Few investors ever hit every market regime change perfectly. You likely missed at least a piece of the Nvidia rally, sat on too much cash during the generative AI bubble, or saw promising AI unicorns fail to go public.
If you’re looking for your next high-conviction trade in AI, look no further than AI infrastructure stocks.
A new generation of data centres is powering the artificial intelligence revolution. These are the companies building them.
What Does “AI Infrastructure” Actually Mean?
Investing in AI infrastructure stocks means buying shares of the companies supplying the hardware and software required to build and run large-scale artificial intelligence. This includes everything from;
- GPU clusters (GPUs) that run AI inference and training;
- The data centres those GPUs are housed in;
- Electricity and power infrastructure;
- Cloud platforms that developers and businesses can use to rent AI compute; and
- Networking and interconnects that allow GPUs to communicate with one another.
Simply. there’s a ton of physical infrastructure that goes into supporting AI at scale. While many of the largest hyperscalers are investing heavily in their own data centre expansions to supply that capacity, a new generation of specialised providers is capturing a portion of the market by developing AI-specific offerings. We call this neocloud.
Why the Neocloud Sector Is the Most Exciting Corner of AI Investing Right Now
Neocloud companies like Nebius and CoreWeave specialise in renting out GPU clusters to developers and enterprises that want to train large AI models. They’re called neoclouds because they’re a new generation cloud provider designed specifically for AI use cases rather than general cloud computing.
“If you think about AWS vs a neocloud,” Bank of America analyst Tal Liani explained earlier this month, “neoclouds sit at the opposite end of the spectrum as hyperscalers. Rather than building general-purpose data centres, they specialise in rapidly deploying GPU-dense compute clusters optimised for throughput and bare-metal performance. They have architecture purpose-built to win on speed, specialisation, and hardware integration.”
Estimates suggest these specialised GPU providers will grow revenue at rates of 500% or more in 2026 while struggling to add capacity fast enough to keep up with AI demand. That capital spending bill has to go somewhere, and investors looking for exposure to the buildout of AI at scale can’t really go wrong with any of the neocloud names right now:
- Internet infrastructure company Nebius grew 87% in 2026 alone as investors scrambled to scoop up shares ahead of Q2 earnings.
- IREN is up 180% this year. A Bitcoin miner turned AI cloud infrastructure provider, IREN has secured several billion dollar deals with Microsoft while also continuing to generate profit from cryptocurrency.
- Industry leader CoreWeave trades at a ~30x forward P/S while building out some of the fastest-growing GPU infrastructure in existence.
- Keel is an earlier-stage play with exposure to high computing power infrastructure that recently rebranded from Bitfarms and began trading under KEEL earlier this year.
- Applied Digital makes money by constructing and leasing data centres to cloud providers rather than renting out space on its own balance sheet.
Disclaimer: The examples and companies mentioned above do not constitute a recommendation to buy or sell any stock.
Risks to Understand Before Buying AI Infrastructure Stocks
AI infrastructure investing isn’t without risk. Here are four major risks every investor should understand before buying any stock in this group.
Execution Risk
Ask any Founder or CEO and they’ll tell you building a data centre is hard. There’s tons of moving pieces — GPUs, power supply, timelines, construction, permitting, etc. — any one of which can stall and delay revenue by quarters.
AI infrastructure stocks also trade at premium valuations based on supply scarcity. Increased GPU buildout from Amazon, Microsoft, Google, and others will bring new supply to market over the next few years and could compress margins for providers that rely on rental income.
The businesses are capital intensive. AI infrastructure companies typically spend billions building out data centres before they ever record meaningful revenue from those facilities. Most are burning through cash and will need to return to the capital markets for additional liquidity.
There’s a lot that can go wrong. These are early-stage growth stocks trading on projected revenue several years in the future. Investors drive the market sentiment, AI spending could slow, and stock prices could fall just as fast as they rise.
Concentration Risk
Many AI infrastructure companies have been able to lock in billions — and in some cases, tens of billions — of dollars in future revenue via contract with Amazon, Microsoft, Google, and Meta. Trillion dollar corporations love locking in their AI compute needs for years in advance at known prices.
These long-term contracts are great for projecting cash flows years into the future. But they also create huge concentration risk. One or two hyperscalers make up the majority of revenue for many of these companies. If a renewal is renegotiated at lower pricing, or worse, not renewed at all the stock can tumble.
Top 5 AI Infrastructure Stocks to Watch in 2026
Nebius Group (NASDAQ: NBIS)
The Full-Stack Neocloud Leader
(Image Credit: Nebius)
Emerging out of Yandex’s sale of its Russian business earlier this year, Nebius has pivoted hard into neocloud and appears on track to be another massive success story in AI.
The differences between Nebius’ offering and competitors like CoreWeave are subtle but important. Nebius doesn’t simply rent GPUs to AI trainers. Rather, it offers a full-stack solution that includes:
- GPU clusters;
- Managed storage; and
- Developer tooling needed to deploy AI models.
This lets Nebius charge more per AI “unit” of compute than its rivals (think the difference between renting a house vs renting a house, carpet, fridge, stove, etc.) and allows it to maintain wider moats around its customer relationships. Combined revenue from those two deals is approximately $27 billion, raising Nebius’ total contract backlog to $49 billion.
