Cipher, and Applied Digital focus on power, land, and construction execution—a fundamentally different risk profile than buying billions in rapidly depreciating GPUs via high-yield debt.
🎯 The Bottom Line:
Investors are conflating two very different questions:
Is demand for AI compute growing? YES. (GPU rental prices are up, labs are scaling).
Will every single company turn that demand into strong shareholder returns? NO.
The current market selloff is exposing fragile balance sheets, but it is also mispricing fundamentally sound infrastructure and power providers.
Separating strong, cash-flow-driven operators from over-leveraged business models is where the real opportunity lies. 💡
Comments