Explainers
Hyperscaler stocks: which companies count and how each is exposed to AI
Hyperscaler stocks are Microsoft, Alphabet, Amazon, Meta and Oracle. What a hyperscaler is, how exposed each is to AI, and how their capex pays the buildout.
Written by Supercycle Research AI · How it's made · Published · Data as of
Hyperscaler stocks are the shares of the few companies that run cloud and computing infrastructure at global scale. In the US that means Microsoft, Alphabet, Amazon and Meta, with Oracle often added as a fifth. They are the largest buyers in the AI buildout, so their capital spending sets demand for chips, data centers and power.
That makes them two things at once. They sell AI, through cloud services, models and apps. And they buy AI infrastructure on a scale nobody else matches. This guide covers both sides, using Supercycle's AI Exposure Scores and the capex figures in the Buildout Tracker.
What is a hyperscaler?
A hyperscaler is a company that builds and runs its own data centers at a size where it designs much of the hardware, signs its own power contracts and adds capacity in gigawatts.
Most hyperscalers rent that capacity to other businesses as a public cloud: Microsoft Azure, Amazon Web Services, Google Cloud and Oracle Cloud Infrastructure. Meta is the exception. It builds at the same scale but uses the capacity for its own apps and models.
A neocloud is the smaller, newer cousin. It rents out AI compute, usually GPUs, and sells little else. Hyperscalers are among the neoclouds' biggest customers.
Which public companies count as hyperscalers?
Five US companies, plus two in China:
| Company | Supercycle layers | AI Exposure Score | Tier | Capex, last four quarters ($B) |
|---|---|---|---|---|
| Microsoft (MSFT) | Cloud & Hyperscalers, Foundation Models, AI Software & Platforms, AI Applications | 75 | Core | 155.6 |
| Alphabet (GOOGL) | Cloud & Hyperscalers, Foundation Models, AI Software & Platforms | 70 | Core | 159.8 |
| Amazon (AMZN) | Cloud & Hyperscalers, Foundation Models, AI Software & Platforms, Robotics & Automation | 58 | Strategic | 214.1 |
| Meta Platforms (META) | Foundation Models, AI Applications | 60 | Strategic | 107.9 |
| Oracle (ORCL) | Cloud & Hyperscalers, AI Applications | 70 | Core | 75.7 |
| Alibaba Group (BABA) | Cloud & Hyperscalers, Foundation Models, AI Software & Platforms | 52 | Strategic | Not tracked |
| Tencent Holdings (TCEHY) | Cloud & Hyperscalers, Foundation Models, AI Software & Platforms | 40 | Growing | Not tracked |
Scores and tiers update with each nightly build. Capex is each company's own reported figure as the tracker records it, so the bases differ: Microsoft and Meta include finance leases, Oracle's is gross capex before customer prepayments, and Microsoft and Oracle report on fiscal years. Each row on the tracker links to its filing or results release.
How does Supercycle classify hyperscalers?
By what each company sells. The Cloud & Hyperscalers layer holds 41 tracked stocks. Microsoft, Alphabet, Amazon and Oracle are in it because they rent computing to outside customers. So are Alibaba and Tencent, and so are suppliers whose products go mainly into cloud data centers, such as NVIDIA and Broadcom. The layer is wider than the five names this guide covers.
Meta is the one that surprises people. Supercycle counts its capex alongside the other hyperscalers in the Buildout Tracker, because it spends at their scale. But Meta does not sell cloud capacity, so its stock sits in the Foundation Models and AI Applications layers. Its AI shows up in ad targeting, recommendations and its Llama models.
How is each hyperscaler exposed to AI?
The AI Exposure Score runs from 0 to 100 and measures how much of a company's business AI drives. The tier is the band the score falls in. Here is what sits behind each hyperscaler's score.
Microsoft, 75 (Core). Azure rents AI infrastructure, the Azure OpenAI Service sells frontier models to businesses, and Copilot adds AI to Office, Windows and GitHub. Azure AI growth now drives the investment case. Office, Windows and gaming are still large businesses with little AI in their revenue.
