Beak
Loading ..

table of contents feature [open]

How to Invest in AI Stocks in 2026: A Beginner's Guide

0 Comments
Investor reviewing a company annual report and stock charts next to a laptop showing a data centre, illustrating how to research AI stocks

Last reviewed: October 1, 2026

Short answer: "AI stocks" are shares of companies whose revenue depends, in whole or in part, on artificial intelligence: chip designers and manufacturers, cloud platforms, data-centre and power suppliers, model developers, software companies and businesses that use AI to run better. You can invest through individual shares, a broad index fund that already holds the largest AI-related companies, a thematic AI exchange-traded fund (ETF) or a mutual fund. Private AI companies are mostly limited to wealthy or professional investors.

A careful approach is to decide how much of your portfolio you will expose to one industry, read what companies file with the regulator instead of headlines, remember that high growth expectations make prices fragile, and keep costs, leverage and single-stock bets low. Nobody can guarantee a return, and offers that promise one are a warning sign.

This guide draws on the US Securities and Exchange Commission (SEC), FINRA, the CFTC, the IRS, SIPC, the European Commission, the US Bureau of Industry and Security, the US Department of Energy, the Federal Reserve system, GOV.UK, the FCA, the FSCS and the Government of Canada. We do not recommend any share, fund, broker or adviser, and we show no real prices, price targets or forecasts.

This article is general information, not investment advice. It is also not financial or tax advice. Rules differ by country and by account type, and your own circumstances matter.

What "AI stocks" means and the AI value chain

An AI stock is any publicly traded company that earns, or expects to earn, a meaningful share of its revenue from artificial intelligence. There is no official definition and no regulator maintains a list. The label is a marketing and analysis shortcut, not a legal category.

The most useful way to understand the sector is as a value chain: layers that each buy from the one below and sell to the one above. Spending flows down the chain and has to come back as revenue from end users. If end users do not eventually pay enough, every layer feels it.

Layer What these companies do Typical risk profile
Semiconductors and equipment Design or manufacture processors, memory and networking chips, or build the machines used in chip factories. Well-known examples of the category include Nvidia, AMD, TSMC and ASML. Highly cyclical, capital-intensive, exposed to export controls and to a small number of large buyers.
Cloud platforms ("hyperscalers") Operate very large data centres and rent computing, storage and AI services. Examples of the category include Microsoft, Alphabet and Amazon. Diversified revenue, but very heavy capital spending that must be justified by future demand.
Data-centre infrastructure and power Build or supply buildings, servers, cooling, networking gear, electrical equipment and electricity. Tied to the building cycle; exposed to grid, permitting and supply bottlenecks.
Foundation-model developers Train and run large general-purpose AI models. Many are private or sit inside larger groups; costs are high and business models are still developing.
Enterprise software Add AI features to business software such as productivity, design, customer service, security or data tools. Depends on customers paying more for AI features; competition can push prices down.
AI-enabled adopters Companies in any industry (retail, finance, health care, advertising, logistics) that use AI internally. AI is only one factor in results; the benefit is hard to measure from outside.

The company names in the table are neutral examples of a category only. They are not recommendations, and a company's place in the chain can change over time.

Each layer has its own economics, so before you invest in anything described as an AI stock, identify its layer and its customers.

Pure-play vs diversified AI exposure

A pure-play AI company gets most of its revenue from AI-related products, so its share price moves closely with the AI theme. A diversified company has AI as one business line among several, so the theme affects it less, in both directions.

  • Pure-play exposure is a concentrated bet. If AI demand pauses, there is no other business to cushion the fall.
  • Diversified exposure is diluted. A large technology group may have advertising, devices, office software or retail operations alongside AI, which can remain a minority of total revenue.

To tell which is which, look at segment reporting in the annual report, which breaks revenue and profit down by business line. If the company does not separate AI revenue from the rest, treat claims about how much of the business is "AI" with caution.

Many investors already have AI exposure without choosing it, because broad index funds are weighted by company size and the largest technology and chip companies are a sizeable part of them. See diversification.

Ways to invest in AI compared

There are five main routes: individual shares, broad index funds, thematic AI ETFs, mutual funds and private or venture investments. They differ mostly in diversification, cost and how much work and risk you take on yourself.

