Company pages · from SEC filings
The companies behind the AI build-out
53 of the most-watched AI companies, at least three in every layer, each read from its own filings: what it does in the AI stack, how fast it is growing and how much it keeps, who its customers are, the financing links it has with other companies, and the commitments that don’t show up in its income statement.
AI clouds: the buyers
Microsoft, Amazon, Alphabet, Meta and Oracle, plus the GPU-only clouds: the companies whose capital spending is everyone else’s revenue. Read the layer analysis.
Microsoft MSFT
Azure, Microsoft 365 and OpenAI: the buyer with the largest contracted backlog and the clearest cloud return so far.
$346B revenue · +18% · 68% gross margin
Amazon AMZN
AWS, Trainium and Anthropic: the biggest spender in the stack, now spending more than it generates.
$776B revenue · +16% · 51% gross margin
Alphabet GOOGL
Google Cloud and TPUs: the buyer whose cloud profits are growing fastest.
$471B revenue · +22% · 61% gross margin
Meta META
Building for its own apps: a buyer with no cloud business to sell capacity through, and large off-balance-sheet commitments.
$242B revenue · +28% · 81% gross margin
Oracle ORCL
The largest contracted backlog in the stack, and the first big buyer spending more than its business generates.
$71.8B revenue · +22% · — gross margin
CoreWeave CRWV
The largest listed GPU-only cloud: fast growth, borrowed capital and a few very large customers.
$7.6B revenue · +115% · 67% gross margin
Servers and systems
Dell, Supermicro, HPE and the contract manufacturers who turn chips into racks, carrying huge revenue on thin margins. Read the layer analysis.
Dell DELL
The largest server maker in the stack, turning AI volume into operating profit.
$151B revenue · +49% · 20% gross margin
Supermicro SMCI
GPU servers and liquid-cooled racks: fast growth that consumes cash.
$39.1B revenue · +78% · 11% gross margin
HPE HPE
Servers, AI systems and, since Juniper, data-center networking: low margins, a growing AI backlog.
$41.9B revenue · +27% · — gross margin
Chips: accelerators and custom silicon
Nvidia, Broadcom, AMD, Marvell and the CPU makers: the toll booth where most of the build-out’s profit has collected so far. Read the layer analysis.
Nvidia NVDA
The largest seller in the AI stack: accelerators, networking and full systems, with software-like margins on hardware.
$303B revenue · +83% · 75% gross margin
Broadcom AVGO
Custom AI accelerators and Ethernet switch chips for a few very large customers.
$89.1B revenue · +49% · 69% gross margin
AMD AMD
The main merchant challenger to Nvidia, winning orders partly with stock warrants.
$41.3B revenue · +40% · 53% gross margin
Arm ARM
The CPU designs inside most custom cloud chips: a small revenue line with a toll on nearly every one.
$4.9B revenue · +23% · 98% gross margin
Marvell MRVL
Custom silicon and optical DSPs for cloud data centers.
$9.5B revenue · +31% · 52% gross margin
Memory and storage
Micron, Sandisk, Western Digital and Seagate: high-bandwidth memory turned the industry’s most cyclical layer into the build-out’s second toll booth. Read the layer analysis.
Micron MU
The only US-listed maker of high-bandwidth memory, in a memory cycle like no other.
$133B revenue · +256% · 81% gross margin
Sandisk SNDK
NAND flash for data-center SSDs, now selling much of its future output under multi-year agreements.
$20.2B revenue · +175% · 71% gross margin
Western Digital WDC
Hard drives for the bulk data AI is trained and served from, with most revenue from a few cloud buyers.
$14.5B revenue · +43% · 53% gross margin
Networking and optics
Arista, Cisco, Ciena, Coherent, Lumentum, Credo, Astera Labs, Amphenol and Corning: the switches, optics and cables that make thousands of GPUs act as one. Read the layer analysis.
Arista ANET
Ethernet switches for AI networks, with software-backed margins.
$10.5B revenue · +33% · 63% gross margin
Cisco CSCO
The largest networking company on the map, selling switches, routers and its own silicon to hyperscalers and enterprises.
