The Market Just Gave You a 30% Discount on AI Infrastructure
CoreWeave's bad earnings triggered panic. Morgan Stanley says the power shortage is still coming
The market is having a moment. And not the good kind.
So-called NeoClouds like $IREN and $NBIS are down 30% from their peaks. Energy stocks are being sold indiscriminately. Even Microsoft and META – companies with completely different business models and solid earnings – are getting hammered.
Welcome to deleveraging season.
What’s driving the selloff?
First: CoreWeave’s (CRWV) terrible earnings spooked the entire sector. The company is debt-laden and struggling to scale. Now the market is asking: can NeoClouds actually make money from their data centers?
But here’s the thing – you can’t compare CRWV’s 4% margin business model with something like Nebius. Different animals entirely.

Second: The market ran non-stop since April. Investors are out of ideas and selling everything. It’s indiscriminate rotation – out of tech and energy, into banks, insurance, and industrials. But even that rotation isn’t particularly strong. There’s no massive inflow into DOW components.
We saw this exact playbook in summer 2024 when there was mass rotation out of semiconductors. Companies like ARM lost 30% too. Sound familiar?
The Denial Phase
Here’s what makes this fascinating: the news keeps getting better, and the market doesn’t care.
Just today:
Anthropic announced plans to invest $50 billion in Texas data centers
TeraWulf expanded its partnership with Fluidstack
Microsoft detailed plans for a massive AI factory
TeraWulf’s response to good news? Up 2%. That’s it.
The market has entered the denial phase. When great news gets ignored, you’re either at a major turning point or in the middle of forced deleveraging.
Meanwhile, Morgan Stanley Sees the Bigger Picture
While the market panics, Morgan Stanley just published research warning that the US faces a 20% power shortage for AI data centers by 2028 – roughly 13 GW of deficit.
They call AI computing demand “the most important technological shift in modern history.”
Who fills the power gap?
Bitcoin miners converting to AI centers ($IREN, $CIFR)
Natural gas turbines: 15-20 GW potential
Bloom Energy fuel cells: 5-8 GW
Nuclear-powered deals: 5-15 GW
Two winning business models are emerging:
Short-term leases (like IREN’s 5-year deal with Microsoft)
Long-term REIT structures (like APLD’s 15-year agreement)
Morgan Stanley’s conclusion: “AI infrastructure stocks are at the center, with non-linear rate of AI improvement creating broader valuation impacts.”
Translation: if you control power infrastructure for AI, you’re positioned to print money over the next 3-5 years.
So What Do You Do?
Fundamentally, nothing has changed for NeoClouds and energy companies. The thesis is intact. The demand is real. The power shortage is coming.
But the market is in deleveraging mode, shaking out anyone who bought on margin.
My approach:
This is a time to carefully accumulate quality names, not panic sell
Use the selloff to upgrade your positions – swap weaker names for stronger ones
Don’t try to catch falling knives, but do have a shopping list ready
Remember: in summer 2024, ARM fell 30% during sector rotation, then recovered completely
The market cycles through these phases. Denial always feels like the end. It rarely is.
When everyone else is capitulating on a thesis that’s actually strengthening (Morgan Stanley just told you there’s a 13 GW shortage coming), that’s usually when the best opportunities appear.
Just don’t use leverage while you wait for the market to figure it out.
Not Everyone Needs to Swing for the Fences
If the volatility in NeoClouds and energy names is keeping you up at night, there are calmer waters in the AI infrastructure space.
Names like $VRT and $VST have pulled back from local highs but offer more stable exposure to the theme. You won’t see 300% runs, but you also won’t see 30% drawdowns in a week.

Hell, even $CAT (Caterpillar) plays into the infrastructure build-out, though I wouldn’t expect massive upside there.
But here’s the thing: where there’s no massive upside, you can preserve your mental health and stick to your risk model.
Not every position needs to be a 10-bagger. Sometimes the best trade is the one that lets you sleep at night while still participating in the broader theme.
Figure out what kind of volatility you can handle, then position accordingly. The AI infrastructure thesis doesn’t require you to white-knuckle through -30% swings if that’s not your style.
This publication is for educational and informational purposes only and does not constitute financial, investment, or trading advice. Readers are solely responsible for their own investment decisions. The author may hold positions in the securities mentioned.





The denial phase observation is spot on. When good news gets ignored and Morgan Stanley is literaly telling us theres a 13 GW shortage coming, that disconnect usualy means the market is just cleansing leverage. CoreWeave's troubles dont invalidate the broader power thesis, they just remind us that not all data center plays are created equal.
flawless bro, i disagree with the last part: "not eveerything needs to be a 10-bagger"... jk