Route What you own Typical costs Concentration
Individual stocks Shares of specific companies you choose. Any trading commission, the bid-ask spread and, for foreign shares, currency conversion costs. No ongoing fund fee. Very high unless you hold many companies across industries.
Broad index funds (ETF or mutual fund) A slice of hundreds or thousands of companies, including the largest AI-related ones. An annual expense ratio, generally lower for index-tracking funds than for actively managed ones. Low at company level, but size-weighted indexes can lean heavily towards a few very large technology companies.
Thematic AI ETFs A basket of companies selected by an index provider or manager as AI-related. An expense ratio that is often higher than a broad index fund, plus the bid-ask spread. High at theme level; many hold the same handful of large companies.
Mutual funds (technology or growth funds) A professionally managed pool, priced once a day. An expense ratio and sometimes sales charges; actively managed funds usually cost more. Depends on the manager; technology funds are concentrated by design.
Private and venture exposure Shares in companies that are not listed on an exchange. Can be high and layered (fund fees, performance fees, platform fees). Very high; often a single company.

Individual stocks and broad index funds

Individual shares give you control and no ongoing fund fee, but the research and the company-specific risk are yours. FINRA defines concentration risk as having a large portion of your portfolio in a single investment, asset class or market segment, which can amplify losses. A handful of AI shares is concentrated by company and by industry.

A broad index fund is the simplest alternative. FINRA describes ETFs and mutual funds as pooled funds offering an interest in a professionally managed, diversified portfolio, and most ETFs passively track an index. Because the largest listed companies include major AI chip, cloud and software businesses, a broad fund already carries the theme without you having to pick winners.

Thematic AI ETFs

A thematic ETF holds companies that an index provider or manager classifies as AI-related, and there is no standard classification. One fund may be dominated by chips, another by software or robotics. Two funds with "AI" in the name can behave very differently.

FINRA suggests reading a fund's prospectus to see what it holds and to spot overlap with your other investments. Check:

  • The top ten holdings and their share of the fund.
  • How the index or manager decides what counts as an AI company.
  • The expense ratio. FINRA notes that actively managed ETFs typically carry higher expense ratios, which eat into returns over time.
  • The bid-ask spread, which FINRA says can be close to zero for some products and much wider for others.

Be especially careful with leveraged and inverse products that promise a multiple, or the opposite, of an index's move. FINRA cautions that they are generally not designed to be held for longer than their stated period, so they are not long-term investments.

Why costs matter: an illustrative example

Fees look small as percentages but compound. This example is fictional and illustrative only; the returns and fee levels are invented.

Suppose you invest 10,000 for 20 years and earn 6% a year before fees. With a fund charging 0.10% a year, the net return is 5.9% and the pot grows to about 31,472. With a fund charging 0.70%, the net return is 5.3% and the pot grows to about 28,091. The difference is roughly 3,381, about a third of the original investment.

Private and venture exposure, and why it is restricted

Some well-known AI developers are private companies. In the United States many private offerings rely on exemptions that limit participation to accredited investors. The SEC explains that individuals generally qualify through net worth above 1 million US dollars excluding their primary residence, income above 200,000 US dollars (300,000 with a spouse or partner) for two years, or certain professional licences.

Private companies do not publish the regular, detailed disclosures that listed companies do, and their shares cannot easily be sold. FINRA lists illiquid assets such as private placements as a source of concentration risk. If an unknown person offers you "pre-IPO" shares in a famous AI company, treat it as a red flag.

How to read an earnings report and SEC filings

The most reliable information about a US-listed company is in the documents it files with the SEC. Three forms matter most: the annual 10-K, the quarterly 10-Q and the 8-K for significant events. All are free on the SEC's EDGAR system.

Document When it appears What it tells you
Form 10-K (annual report) Once a year A detailed picture of the business, its risks and its audited financial results, plus management's discussion of what drove them.
Form 10-Q (quarterly report) For interim quarters Updated financial statements and management discussion during the year.
Form 8-K (current report) After significant events, often within four business days according to the SEC Events such as acquisitions, leadership changes and material agreements. Quarterly earnings press releases are commonly furnished this way.
Earnings press release and call Each quarter Headline numbers, management commentary and often guidance.