$63.3B revenue · +12% · 65% gross margin
Credo CRDO
Copper and optical connections inside AI clusters: the fastest grower in networking, with very concentrated customers.
$1.6B revenue · +165% · 67% gross margin
Power, cooling and data centers
Vertiv, Eaton, GE Vernova, Quanta, the nuclear generators and the data-center landlords: the slow, physical layer that sets how fast everything else can grow. Read the layer analysis.
Vertiv VRT
Power distribution and liquid cooling for AI racks.
$13.1B revenue · +35% · 38% gross margin
Constellation CEG
The largest US nuclear generator, selling long-term power to data centers.
$24.8B revenue · +20% · — gross margin
GE Vernova GEV
Gas turbines and grid equipment: a backlog that runs for years as data centers chase power.
$44.0B revenue · +17% · 22% gross margin
Foundry, equipment and design tools
TSMC, ASML, Applied Materials, Lam Research, KLA, Teradyne, Amkor, Synopsys and Cadence: the companies every chip has to pass through. Read the layer analysis.
TSMC TSM
The foundry that makes nearly every leading AI chip, whoever designs it.
$121B revenue · +38% · 60% gross margin
ASML ASML
The sole supplier of EUV lithography: a toll on every leading-edge chip factory.
€32.7B revenue · +16% · 53% gross margin
Lam Research LRCX
Etch and deposition tools, with heavy exposure to the memory makers AI has turned into buyers again.
$26.1B revenue · +33% · 51% gross margin
Software and applications: the return
Salesforce, ServiceNow, Palantir, Adobe and the rest: the layer whose customers have to pay for AI if the spending below is to earn a return. Read the layer analysis.
Palantir PLTR
Data and AI platforms for governments and enterprises, with growth and margins that stand out in software.
$6.2B revenue · +79% · 85% gross margin
ServiceNow NOW
Workflow software for large enterprises, selling AI agents into its own platform.
$15.4B revenue · +22% · 73% gross margin
Salesforce CRM
The largest application company on the map, betting on Agentforce to turn AI agents into revenue.
$43.9B revenue · +11% · 77% gross margin
Adobe ADBE
Creative and document software with its own generative models: the clearest test of whether AI widens a software franchise or fills it in.
$26.0B revenue · +12% · 89% gross margin
Intuit INTU
TurboTax, QuickBooks and Credit Karma: years of consumer and small-business financial data, now put to work by AI agents.
$21.4B revenue · +14% · — gross margin
Workday WDAY
HR and finance software priced on customers’ headcount, adding AI agents that may reduce it.
$10.2B revenue · +13% · — gross margin
Snowflake SNOW
The data cloud AI workloads read from, priced on consumption, so AI usage shows in revenue as soon as customers run it.
$5.4B revenue · +32% · 67% gross margin
Datadog DDOG
Monitoring software whose filings tie growth directly to AI-native customers, including its largest.
$4.0B revenue · +32% · 80% gross margin
MongoDB MDB
A developer database adding vector search and retrieval models, betting AI applications run on general-purpose databases.
$2.8B revenue · +25% · 73% gross margin
CrowdStrike CRWD
Endpoint and cloud security that uses AI against attackers and sells protection for companies’ own AI.
$5.4B revenue · +24% · 75% gross margin
Palo Alto Networks PANW
Network, cloud and AI security sold as consolidated platforms, mostly through distributors.
$11.5B revenue · +24% · 70% gross margin
AppLovin APP
An ad platform run by a machine-learning engine: AI showing up directly in the income statement.
$6.8B revenue · +61% · 88% gross margin
Shopify SHOP
The commerce platform plugging merchants into AI assistants, with payments behind most of its revenue.
$13.3B revenue · +33% · 48% gross margin
Atlassian TEAM
Jira and Confluence, moving customers to the cloud where its AI, Rovo, runs.
$6.6B revenue · +26% · 85% gross margin
HubSpot HUBS
CRM and marketing software for mid-sized businesses, with AI built across the platform.