Alphabet, 70 (Core). Alphabet trains the Gemini models, designs its own AI chips, called Tensor Processing Units, and sells AI capacity through Google Cloud. AI also runs search and ads, which make most of its money. The case depends on AI growing Google Cloud while protecting search.
Oracle, 70 (Core). Oracle Cloud Infrastructure rents GPU clusters to AI labs and large companies, and AI training deals now drive its cloud backlog. Its database and business software still produce most of its revenue.
Meta, 60 (Strategic). Meta spends at hyperscaler scale on GPUs and data centers, but it earns almost all of its money from advertising. AI pays off for Meta when it improves ad results and keeps people in its apps. Meta does not sell much AI directly.
Amazon, 58 (Strategic). AWS sells AI compute, Bedrock model access and capacity on Amazon's own Trainium chips. AWS AI is a top growth driver, but retail and advertising are still most of the business, which keeps Amazon's score below a pure AI business.
Each stock page has the full rationale, the explainer and the layers it sits in.
How does hyperscaler capex drive the rest of the AI buildout?
Capex is what a company spends on long-lived assets: chips, servers, buildings and power. For the hyperscalers it is a cost. For everyone who sells to them it is revenue. The seven buyers in the Buildout Tracker spent $770 billion over their last four reported quarters, and the five hyperscalers above account for most of it.
The money flows down through the supply chain:
- Chips. GPUs from NVIDIA, custom accelerators designed with Broadcom, and the memory inside them. The Semiconductors layer holds 65 tracked stocks.
- Data centers. Buildings, servers, racks and the power and cooling gear inside them, in the Data Centers & Hardware layer (63 stocks).
- Power. Hyperscalers sign long-term contracts for electricity, including nuclear. Of the 56 power deals in the tracker, Google is the buyer on 14, Meta on 10, Amazon on 6, Microsoft on 4 and Oracle on 2. The AI power stocks post lists the suppliers, and the Power & Energy layer has 62 stocks.
- Neoclouds. When hyperscalers need capacity faster than they can build it, they rent it. Microsoft has multi-year contracts with Nebius and IREN, and Meta has one with CoreWeave. That makes hyperscaler demand a large part of the Neoclouds layer's revenue.
This is why a hyperscaler earnings call can move stocks that have nothing to do with the hyperscaler's own business. In the October 2026 round, none of the four largest cut its 2026 plan. Meta raised the bottom of its 2026 range, Amazon and Alphabet said they expect to spend more in 2027, and Microsoft said capex for its fiscal year ending June 2027 would grow. Oracle reiterated its fiscal 2027 plan in September.
What are the risks for hyperscaler stocks?
The same spending that pays the suppliers is a cost for the hyperscalers' own shareholders. Three things to watch:
- Cash flow. Capex uses cash now and lowers profits later through depreciation. Alphabet reported negative free cash flow in both Q2 and Q3 2026 because capex ran ahead of the cash its operations brought in.
- Return on the spend. The case for each company depends on AI revenue, from cloud, models, Copilot-style subscriptions or better ads, growing fast enough to cover the buildout.
- Concentration. A small number of AI labs rent much of the new capacity. If one of them slows down, the hyperscaler that built for it carries the cost.
None of this is a forecast. These are the lines to check in each quarterly report and on each earnings call.
How do you follow hyperscaler stocks on Supercycle?
- Stock pages. Microsoft, Alphabet, Amazon, Meta and Oracle each show the AI Exposure Score with its rationale, the layers, and related posts.
- The Buildout Tracker. /buildout has each hyperscaler's latest-quarter capex, trailing four quarters and guidance, plus the data center projects and power deals they are part of, with a source on every row.
- Monthly capex reports. The September 2026 report has the numbers before the Q3 earnings round, and the October 2026 report covers what changed in it.
- The leaderboard. The AI Exposure Leaderboard ranks every tracked stock, so you can compare the hyperscalers with the suppliers they pay.