The 10-K: where to start

The SEC's investor bulletin "How to Read a 10-K" explains that 10-Ks follow a set order of topics. The company writes the report; the SEC sets the disclosure requirements but does not vouch for its accuracy. Materially false or misleading statements are prohibited, and the chief executive and chief financial officer must certify the report.

The sections most useful to an ordinary investor are:

  • Item 1, Business. Main products and services, markets, competition and regulation. The SEC calls it a good place to start.
  • Item 1A, Risk Factors. The most significant risks, generally in order of importance. Read it for specifics: dependence on a few customers or suppliers, export restrictions, energy, competition and regulation.
  • Item 7, Management's Discussion and Analysis (MD&A). Management's explanation of results, liquidity, capital resources and known trends or uncertainties.
  • Item 8, Financial Statements. The audited income statement, balance sheet, cash flow statement and notes. The SEC advises looking carefully if the auditor gives anything other than an unqualified opinion.

Segment reporting

Segment information, in the notes to the financial statements and the MD&A, splits revenue and operating profit by business line. It is the best check on AI claims. If a company says AI is transforming its business but the segment containing its AI products is small or shrinking, the numbers and the story do not match.

Capital expenditure

Capital expenditure ("capex") is money spent on long-lived assets such as data centres, servers and factories. It appears in the investing section of the cash flow statement, which the SEC's beginner's guide to financial statements says reports cash used to buy or sell long-term assets. For buyers of AI hardware, capex is the main cost of the build-out; for sellers, it is the source of revenue.

Guidance

Guidance is management's own estimate of future revenue, margins or spending, usually given with quarterly results. It is a forward-looking statement, not a promise. Share prices often react more to guidance than to the quarter just reported.

GAAP and non-GAAP numbers

US companies report under Generally Accepted Accounting Principles (GAAP). Many also highlight "non-GAAP" or "adjusted" figures that leave out certain costs, commonly share-based pay. The SEC bulletin notes that companies must show how such measures differ from the most comparable GAAP measure, and that it is up to you how much weight to give them.

A five-minute earnings read

Check six things in order: revenue growth against the same quarter a year earlier, margins, cash flow after capex, which segments produced the growth, guidance compared with last time, and any new risk factors.

Valuation basics, with a worked example

Valuation is the relationship between a share's price and what the business earns. A great company can still be a poor investment if the price already assumes more growth than it delivers. Valuation ratios do not tell you what will happen; they tell you what expectations are built into the price.

Measure How it is calculated What it tells you
Earnings per share (EPS) Net income divided by shares outstanding Profit attributable to each share.
Price-to-earnings (P/E) Share price divided by EPS over the past year How many times past earnings the price represents.
Forward P/E Share price divided by expected EPS for the next year What investors pay for expected earnings.
Price-to-sales (P/S) Market value divided by annual revenue Useful when a company has little or no profit yet.
Free cash flow (FCF) Cash from operations minus capital expenditure Cash left after funding the business's investment.
Operating margin Operating income divided by revenue How much of each unit of sales becomes operating profit.

The definitions of EPS, P/E and operating margin follow the SEC's beginner's guide to financial statements.

Growth versus margins

Two things drive what investors will pay: how fast earnings are expected to grow and how profitable each unit of revenue is. A fast-growing, high-margin company tends to trade on a high multiple, which only holds if growth and margins arrive as expected. When growth slows, expected earnings fall and the multiple investors will pay can fall too. That double effect is why fast-growing shares can drop far more than their earnings.

Illustrative worked example (fictional company)

The following example uses an invented company, "Northfield Compute", and invented numbers. It is for illustration only and does not describe any real company, price or forecast.

Starting position. Northfield Compute has 500 million shares, trading at 160 each. Over the past year it earned revenue of 10 billion and net income of 2 billion.

  • Market value: 500 million x 160 = 80 billion.
  • EPS: 2 billion / 500 million = 4.00.
  • Trailing P/E: 160 / 4.00 = 40.
  • Price-to-sales: 80 billion / 10 billion = 8.
  • Net margin: 2 billion / 10 billion = 20%.

Forward view. Investors expect EPS to grow 25% next year, to 5.00. The forward P/E is 160 / 5.00 = 32.