$3.4B revenue · +21% · 83% gross margin
Zscaler ZS
Cloud security that inspects traffic between users, AI agents and applications.
$3.4B revenue · +25% · 77% gross margin
Okta OKTA
Identity and access management, pitching itself as the control plane for AI agents’ identities.
$3.1B revenue · +11% · 78% gross margin
Dynatrace DT
Observability for large enterprises, built on one data store its AI reasons over.
$2.1B revenue · +18% · 81% gross margin
Twilio TWLO
The messaging and voice APIs AI agents talk to customers through.
$5.6B revenue · +18% · 49% gross margin
GitLab GTLB
A DevSecOps platform betting AI moves software’s bottleneck from writing code to shipping it.
$1.1B revenue · +23% · 86% gross margin
Elastic ESTC
Elasticsearch: the search engine many companies already run, now pitched as a vector database for AI retrieval.
$1.8B revenue · +16% · 76% gross margin
UiPath PATH
Automation software moving from scripted bots to AI agents, and exposed if agents can operate software on their own.
$1.7B revenue · +15% · 83% gross margin
C3.ai AI
Enterprise AI software from before the generative boom, with shrinking revenue and concentrated customers.
$232M revenue · −38% · 29% gross margin
Pegasystems PEGA
Workflow and decisioning software for large enterprises, using generative AI to design applications.
$1.7B revenue · +1% · 76% gross margin
SoundHound AI SOUN
Voice AI for carmakers, restaurants and customer service, grown largely through acquisitions.
$203M revenue · +55% · 42% gross margin
Teradata TDC
The incumbent enterprise data platform, moving to the cloud while arguing AI belongs next to on-premises data.
$1.7B revenue · +1% · 61% gross margin
CCC Intelligent Solutions CCC
Insurance-claims software built on decades of repair-estimating data, with AI estimating from photos.
$1.1B revenue · +11% · 74% gross margin
The 53 at a glance
The fastest growers over their latest four quarters were Micron (+256%), Sandisk (+175%), Credo (+165%). The highest gross margins belong to Arm (98%), Adobe (89%), AppLovin (88%), the clearest sign of where scarcity and pricing power sit. The most capital-intensive, measured by capex as a share of revenue, are CoreWeave, Oracle and Meta, and Amazon, Oracle and CoreWeave spent more on capital projects than they generated from operations. Supermicro, C3.ai and SoundHound AI used more cash in their operations than they took in.
| Company | Layer | Revenue | Growth | Gross margin | Free cash flow |
|---|---|---|---|---|---|
| Amazon | AI clouds | $776B | +16% | 51% | −$11.6B |
| Alphabet | AI clouds | $471B | +22% | 61% | $19.7B |
| Microsoft | AI clouds | $346B | +18% | 68% | $60.0B |
| Nvidia | Chips | $303B | +83% | 75% | $127B |
| Meta | AI clouds | $242B | +28% | 81% | $28.6B |
| Dell | Servers | $151B | +49% | 20% | $8.6B |
| Micron | Memory | $133B | +256% | 81% | $59.0B |
| TSMC | Foundry & tools | $121B | +38% | 60% | $32.0B |
| Broadcom | Chips | $89.1B | +49% | 69% | $39.4B |
| Oracle | AI clouds | $71.8B | +22% | — | −$28.7B |
| Cisco | Networking | $63.3B | +12% | 65% | $12.8B |
| GE Vernova | Power & DCs | $44.0B | +17% | 22% | $17.1B |
| Salesforce | Software | $43.9B | +11% | 77% | $15.2B |
| HPE | Servers | $41.9B | +27% | — | $4.2B |
| AMD | Chips | $41.3B | +40% | 53% | $8.4B |
| Supermicro | Servers | $39.1B | +78% | 11% | −$7.0B |
| ASML | Foundry & tools | €32.7B | +16% | 53% | €11.1B |
| Lam Research | Foundry & tools | $26.1B | +33% | 51% | $6.5B |
| Adobe | Software | $26.0B | +12% | 89% | $10.6B |
| Constellation | Power & DCs | $24.8B | +20% | — | $309M |
| Intuit | Software | $21.4B | +14% | — | $8.7B |
| Sandisk | Memory | $20.2B | +175% | 71% | $11.5B |