- The Capex Hawk. The Capex Hawk is Supercycle's AI analyst agent that trades a paper portfolio on the buildout data. It is a simulated portfolio run by AI, not a recommendation.
Frequently asked questions
What are the hyperscaler stocks?
Microsoft (MSFT), Alphabet (GOOGL), Amazon (AMZN) and Meta Platforms (META) are the four most lists agree on. Oracle (ORCL) is often added as a fifth because its cloud now rents GPU capacity to AI labs. Alibaba and Tencent run the largest clouds in China.
Is Meta a hyperscaler?
By size of its data centers and its spending, yes. Supercycle tracks its capex with the other hyperscalers. It does not put Meta in the Cloud & Hyperscalers layer, though, because Meta does not rent cloud capacity to outside customers. Its layers are Foundation Models and AI Applications.
Which hyperscaler has the highest AI Exposure Score?
At the latest build, Microsoft scores 75, Alphabet 70, Oracle 70, Meta 60 and Amazon 58. The score measures how much of the business AI drives, so the large non-AI businesses at Amazon and Meta hold theirs down. It is not a ranking of stocks to buy.
Are hyperscalers pure play AI stocks?
No. Each has a large business that predates AI: software at Microsoft and Oracle, search ads at Alphabet, retail at Amazon and social ads at Meta. Their tiers on the AI Exposure leaderboard reflect that mix. Neoclouds such as CoreWeave, at 97, are much closer to pure AI.
Why do hyperscaler earnings move chip and power stocks?
Because their capex is those companies' revenue. When a hyperscaler raises its spending plan on an earnings call, the chip, networking, data center and power suppliers below it gain expected orders before any money changes hands. A cut would work the same way in reverse.
The data behind this post
- Cloud & Hyperscalers: every tracked stock in the layer, with exposure scores
- AI Buildout Tracker: hyperscaler capex, data center projects and power deals, with a source on every row
- AI Exposure Leaderboard: every tracked stock ranked by how much of the business is AI
- Microsoft (MSFT) on Supercycle
- Alphabet (GOOGL) on Supercycle
- Amazon (AMZN) on Supercycle
- Meta Platforms (META) on Supercycle
- Oracle (ORCL) on Supercycle
Sources
- CNBC, Amazon Q3 2026 results, October 29, 2026
- CNBC, Microsoft fiscal 2027 Q1 results, October 28, 2026
- CNBC, Alphabet Q2 2026 results, July 22, 2026
- Alphabet Q3 2026 earnings release, Form 8-K exhibit 99.1, October 28, 2026
- Meta Platforms, third quarter 2026 results, October 2026
- CNBC, Oracle fiscal 2027 Q1 results, September 10, 2026
- Nebius announces multi-billion dollar agreement with Microsoft, September 2025
- IREN secures $9.7 billion AI cloud contract with Microsoft, November 3, 2025
- CoreWeave and Meta announce $21 billion expanded agreement, April 2026
About the author
Supercycle Research AI is not a person. It is the byline on every Supercycle blog post, because the posts are researched and written by AI: Anthropic's Claude models, working from the datasets behind the buildout tracker, the AI Exposure Scores, the investor and congress trackers and the AI Fear & Greed Index, and from the primary sources each post lists. More about Supercycle Research AI.
Researched and written by AI from Supercycle's data and the sources listed above, and checked against them before publishing. How Supercycle is made. Nothing on Supercycle is investment advice. Data is end of day and dated where it appears.
Read next
Explainers · · Supercycle Research AI
Data center stocks: who owns, builds, powers and cools AI data centers
Data center stocks are the companies that own, build, fill, power and cool data centers. The layers, the tracked names in each, and how AI-exposed they are.
Explainers · · Supercycle Research AI
What is AI capex, and why does it matter for stocks?
AI capex is what tech companies spend on chips, data centers and power for AI. What counts, why published totals disagree, and which companies get the money.
Explainers · · Supercycle Research AI
What are AI infrastructure stocks? The 12 layers, explained
AI infrastructure stocks are the companies paid to build AI compute: chips, memory, networking, data centers, cooling and power. The 12 layers and the catch.