10 Publicly Traded AI Companies to Research in 2026 (Not Investment Advice)

People often ask for the 'best AI stocks'. We will not answer that question, and the reason is part of the lesson. A list of best stocks is a forecast, and a forecast needs a tested method that we do not have. What we can do honestly is show where AI money and spending appear in company reports, and teach you to read those reports yourself.

Below are ten listed companies at different layers of AI: the cloud platforms that rent out computing, the chip designers, a networking and custom-chip maker, a specialist AI cloud, a data and AI software company and an enterprise database and cloud group. Each has one dated fact from its own statement. The facts are not equal in kind. Some are revenue, some are spending, some are contracted backlog. Compare them with care.

How do you research a public AI company yourself?

Which ten companies are worth reading about?

What do the cloud platforms report?

Microsoft (MSFT). In its fiscal fourth-quarter release of July 29, 2026, Microsoft said Azure and other cloud services revenue increased 43%, and quoted its chief executive saying Azure revenue passed $100 billion in a year for the first time. The same release reported that Microsoft 365 Copilot had reached over 30 million paid seats. The cash flow statement shows additions to property and equipment of $35.8 billion in the quarter and $115.9 billion for the fiscal year. Caution: the release gives no capital expenditure sentence, so the spending figure comes from a statement table, and 'Azure' is a segment the company defines.

Alphabet (GOOGL). Alphabet's second-quarter release of July 22, 2026 said Google Cloud saw a meaningful acceleration as revenues increased 82% to $24.8 billion. Its cash flow statement shows purchases of property and equipment of $44.924 billion for the quarter. The release also says a June 2026 equity raise was for general corporate purposes, including capital expenditures to scale AI infrastructure. Caution: a growth rate of 82% is measured from a smaller base, and the release does not itself break out how much is AI.

Amazon (AMZN). Amazon's second-quarter release of July 30, 2026 said AWS segment sales increased 37% year over year to $42.2 billion, with AWS operating income of $16.6 billion, against $10.2 billion in the second quarter of 2025. Its chief executive said the company's AI and chips businesses each passed run rates of more than $25 billion. The release also reports trailing-twelve-month free cash flow as an outflow of $7.6 billion. Caution: 'run rate' is the company's own annualised measure, not reported revenue.

Meta Platforms (META). Meta's second-quarter release of July 29, 2026 said capital expenditures, including principal payments on finance leases, were $31.08 billion. Its chief executive said AI is accelerating the core business today. Caution: Meta sells advertising, not AI directly, so its spending is an investment in products, not a revenue line you can read off the page.

What do the chip and hardware companies report?

NVIDIA (NVDA). For the quarter ended July 26, 2026, NVIDIA's release of August 26 reported revenue of $96.2 billion and Data Center revenue of $89.0 billion, up 117% from a year earlier. Its chief executive said Vera Rubin, now in full production, was built for this moment. Caution: the risk factors in the same release note NVIDIA's reliance on third parties to manufacture, assemble, package and test its products.

AMD (AMD). AMD's second-quarter release of August 4, 2026 reported Data Center segment revenue of $6.7 billion, up 107% year over year, and said it launched the Instinct MI400 Series and the Helios rack-scale solution. Caution: 'the world's most powerful AI server rack' is a vendor claim we cannot test, and the release reports gross margin and earnings per share on both GAAP and non-GAAP bases, which differ.

Broadcom (AVGO). Broadcom's third-quarter release of September 2, 2026 reported revenue of $29.6 billion, up 86% from the prior year, and quoted its chief executive saying AI semiconductor revenue of $16.7 billion grew 221% year over year. Caution: 'AI semiconductor revenue' is Broadcom's own category, so check its definition before setting it beside another company's AI figure.

What do the specialists and software companies report?

Oracle (ORCL). Oracle's first-quarter release of September 10, 2026 said remaining performance obligations were up $209 billion year over year to $664 billion, and cloud infrastructure revenue rose 121% to $7.4 billion. It also said free cash flow was negative $5 billion for the quarter, and its cash flow statement lists capital expenditures of $28.499 billion against $8.502 billion a year earlier. We read this from a verbatim copy of Oracle's release distributed by PR Newswire, because Oracle's own page blocked our reader. Caution: 'remaining performance obligations' are contracted but not yet recognised revenue, and the company says demand is growing faster than supply.

CoreWeave (CRWV). CoreWeave's second-quarter release of August 11, 2026 said revenue backlog was approximately $104 billion as of June 30, 2026. Its statement of operations shows revenue of $2.575 billion and a net loss of $626 million for the quarter. The company defines revenue backlog to include remaining performance obligations plus other amounts it estimates will be recognised under committed contracts. Caution: backlog is not revenue, and the same word is defined differently by different companies.

Palantir (PLTR). Palantir's second-quarter release of August 3, 2026 said revenue grew 93% year over year to $1.935 billion, with U.S. commercial revenue up 149% to $764 million, and that it closed 220 deals of at least $1 million. We read this from a copy of the release reproduced on a filings aggregator, which links the 8-K, not from Palantir's own page. Caution: a deal count says nothing about how long contracts last or whether they renew.

What do these companies have in common?

• Spending is large and rising. Several reported capital spending in the tens of billions of dollars in a single quarter, and two (Amazon on a trailing-twelve-month basis, Oracle in the quarter) reported negative free cash flow.
• Backlog is a promise, not a payment. Alphabet, Oracle and CoreWeave each cite contracted future revenue. The three define it in their own words.
• Everyone sells to everyone. The cloud platforms buy chips from the chip companies and also design their own, which makes the supply chain circular and hard to read from one filing.
• 'AI' is a label companies choose. A line item called AI at one company may include products another company does not count.

For the hardware side in more depth, see our companion piece on AI hardware and chip companies, and for the wider field our list of companies shaping AI.

How should you read an AI claim in an earnings release?

Treat an earnings release as a vendor claim. Our guide to vendor claims versus verified findings explains why a company statement is not an independent finding, and the same logic applies to performance claims such as 'most powerful'. For the habit of asking what was measured and by whom, see reading benchmark claims critically.

• Find the definition of any measure labelled 'AI', 'backlog' or 'run rate'.
• Separate reported results (what happened) from outlook statements (what the company hopes).
• Note whether a figure is GAAP or non-GAAP.
• Look for the date. A fact from last quarter may already be out of date.

What are the limits of this list?

• It is a selection by layer, not a measure of importance, and ten names cannot cover a field of thousands.
• Each fact is the company's own statement. We did not audit any figure, and we did not test any product.
• Oracle and Palantir were read from copies of the companies' releases, because their own pages could not be read by our tools. We say so in each entry.
• Facts age quickly. Company results change every quarter, and the next reports are due soon. Check the latest filing before relying on anything here.
• Nothing in this article is a prediction. If you spot an error, tell us through our corrections page.