| ServiceNow | Software | $15.4B | +22% | 73% | $4.8B |
| Western Digital | Memory | $14.5B | +43% | 53% | $4.3B |
| Shopify | Software | $13.3B | +33% | 48% | $2.4B |
| Vertiv | Power & DCs | $13.1B | +35% | 38% | $3.1B |
| Palo Alto Networks | Software | $11.5B | +24% | 70% | $4.1B |
| Arista | Networking | $10.5B | +33% | 63% | $5.2B |
| Workday | Software | $10.2B | +13% | — | $2.8B |
| Marvell | Chips | $9.5B | +31% | 52% | $1.7B |
| CoreWeave | AI clouds | $7.6B | +115% | 67% | −$13.7B |
| AppLovin | Software | $6.8B | +61% | 88% | — |
| Atlassian | Software | $6.6B | +26% | 85% | $1.3B |
| Palantir | Software | $6.2B | +79% | 85% | $3.4B |
| Twilio | Software | $5.6B | +18% | 49% | — |
| Snowflake | Software | $5.4B | +32% | 67% | $1.2B |
| CrowdStrike | Software | $5.4B | +24% | 75% | $1.6B |
| Arm | Chips | $4.9B | +23% | 98% | $979M |
| Datadog | Software | $4.0B | +32% | 80% | $1.2B |
| HubSpot | Software | $3.4B | +21% | 83% | $797M |
| Zscaler | Software | $3.4B | +25% | 77% | $852M |
| Okta | Software | $3.1B | +11% | 78% | $979M |
| MongoDB | Software | $2.8B | +25% | 73% | $659M |
| Dynatrace | Software | $2.1B | +18% | 81% | $571M |
| Elastic | Software | $1.8B | +16% | 76% | $349M |
| Pegasystems | Software | $1.7B | +1% | 76% | $442M |
| UiPath | Software | $1.7B | +15% | 83% | $363M |
| Teradata | Software | $1.7B | +1% | 61% | $736M |
| Credo | Networking | $1.6B | +165% | 67% | $439M |
| CCC Intelligent Solutions | Software | $1.1B | +11% | 74% | — |
| GitLab | Software | $1.1B | +23% | 86% | $214M |
| C3.ai | Software | $232M | −38% | 29% | −$156M |
| SoundHound AI | Software | $203M | +55% | 42% | −$116M |
Latest four quarters for 10-Q filers; latest fiscal year for TSMC, ASML (in euros) and Arm.
How to read a company page
Every page follows the same structure, so companies can be compared on the same terms. It opens with the company’s revenue, growth and margins over its latest four reported quarters (or its latest fiscal year for annual filers), then describes what the business does in the stack and what its filings show. Charts and tables give the quarter-by-quarter figures, with gross and operating margin for each quarter, so you can see whether a company keeps more of each dollar as it grows or passes it through to its own suppliers.
The middle of each page places the company among the others in its layer, with a peer table and its rank by revenue and growth. Then come the disclosures that matter most in an investment cycle built on a handful of buyers: customer concentration, quoted from the filing; financing links, such as investments in customers, guarantees and warrants; and commitments outside the income statement, including backlog, purchase obligations, leases signed but not yet started and the useful lives used to depreciate servers.
Each page ends with what to watch in the next filings. The aim is not to tell you whether a stock is cheap or expensive (the pages don’t look at valuation) but to make sure you know what the company itself reports before you form a view.
How the companies connect
These companies are each other’s customers and suppliers, and increasingly each other’s financiers. Microsoft, Amazon, Alphabet, Meta and Oracle spend on the chips, memory, servers, networking and power the others sell. Nvidia invests in and buys capacity back from the AI clouds that buy its chips; Amazon finances Anthropic as AWS delivers compute; AMD gives OpenAI and Meta warrants tied to its accelerators; CoreWeave depends on Microsoft for most of its revenue and on Nvidia for all of its GPUs. TSMC and ASML sit underneath all of them.