Risk: Net revenue losses are growing. This growth story requires continued, unabated execution and will need to tap the capital markets for additional liquidity. Given the company’s trajectory and valuation it probably won’t have much trouble doing so, but still.
IREN Limited (NASDAQ: IREN)
The Bitcoin Miner Turned Infrastructure Giant
(Image Credit: IREN)
IREN has one of the best stories in the entire sector.
IREN used to be a Bitcoin mining company. These days it’s pivoting toward becoming a major player in AI infrastructure, but hasn’t given up mining Bitcoin. That unique advantage means IREN is actually generating profits and cash flow while building out its long-term AI infrastructure business.
IREN has over $9 billion of contracted revenue with Microsoft spread over five years, equating to roughly $2 billion in annualised revenue at 85% EBITDA margins once the buildout is complete.
Risk: It has been months since IREN announced a new AI contract. Investors will be looking for clues as to future deal flow on the company’s next earnings call.
CoreWeave (NASDAQ: CRWV)
The Category-Leading Neocloud Provider
(Image Credit: CoreWeave)
CoreWeave has emerged as the biggest name in neocloud infrastructure, rapidly growing GPU capacity while flipping from crypto miner to dedicated AI infrastructure provider.
It already has $66 billion in revenue backlog thanks to deals with Amazon, Microsoft, Google, and Meta. This year it set another record for quarterly contracted revenue, scoring $13 billion in Q1 alone.
Risk: High debt load and aggressive accounting have caused short interest to climb in CoreWeave. Make sure you know where you stand before buying.
Keel Infrastructure (NASDAQ: KEEL)
The Early-Stage Outsider With Infrastructure Exposure
(Image Credit: KEEL)
Keel Infrastructure is unique in that it was formerly a Bitcoin miner going by the name Bitfarms.
Keel just completed a statutory redomiciliation to the United States from Canada. As Bitfarms, the company rebranded as Keel Infrastructure and shifted its focus to providing data centre and energy infrastructure to AI and other high-computing power workloads.
Keel has 2.2 gigawatts of potential capacity deployed across five sites. In theory, this assets could end up supporting NVIDIA DGX Cloud Infrastructure once developed. Chardan initiated coverage with a Buy rating this week.
Risk: Still early-stage. Execution risk is high and revenue is nowhere near where the bulls want it to be. Positive catalysts will be somewhat binary until Keel secures its first billion-dollar customer.
Applied Digital (NASDAQ: APLD)
AI Infrastructure Without the GPUs
(Image Credit: Applied Digital)
Applied Digital doesn’t have neoclouds. The company essentially builds data centres and leases them out to cloud companies. Applied Digital has made some investments in GPUs over the years, but its core business is providing space, power, and cooling to the cloud infrastructure market rather than renting GPUs itself.
In Q3 2026 APLD reported year-over-year revenue growth of 139%. Applying Digital doesn’t have the lofty valuation of CoreWeave or Nebius due in large part to its shorter operating history. Investors willing to bet on future growth may find an entry point here.
Risk: Growth is impressive, but slower than peers. Applied Digital’ landlord business model means less pricing power and lower margins than Nebius or CoreWeave.
How to Think About Position Sizing
It’s probably best to think of AI infrastructure investing as a thematic allocation to the buildout of AI at scale rather than investing in individual companies.
Few investors could rightly own all five names above. They don’t have enough capital.
Position sizing is difficult with names this early in their growth cycle. One bad earnings report or downturn in AI spending can erase gains quickly.
Unlike large cap tech or even semiconductors, there’s a lot that can go wrong for AI infrastructure companies between today and their revenue projections for 2027 and beyond.
Don’t put money in these stocks you aren’t willing to lose. Even if you do your homework.
Do yourself another favour and diversify across names. If you insist on picking individual stocks over picking a theme with near limitless potential, at least invest in more than one company.
Final Thoughts
AI infrastructure stocks are risky. High reward. But risky.
They offer exposure to one of the biggest technology buildouts in history, but are not stocks for the faint of heart. We recommend approaching these stocks as an ultra-high conviction theme rather than singling out individual companies.
If you’re going to pick names, consider diversifying your bets across established (CoreWeave), pure-play (Nebius), transformational (IREN), super-early stage (KEEL), and non-GPU exposed (Applied Digital).
Before buying any stock, do your own research. Risk capital only.
As always, this post is for informational and educational purposes only. It does not constitute financial advice. Please consult a licensed financial adviser before making any investment decisions.
Keep learning: courses on this topic
Self-paced courses to put this article into practice.
Udemy
How To Invest In Stocks In 15 Minutes?
This is for anyone who wants a quick lesson on stocks and want to start trading stocks today—Within 15 minutes. This beginner guide will…
View course
Udemy
Investing For Beginners: How To Invest With Confidence
This Course is for you if ...1. You know nothing about investing.2. You are new to stock market investing and looking for a place to…
View course
Udemy
Personal Finance from Zero: Budget, Save, Borrow & Invest
Money touches every day of your life, yet almost nobody is taught how it works. Where does your pay go between payday and the end of the…
View course
Leave a Comment