Three AI Infrastructure Stocks Set to Triple by 2029

FinanceGET IN EARLY! These 3 Stocks Will Make Millionaires By 2029

TL;DR

CoreWeave, Nebius, and IronNet are positioned to triple their AI compute capacity by 2026 as ACIE (non-hyperscaler AI) spending grows 138% year-over-year, making them critical picks in the AI infrastructure buildout beyond the tech giants.

Key Takeaways

1ACIE (AI clouds, industrial, enterprise) revenue grew 138% year-over-year at Nvidia, outpacing hyperscaler growth of 102% for the first time, signaling AI spending is broadening beyond Microsoft, Amazon, Google, and Meta.
2CoreWeave has $14 billion in revenue backlog (up 246% YoY) with 4.2 GW contracted power capacity, but carries $46 billion in net debt where interest expenses consume 25% of quarterly revenues.
3Nebius grew revenues 454% year-over-year to $582 million and received $9 billion in customer prepayments covering 50-60% of equipment costs, reducing reliance on high-interest debt with only $2 billion net debt.
4IronNet's AI cloud revenue more than doubled in 90 days to $70.5 million and now exceeds Bitcoin mining revenues, with 5 GW power pipeline already secured from Bitcoin mining operations at only $1.9 billion net debt.
5Nvidia partnered with all three companies: buying $6.3 billion in unused CoreWeave capacity through 2032, investing $2 billion each in CoreWeave and Nebius, and committing $3.4 billion with IronNet for internal GPU deployment.
6Neo-cloud providers are the fastest-growing AI infrastructure segment as Nvidia expects partner capacity to grow from 3 GW at end of 2025 to 8 GW by end of 2026, creating virtuous cycle of increased AI adoption.
7CoreWeave trades at $62 million per megawatt of active power (6x cheaper than Nebius/IronNet), while IronNet costs $3 million per megawatt of contracted power (7x cheaper than CoreWeave), creating different risk/reward profiles.

Notable Quotes

“For every dollar they all make, Nvidia keeps twice as much as Google and almost four times as much as AMD.”
— Alex

“This is the first time that hyperscalers weren't the fastest growing segment, which means the AI buildout is broadening well beyond the trillion dollar tech giants.”
— Alex

“Nvidia will purchase any unsold cloud capacity from Cororeweave through April of 2032—that's a buyer of last resort.”
— Alex

“Customers are funding their AI buildout instead of high-interest loans with 50 to 60% of equipment cost prepayments.”
— Alex

Chapters

1. Nvidia Earnings Signal Broader AI Buildout

Nvidia reported record $96.2 billion quarterly revenue (106% YoY growth) with 66.2% operating margin—twice Google's and four times AMD's—but the key insight is splitting data center revenue into hyperscale versus ACIE, where ACIE grew 138% YoY, faster than hyperscaler growth for the first time.

2. Why Neo-Clouds Matter

Neo-cloud providers (CoreWeave, Nebius, IronNet) are specializing in AI infrastructure for non-hyperscaler buyers; they have customer contracts from competing firms, direct Nvidia partnerships, and are positioned to grow from 3 GW to 8 GW installed capacity by end of 2026.

3. CoreWeave: Largest Scale with Debt Risk

CoreWeave operates 51 data centers across North America/Europe with 1.5 GW active power and 4.2 GW contracted; quarterly revenue of $2.6 billion (112% YoY) backed by $14 billion backlog, but $46 billion net debt where $640 million quarterly interest equals 25% of revenues.

4. Nebius: Strongest Balance Sheet with Fastest Growth

Nebius grew revenues 454% YoY to $582 million with $37.5 billion contracted work including $27 billion Meta deal; customers provide 50-60% equipment cost prepayments ($9 billion expected this year), resulting in only $2 billion net debt and 20% interest-to-revenue ratio.

5. IronNet: Smallest with Lowest Cost Per Megawatt

IronNet's AI cloud revenue more than doubled in 90 days to $70.5 million and now exceeds Bitcoin mining; leverages 5 GW pre-existing power pipeline, holds $16.6 billion backlog including $9.7 billion Microsoft and $3.4 billion Nvidia deals, with only $1.9 billion net debt at 18% interest-to-revenue ratio.

6. Valuation Comparison and Investment Recommendation

CoreWeave costs $62M per active megawatt (6x cheaper than competitors) but has 46B net debt; Nebius has 454% growth but 7.4x year-end run rate valuation; IronNet is cheapest at $3M per contracted megawatt despite lowest revenue, making CoreWeave lowest-risk, IronNet highest-upside, Nebius middle-ground choice.

Key People & Entities

Alex: Content creator with 8 years experience as electrical engineer and AI researcher at MIT; host of the video
Jensen Huang: Nvidia CEO; invested $2 billion in CoreWeave stock at $87 per share
Nvidia: AI chip manufacturer and infrastructure partner; directly invested in and contracted with all three neo-cloud companies
CoreWeave: Neo-cloud provider with 51 AI-optimized data centers, 1.5 GW active power, $14B revenue backlog
Nebius: Neo-cloud provider with unified platform for training (Aether) and inference (Token Factory), 454% YoY revenue growth
IronNet: Former Bitcoin miner pivoting to AI cloud infrastructure with 5 GW power pipeline and 10x+ AI revenue growth
Microsoft: Hyperscaler with multi-billion dollar contracts across all three neo-clouds; $17.4B deal with Nebius, $9.7B with IronNet
Meta Platforms: Hyperscaler with $21B CoreWeave contract and $27B Nebius contract for AI infrastructure

Glossary

ACIE: AI Clouds, Industrial and Enterprise—Nvidia's data center revenue category excluding hyperscalers, including healthcare, financial services, automotive, robotics, and specialized AI cloud providers; grew 138% YoY.
Neo-Cloud: Specialized AI infrastructure provider offering compute capacity through managed cloud platforms; includes CoreWeave, Nebius, and IronNet.
Contracted Power: Future electricity promises to data center sites that may not be built yet; differs from active power which generates current revenue.
Active Power: Electricity actually flowing into servers running today; the basis for current revenue generation.
Net Debt: Total debt minus cash on hand; represents what a company would owe if it emptied its bank accounts immediately.
Enterprise Value: Market capitalization plus total debt minus cash; better metric than market cap for comparing debt-heavy companies.
GW (Gigawatt): Unit of electrical power equal to 1 billion watts; used to measure data center power capacity.
Vera Rubin / Blackwell: Nvidia's latest GPU architectures; Vera Rubin racks hold 72 GPUs each requiring ~120 kW of power.

Explore