That is why a company page links to the money-loop map and to its layer. The strength of any one of these businesses depends on the spending decisions of a few of the others, and the capital-flow analysis shows how much of each new dollar each layer keeps.
Why these companies
We started with the AI companies investors search for most and whose filings tell the richest story: the five largest buyers and the leading GPU cloud, the main chip designers, the foundry and the lithography supplier every chip depends on, the memory maker at the center of the HBM cycle, the two largest AI server makers, the leading AI networking supplier, the largest power-equipment and nuclear names in the build-out, and the software company whose numbers show AI most clearly. We then made sure every layer has at least three, so each can be read through companies that make different bets: flash and hard drives alongside DRAM in memory, the largest incumbent and a fast-growing specialist in networking, HPE beside Dell and Supermicro, the largest gas-turbine maker in power, Lam Research for the toolmakers most tied to memory, and ServiceNow and Salesforce for the incumbents selling AI agents. Then we covered the whole software layer, from Adobe to SoundHound AI, because software is where the return on all the spending below it has to appear, and because the reasons a software company grows or doesn’t differ more from one company to the next than in any other layer. Every other company on the map appears in its layer’s table, and more company pages will follow where the filings support a page of real substance. Readers can suggest a company through the contact form on the work-with-us page; we add one when its filings can carry the same depth of analysis as the ones here. We read every suggestion.
Free · quarterly · from SEC filings
Get the AI Stack Report every quarter
After each Big Tech earnings season we update the map from the new filings and publish what changed: who spent, who got paid, who kept it and what the filings reveal about the loop. Free, in Rob May’s Investing in AI newsletter.
Form not showing? Subscribe on Substack. No spam; unsubscribe any time.
Questions about the company pages
Which AI companies have their own pages?
Nvidia, Microsoft, Amazon, Alphabet, Meta, Oracle, CoreWeave, Broadcom, AMD, TSMC, Micron, ASML, Palantir, Dell, Supermicro, Arm, Marvell, Arista, Vertiv, Constellation, Sandisk, Western Digital, Cisco, Credo, HPE, GE Vernova, Lam Research, ServiceNow, Salesforce, Adobe, Intuit, Workday, Snowflake, Datadog, MongoDB, CrowdStrike, Palo Alto Networks, AppLovin, Shopify, Atlassian, HubSpot, Zscaler, Okta, Dynatrace, Twilio, GitLab, Elastic, UiPath, C3.ai, Pegasystems, SoundHound AI, Teradata, CCC Intelligent Solutions. The first twenty were the most-searched AI names with the richest filings data; more were added so every layer of the stack has at least three, and then every company in the software layer, where the return on AI spending has to show up. The full list of companies on the map, with their numbers, is on each layer page.
What does each company page cover?
What the company does in the AI stack and what its filings show, its quarterly or annual revenue, the segments most tied to AI, how it compares with the other companies in its layer, its customer concentration, any financing links with other companies, its backlog and commitments, and what to watch in its next filings.
Will more companies get their own pages?
Yes, where the filings support a page of real substance: quarterly figures, segment data, customer disclosures and enough history to describe a trend. Every company on the AI stack map already appears, with its latest figures, in its layer’s table.
Are these stock recommendations?
No. The pages describe what each company discloses and where it sits in the AI build-out. They don’t consider valuation or anyone’s circumstances, and they don’t recommend buying or selling any security.
How often are the company pages updated?
Whenever the data is refreshed from SEC filings, normally after each earnings season. Figures in the text are computed from the same data as the tables, so they update together.
Why are some figures annual rather than quarterly?
TSMC, ASML and Arm file annual reports on Form 20-F rather than quarterly 10-Qs, so their pages use fiscal-year figures in the currency they report. ASML reports in euros.
Research and education, not investment advice: nothing here is a recommendation to buy or sell any security. Figures come from companies’ SEC filings as reported; companies named were chosen because they show how the stack works, not as a view on their shares.