Free cash flow. Cash from operations was 3.0 billion and capital expenditure was 1.8 billion, so free cash flow was 1.2 billion. That is 1.5% of the 80 billion market value (a price-to-free-cash-flow ratio of about 67).

What the price assumes. A P/E of 40 only looks reasonable if growth continues. The table shows EPS after five years under two growth paths.

Scenario (illustrative) Annual EPS growth EPS after 5 years Price of 160 as a multiple of year-5 EPS
Growth meets expectations 25% About 12.21 About 13.1
Growth slows 10% About 6.44 About 24.8

If growth runs at 25%, the price paid looks modest in hindsight. If it runs at 10%, the buyer has paid nearly 25 times earnings five years out. The same price is cheap or expensive depending on a future nobody knows.

A guidance miss. Now suppose Northfield reports a solid quarter but guides to next-year EPS of 4.40 instead of the 5.00 investors expected. Expected growth has fallen from 25% to 10%, and the earnings estimate is 12% lower (4.40 / 5.00 = 0.88).

  • If investors still paid 32 times forward earnings, the price would be 4.40 x 32 = 140.80, a fall of 12%.
  • But slower growth usually earns a lower multiple. If investors now pay 25 times, the price would be 4.40 x 25 = 110, a fall of 31.25% from 160.

A 12% cut to expected earnings produced a fall of more than 31% because the multiple shrank as well. The company is still profitable and still growing; the price fell because expectations changed. The same mechanism works in reverse when guidance beats expectations.

The capital-expenditure cycle and why results are volatile

AI requires enormous up-front spending on chips, data centres and power before the revenue arrives. That makes one company's spending plan another company's revenue forecast, so changes in spending plans ripple through the whole chain and make results volatile.

The cycle works roughly like this:

  1. Demand signal. Cloud platforms and large technology companies expect strong demand for AI services and raise capital spending.
  2. Orders. They order chips, servers, buildings and power. Suppliers' revenue rises quickly.
  3. Capacity build. Suppliers expand their own capacity, which takes time and money.
  4. Digestion. Buyers pause to absorb what they have built. Orders slow even if long-term demand is intact.
  5. Adjustment. Suppliers that expanded for peak orders face lower revenue against higher fixed costs, and margins fall.

Three features sharpen the cycle:

  • A few large buyers. A change of plan by one very large customer affects many suppliers at once.
  • Depreciation. Capital spending is charged against profit over the asset's estimated life, so heavy spending today lowers reported profits for years.
  • Cash versus profit. A company can report rising profits while free cash flow falls because capex is consuming the cash.

As the bubbles section shows, investment booms can produce more capacity than near-term demand can use, even when the technology is genuinely important.

Key risks of AI investing

The main risks are concentration, valuation, the semiconductor cycle, competition, regulation, export controls, energy and supply constraints, and dependence on a few customers. None of them means AI is a bad investment; each is something the price may or may not already reflect.

Concentration risk

FINRA identifies several ways concentration builds up: deliberate choices, strong performance that makes a winner grow into an outsized share of the portfolio, employer stock, correlated holdings that move together, and illiquid assets. For AI investors the second and fourth are the quiet ones: a winning position grows without you buying more, and several technology shares or funds can depend on the same demand.

Valuation risk and cyclicality in semiconductors

Shares priced for rapid growth are sensitive to any change in expectations: the risk is that a company merely does well instead of extremely well.

Chipmaking combines long lead times, high fixed costs and demand that can change faster than capacity. The industry has been through alternating periods of shortage and oversupply, and strong demand for AI computing does not remove that cycle.

Competition and technological change

Leadership in technology is not permanent. Large customers can design their own chips, rivals improve, open-source models compete with paid ones, and more efficient methods can reduce the hardware needed for a task. Any of these can change who captures the profit, even while total AI use grows.

Regulation: the EU AI Act

The European Union's AI Act is a risk-based law. The European Commission describes four levels: unacceptable-risk practices that are banned, high-risk systems that must meet strict requirements before being placed on the market, systems with transparency obligations such as disclosing that a person is interacting with AI and labelling AI-generated content, and minimal-risk systems with no specific obligations. Providers of general-purpose AI models have their own duties around transparency and copyright, with additional risk assessment for the most capable models.

