How the Iran War Quietly Became the Biggest Threat to AI Infrastructure Stocks

The Story Wall Street Missed

While everyone watched oil prices and gas pumps during the six-week Iran war, the real casualty was hiding in plain sight: the neocloud AI infrastructure names that had been 2025's best-performing trade. IREN, CoreWeave, and Nebius got hammered alongside the Magnificent Seven, and most retail investors chalked it up to risk-off sentiment. That was only half the story.

The other half is structural, and it matters far more for how these stocks trade over the next twelve months. When WTI crude spiked from $67 before the war to well over $110 at its peak, every company whose cost structure depends on cheap electrons took a quiet but brutal hit to forward margins. That's the AI infrastructure trade in one sentence.

Why This War Was Always an AI Infrastructure Story

Here's the connection most analysts glossed over. Natural gas generates roughly 40% of US electricity, and electricity accounts for 40% to 60% of data center operating expenses, according to Boston Consulting Group's framework on US data center power economics. Natural gas prices move in lockstep with oil during geopolitical shocks because LNG arbitrage, substitution in industrial processes, and power-generation fuel switching all link them at the margin.

When Iran effectively closed the Strait of Hormuz, it didn''t just spike gasoline prices. It reset the cost curve for every gas-fired power plant in North America, including the ones Microsoft, Meta, and Alphabet have been racing to build behind-the-meter for their AI buildouts. PJM Interconnection capacity auction prices had already exploded from roughly $60/kWh in 2024 to over $300/kWh in 2025, and the war added another layer of gas-price volatility on top of an already broken supply-demand equation for electrons.

That's the bear case no one on CNBC was articulating during the war: every dollar that oil stayed elevated was a dollar compressing gross margins at the neoclouds that sell GPU hours at fixed, multi-year contracted rates.

The Neocloud Margin Math Most Retail Investors Are Missing

The neocloud business model is deceptively simple. Buy Nvidia GPUs on debt, stuff them into data centers, sell compute capacity to hyperscalers and AI labs on multi-year take-or-pay contracts. The gross margin on that contract is essentially the spread between your fixed contracted revenue and your mostly-variable power cost. When power costs jump 15% to 30% in a quarter, there''s no pricing lever to pull, because the contracts are locked.

Look at what each of these companies is actually exposed to:

• IREN Limited is arguably the best-positioned of the three, with a 3 GW grid-connected power portfolio and a $9.7 billion five-year Microsoft contract running at targeted 85% EBITDA margins. That EBITDA margin assumption relies on power costs behaving. Every 100 basis points of margin erosion on a contract that size is roughly $97 million in lost cumulative EBITDA over the life of the deal.
• CoreWeave carries a $55 billion revenue backlog against already-ballooning debt, with 2026 revenue guidance around $12 billion. The company is essentially levered long AI compute demand and levered short cheap power. Its Q4 loss of 89 cents per share already exposed how thin the buffer is. A sustained energy cost shock doesn''t break the company, but it does break the narrative that the path to profitability is a straight line.
• Nebius Group just closed a $10 billion European data center deal on top of a $27 billion Meta contract, with ARR targets of $7 billion to $9 billion by end of 2026. European gas prices are even more exposed to Middle East disruption than US prices, which means Nebius''s Finnish and Dutch sites felt the war more directly than any US competitor.

Now layer in the valuation context. Nebius trades at a price-to-sales multiple that only makes sense if you believe the 500%+ revenue growth projections play out exactly as guided, with margins intact. IREN''s EBITDA-margin story is the entire bull case. CoreWeave''s backlog conversion is priced for flawless execution. None of these stocks have room in their valuations for structurally higher power costs.

Is IREN Stock a Buy After the Ceasefire?

The April 8 ceasefire announcement sent oil tumbling 16% in a single session and the Nasdaq up 2.8%, giving the neoclouds their first real bounce in weeks. The question now is whether that bounce has legs or whether it''s a dead-cat move in a structurally damaged trade.

Here''s the realistic read. The ceasefire is two weeks long. BCA Research''s chief geopolitical strategist warned that fighting will likely reignite later this year after US midterm elections, because Tehran acting as gatekeeper over the Strait of Hormuz is politically unsustainable for Washington long-term. Oil prices are still sitting at $94, well above the $67 pre-war baseline. That floor is probably where things settle regardless of what happens in Islamabad negotiations, because governments are now hoarding and restocking in anticipation of renewed conflict.

Translation for the neocloud trade: the energy cost reset isn''t getting fully reversed. The base case is elevated gas prices sticking around for the rest of 2026, with another volatility spike if the ceasefire collapses.