According to the Commission, the Act entered into force on 1 August 2024 and applies in stages: prohibitions from 2 February 2025, general-purpose AI rules from 2 August 2025 and general application from 2 August 2026, with later dates for high-risk systems. Check the Commission's page for the current schedule. For investors, compliance adds cost and can limit some products, and companies describe their exposure in their risk factors.

Export controls

In the United States the Bureau of Industry and Security (BIS), part of the Department of Commerce, administers the Export Administration Regulations. BIS has imposed licence requirements on advanced computing items and semiconductor manufacturing equipment for certain destinations. One example is its December 2024 action, which added controls on 24 types of chipmaking equipment, three types of software tools and high-bandwidth memory, and added 140 entities to the Entity List.

These rules change over time. A company that sells advanced chips or tools abroad can lose or regain a market through a government decision it does not control, so look at the geographic revenue breakdown and the risk factors.

Energy and supply constraints

AI data centres need a great deal of electricity. A US Department of Energy announcement of a report on the subject states that data centres consumed about 4.4% of total US electricity in 2023 and projects that this could rise to approximately 6.7% to 12% by 2028. Power availability, grid connections and specialised components can all limit how fast capacity is built, which is a risk for companies waiting on capacity and an opportunity for those supplying it.

Customer concentration

Many AI suppliers depend on a small number of very large customers. Companies must disclose significant customer concentration, typically in the notes to the financial statements and in risk factors. A supplier whose top few customers make up most of its revenue is, in effect, a bet on those customers' spending plans.

Bubbles and lessons from history

Nobody can reliably say in advance whether a market is in a bubble. History does show a repeated pattern: a real and important technology attracts more investment than near-term demand can support, prices assume everything goes right, and those who paid peak prices lose money even though the technology succeeds.

The late-1990s technology boom

A Federal Reserve Bank of San Francisco Economic Letter describes the period from early 1996 to mid-2000, when real business fixed investment grew at an average compound rate of about 10% a year, roughly 2.5 times faster than the US economy, with much of it going into computers and information technology. It attributes part of the boom to temporary factors, including telecom deregulation, the spread of the world wide web, dot-com buildouts and replacement of hardware ahead of the year 2000.

The letter says firms vastly overspent on new technology and capacity, and that rising share prices and the investment boom reinforced each other because high share prices gave companies a low-cost source of funds. After the stock market peaked in March 2000, investment fell sharply within about two quarters.

The lesson is not that the internet was a mistake. It is that a technology can transform the economy while shares bought at the peak of enthusiasm still lose money, and that infrastructure spending can run ahead of paying demand.

1929 and borrowed money

The Federal Reserve's history of the 1929 crash records that investors commonly bought shares with borrowed money, putting down only a small fraction of the price. The Dow Jones Industrial Average rose from 63 in August 1921 to 381 in September 1929. By the summer of 1932 it stood 89% below its peak, and it did not return to its 1929 level until November 1954.

Two lessons carry over: leverage turns a fall into a forced sale, and recovery from a peak can take far longer than investors assume. The same essay notes that detecting bubbles is difficult even for central banks.

History does not tell you whether AI prices are too high at any given moment. It does tell you that "this technology is real" and "this price is safe" are separate questions, and it suggests practical defences: diversify, avoid borrowing to invest and size positions so that a large fall is survivable.

AI washing and AI-themed investment scams

"AI washing" means overstating or inventing the use of artificial intelligence to attract customers or investors. Regulators have brought enforcement cases over it and have separately warned the public about scams that use AI as bait, including trading bots that promise guaranteed returns.

SEC enforcement against AI washing

In March 2024 the SEC announced settled charges against two investment advisers, Delphia (USA) Inc. and Global Predictions Inc., for making false and misleading statements about their purported use of AI. The firms agreed to pay 400,000 US dollars in total civil penalties. The SEC chair at the time said that investment advisers should not mislead the public by saying they are using an AI model when they are not.

The theme has also appeared in cases involving companies. In June 2024 the SEC charged the founder of an AI recruitment start-up, Joonko, with defrauding investors of at least 21 million US dollars through false statements about its customers and revenue. In January 2025 it settled charges against Presto Automation, a formerly Nasdaq-listed restaurant-technology company, over misleading statements about its AI voice-ordering product, including the extent of human involvement. The company neither admitted nor denied the findings.