Probability-weighted, here''s how the setup looks over the next six months:

• Bull case (30% probability): A permanent US-Iran deal gets hammered out, the Strait reopens fully, gas prices normalize toward $3.50/mmbtu, and the neoclouds resume their late-2025 trajectory. IREN, CRWV, and NBIS all retest recent highs as the AI capex narrative reasserts itself.
• Base case (50% probability): The ceasefire holds imperfectly, gas and oil settle at structurally higher levels than pre-war, margin pressure is real but manageable, and these stocks trade sideways-to-modestly-higher as revenue growth partially offsets margin compression. IREN outperforms due to its diversified power portfolio and Bitcoin hedge.
• Bear case (20% probability): The ceasefire collapses before month-end, oil retests $110+, electricity costs in key data center regions spike another 15%, and neocloud EBITDA guidance gets cut across the board. CoreWeave is the most exposed given its debt load and thin margin buffer.

The asymmetric bet here, in my view, is IREN. The combination of the 3 GW power portfolio (much of it secured at legacy prices from its Bitcoin mining build), the Microsoft contract at 85% targeted EBITDA margins, and a still-profitable Bitcoin mining business that hedges against AI margin pressure makes it the one name that can absorb energy cost shocks without the story falling apart.

Where Nvidia Fits In the Energy Cost Equation

Nvidia sits in a weirdly insulated spot in this whole story. The company sells GPUs upfront at fixed prices to customers who then have to figure out how to power them. The energy cost problem is entirely downstream of Nvidia''s P&L. When CoreWeave, IREN, and Nebius see their margins squeezed by a natural gas spike, Nvidia has already booked the revenue and moved on. That''s why Nvidia held up noticeably better than the neoclouds during the six-week war.

The counterintuitive angle is that elevated energy costs are actually accelerating demand for Nvidia''s newest, most efficient silicon. Blackwell delivers roughly 25x the inference performance per watt of Hopper, and Rubin is projected to push that gap wider. In a power-constrained world, customers don''t stop buying GPUs, they buy the most efficient ones and retire the older ones faster. That''s a tailwind for Nvidia''s average selling price and refresh cycle, not a headwind. Hyperscalers aren''t cancelling orders, they''re prioritizing the newest generation because every watt matters more when gas is $5/mmbtu instead of $3.

The second-order risk is real though. If hyperscaler capex actually cracks in 2026 as energy costs crush data center ROI math, Nvidia eventually feels it through order slippage. New data center deals already fell 40% between Q3 and Q4 2025, and hyperscaler capex could fall by as much as half in 2026 if financing and power costs stay elevated. That''s the only real bear case for NVDA out of this situation, and it''s a second-derivative move that takes quarters to play out, not weeks.

What a Permanent Deal Would Actually Mean

If the Islamabad negotiations produce a real, durable peace agreement rather than another fragile pause, the implications for AI infrastructure stocks are more interesting than most investors realize. A permanent reopening of the Strait of Hormuz would pull global oil back toward the $70s, drag Henry Hub gas prices with it, and restore the pre-war power cost structure that made neocloud unit economics work in the first place.

For IREN, that means the 85% EBITDA margin target on the Microsoft contract stops being aspirational and starts being conservative. For Nebius, it means European data center economics stop bleeding and the $7-9 billion ARR target comes back into focus. For CoreWeave, it''s existential in a different way: lower power costs directly improve the math on debt service and the path to free cash flow positivity.

The counterfactual also matters. Tech capex from hyperscalers could fall by as much as half in 2026 if energy and financing costs stay elevated, according to Information Technology and Innovation Foundation research. That''s the headwind a permanent peace deal would partially reverse, and it''s the primary reason a durable ceasefire is probably more bullish for the neoclouds than a 10% cut in rates.

The Bottom Line for AI Infrastructure Investors

The Iran war wasn''t a sideshow for AI infrastructure stocks. It was a stress test of whether the neocloud business model survives in a world where power costs can spike 30% without warning. The ceasefire rally is giving investors a chance to reposition, not a signal that the energy risk has gone away.

If you believe AI demand is structural (and the $55 billion backlog at CoreWeave, the $27 billion Meta deal at Nebius, and the $9.7 billion Microsoft contract at IREN all suggest it is), the real question is which of these companies has the power cost moat to deliver on its guidance regardless of what happens in the Middle East. On that test, IREN looks meaningfully stronger than its peers, Nebius looks compelling on valuation after the selloff, and CoreWeave remains the highest-risk, highest-reward name with the biggest downside if energy costs don''t cooperate. Nvidia stays the boring, safer way to stay long the theme through the geopolitical noise.

Watch the Strait of Hormuz tanker traffic data over the next two weeks. That''s your leading indicator for whether the neocloud trade works in 2026, no matter what the AI capex headlines say.

---

Disclaimer: This article is for informational purposes only and should not be considered financial advice. Stock investing involves significant risk, including potential loss of principal. Always conduct your own research and consult with a qualified financial advisor before making investment decisions.