The word "AI" in a pitch, press release or fund name proves nothing. Look for evidence in filings: revenue attributed to AI products, clear product descriptions and consistent statements over time. Be wary of vague claims, sudden rebranding around AI and thinly traded shares promoted on social media.

Regulator warnings about AI investment fraud

In January 2024 the SEC's Office of Investor Education and Advocacy, the North American Securities Administrators Association and FINRA issued a joint investor alert on artificial intelligence and investment fraud. It warns about:

  • Unregistered platforms that promote AI trading systems with promises such as guaranteed winners.
  • Pump-and-dump schemes in which promoters spread false positive information about a small company's AI credentials, push up the price and sell. Microcap and penny stocks are particularly vulnerable.
  • Finfluencers and celebrity endorsements on social media.
  • Deepfakes: AI-generated voices and videos impersonating executives, officials or family members.
  • AI-generated information that may be inaccurate, incomplete or fabricated.

Trading bots and "guaranteed" returns

A CFTC customer advisory from January 2024 is blunt: AI technology cannot predict the future or sudden market changes. It warns about bots and algorithms that claim very high or guaranteed returns, and cites a scheme promising bot-driven trading profits that took more than 1.7 billion US dollars in bitcoin from about 23,000 people.

The CFTC suggests researching the company and the people behind it, checking how long the website has existed, getting a second opinion, understanding all fees and being sceptical of influencer pitches. Many of these schemes ask for payment in crypto-assets; our guide on how to spot a crypto scam covers those red flags in detail, and the same pressure tactics appear in the offers described in how to avoid loan scams and fake offers.

In short: check that anyone offering investments is registered (see how to check a broker or adviser), look up the company's filings on EDGAR, treat guaranteed returns and urgency as red flags, and verify any voice or video message through a contact detail you already have.

Diversification and position sizing

Diversification means spreading money across and within asset classes so that no single company or theme can do serious damage. Position sizing means deciding in advance how much any one holding, or one theme, may represent. Together they matter more to long-term outcomes than picking the "right" AI share.

FINRA says diversification reduces the risk of major losses that can result from over-emphasising a single security or asset class. It does not prevent losses.

Measure your real AI exposure: an illustrative example

The following portfolio is fictional and the percentages are invented for illustration.

Holding Share of portfolio Assumed AI-related share of that holding AI exposure as share of portfolio
Broad stock index fund 60% 25% 15%
Thematic AI ETF 20% 100% 20%
One chip company's shares 10% 100% 10%
Bond fund 10% 0% 0%
Total 100% 45%

This investor might say they are "mostly in an index fund with a bit of AI". In fact 45% of the portfolio depends on one theme. Looking through funds to their holdings is the only way to see this.

Principles for sizing

  • Start from the whole portfolio. Decide your mix of shares, bonds and cash first, then how much may sit in one theme.
  • Set a limit before you buy. Choose a maximum for a single company and for the AI theme that you could see fall heavily without changing your plans. There is no official "right" number.
  • Count overlap. Include AI-related companies inside your index and pension funds, and any employer shares.
  • Rebalance. FINRA suggests considering rebalancing about once a year. If a winner has outgrown your limit, direct new money elsewhere or trim it, remembering that selling can trigger tax.

Dollar-cost averaging vs lump sum

Dollar-cost averaging means investing equal amounts at regular intervals regardless of price; lump-sum investing means putting all the money in at once. Averaging reduces the risk of buying everything just before a fall, while lump-sum investing has often produced higher returns over long periods because the money is in the market sooner. Neither can prevent losses.

FINRA explains that with dollar-cost averaging you buy more shares when prices are low and fewer when they are high, which can lower your average cost, and a fixed schedule removes the pressure of deciding when to buy. But holding cash for longer often produces lower returns than lump-sum investing, especially over longer periods, and more transactions can mean more fees.

Illustrative example (fictional prices)

Suppose you have 1,200 to invest in a fund, either all at once at a price of 50 or as 300 a month for four months.

Scenario Monthly prices Shares bought by averaging Value at final price (averaging) Lump sum: 24 shares at 50, value at final price
Price falls then partly recovers 50, 40, 30, 40 6 + 7.5 + 10 + 7.5 = 31 31 x 40 = 1,240 24 x 40 = 960
Price rises steadily 50, 55, 60, 65 6 + 5.45 + 5 + 4.62 = about 21.07 About 21.07 x 65 = about 1,370 24 x 65 = 1,560

In the first scenario averaging wins: the average cost is about 38.71 per share against 50. In the second, the lump sum wins. You cannot know in advance which scenario you are in.

Time horizon and risk tolerance

Your time horizon is when you will need the money; your risk tolerance is how much loss you can bear, financially and emotionally, without abandoning your plan. Both should be settled before you choose any investment, and both point towards caution with money you need soon.

FINRA notes that the right asset allocation depends on factors including risk tolerance and investment horizon: someone saving for retirement decades away may invest differently from someone who needs a house deposit soon. A concentrated or highly valued sector can fall a long way and take years to recover.

Questions to answer honestly:

The UK's Financial Conduct Authority suggests similar questions: can you afford to lose the money, do you understand the investment, is it regulated, and what protection exists if the provider fails?

Taxes in brief

In most countries you may owe tax when you sell shares for more than you paid (a capital gain) and on dividends, unless the investments are held in a tax-advantaged account. The rules, rates and allowances vary by country and change, so treat this section as orientation and check the official source for your situation.

United States

The IRS explains that a capital gain or loss is short-term if you held the asset for one year or less and long-term if you held it for more than one year. Long-term gains are generally taxed at lower rates than short-term gains, which are taxed at ordinary income rates. Capital losses can offset gains, and the IRS limits how much net loss can be deducted against other income in a year, with the excess carried forward.

Two consequences follow: frequent trading tends to produce short-term gains taxed at higher rates, and selling a large winner to reduce concentration can trigger a tax bill. Tax-advantaged retirement accounts have their own rules and limits, set out by the IRS.

Other countries

The UK and Canada have their own tax-advantaged accounts, described in the notes below. Buying foreign shares can add currency conversion costs; see our explainer on foreign transaction fees and dynamic currency conversion. A qualified tax professional can advise on your position.

How to check a broker or adviser, and what SIPC covers

Before you send money to anyone, confirm they are registered. In the United States, FINRA BrokerCheck covers brokers and brokerage firms and links to the SEC's Investment Adviser Public Disclosure (IAPD) database for investment advisers. Both are free.

BrokerCheck and IAPD

FINRA says BrokerCheck lets you research individuals and firms that sell securities or give financial advice. A summary report shows employment history, qualifications and disclosure events. For investment advisers not registered through FINRA, BrokerCheck directs you to the SEC's IAPD database.

Go to the official sites directly instead of following a link in an email or advert, and check that the contact details you were given match those on the register.

What SIPC protects

The Securities Investor Protection Corporation (SIPC) protects customers if a SIPC-member brokerage firm fails financially and customer assets are missing. According to SIPC, protection is up to 500,000 US dollars per customer, including a 250,000 US dollar limit for cash. It covers securities such as stocks, bonds and mutual funds held at the firm.

What SIPC does not protect

SIPC states that it does not protect against a decline in the value of your securities. It also does not cover losses from bad investment advice, commodity futures contracts (with limited exceptions), foreign exchange trades, or digital assets that are unregistered investment contracts. It protects custody of your assets if the firm fails; it is not insurance against the market.

Notes for UK and Canadian readers

The principles in this guide apply everywhere, but regulators, protection schemes and tax-advantaged accounts differ by country.

United Kingdom

  • Regulator. The Financial Conduct Authority (FCA) authorises investment firms. Its Firm Checker lets consumers confirm that a firm is authorised and has permission for the service it offers; fuller detail is on the Financial Services Register.
  • Stocks and shares ISA. GOV.UK explains that you do not pay tax on income or capital gains from investments held in an ISA. A stocks and shares ISA can hold company shares, unit trusts, investment funds and bonds. For the 2026 to 2027 tax year, GOV.UK gives the maximum you can save across ISAs as £20,000. Allowances can change, so check GOV.UK.
  • FSCS. The Financial Services Compensation Scheme can pay up to £85,000 per person per firm (for firms that failed after 1 April 2019) if an authorised investment provider or adviser has gone out of business. It states that it cannot accept claims for poor investment performance.

Canada

  • Regulators and registration. The Government of Canada's guidance on choosing a financial advisor says that, by law, sellers of mutual funds, stocks and bonds must be registered with a provincial or territorial securities regulator, and points to the National Registration Search to verify this. It refers to the Canadian Investment Regulatory Organization (CIRO) and the Canadian Securities Administrators (CSA) for disciplinary information.
  • TFSA. The Canada Revenue Agency describes the Tax-Free Savings Account as a registered account for Canadian residents aged 18 or over. Investment income, capital gains and withdrawals are generally tax-free, and a TFSA can hold cash, GICs, bonds, mutual funds and securities listed on a designated stock exchange. Contribution room is limited; check yours with the CRA.

Common mistakes

Most losses in a hot sector come from behaviour, not from the technology. These are the errors that recur.

  • Chasing headlines. By the time a result or product launch is in the news, the price has usually moved.
  • Using leverage. The SEC warns that buying on margin can produce losses greater than the cash you put in, that losses can mount quickly, and that a broker may be able to sell your securities without consulting you if your account falls below the firm's requirements.
  • Trading options without understanding them. Options can expire worthless, and they put a time limit on a bet that might otherwise have had years to work.
  • Holding leveraged or inverse ETFs for long periods. FINRA notes they are generally designed for short holding periods.
  • Single-stock concentration. Letting one winner grow until it dominates the portfolio, or owning several AI funds that hold the same companies.
  • Ignoring fees. Paying a higher expense ratio for a thematic fund without checking what you get for it.
  • Confusing a great product with a great investment. Liking an AI tool says nothing about its maker's valuation.
  • Believing predictions. Price targets and forecasts are opinions.
  • Trusting unregistered sellers. Acting on tips, bots or "signals" without checking registration.

How to evaluate any AI company, step by step

You can assess any AI-related company with the same nine steps. The aim is not to predict the share price but to understand what you would own and what has to go right.

  1. Place it in the value chain. Chips, equipment, cloud, infrastructure, models, software or adopter? Who pays it?
  2. Read the Business section of the latest annual report. Write down in two sentences what it sells and to whom.
  3. Find how much revenue is really AI. Use segment reporting.
  4. Check growth. Compare revenue growth over several years and quarters. Is it accelerating or slowing?
  5. Check profitability and cash. Look at margins and free cash flow, in both GAAP and adjusted figures.
  6. Look at capital spending and the balance sheet. How much capex, how is it funded, and how much debt and cash?
  7. Read the risk factors. Note customer concentration, export restrictions, regulation, energy and competition, and what changed since last year.
  8. Consider valuation. Work out the P/E, forward P/E, price-to-sales and price-to-free-cash-flow, and ask what growth they assume.
  9. Decide the size. Check your total AI exposure including funds, set a limit, and decide in advance what you will do if the price halves or doubles.

If at any step you cannot find the answer, that is a finding in itself. A diversified fund is a legitimate alternative to doing this work company by company.

Investor's checklist

  • I have an emergency fund and no high-interest debt to pay off first.
  • I know when I will need this money and could leave it invested through a large fall.
  • I know how much AI exposure my existing funds already give me, and I have set a maximum for any single company and for the theme.
  • For each company, I know its layer of the value chain and its main customers.
  • I have read the Business, Risk Factors and MD&A sections of the latest annual report and checked segment revenue.
  • I have looked at capex, free cash flow and what growth the valuation assumes.
  • For each fund, I have checked the top holdings, the expense ratio and overlap with what I own.
  • I am not using margin, options or leveraged ETFs that I do not fully understand.
  • I have verified my broker or adviser on an official register and know what SIPC, FSCS or the local scheme covers.
  • I have ignored any offer of guaranteed returns, AI trading bots or "pre-IPO" shares from unknown sellers.
  • I understand the tax treatment of my account and have a date set to review and rebalance.

FAQ

What are AI stocks?

They are shares of companies whose business depends significantly on artificial intelligence, such as chipmakers, cloud platforms, data-centre suppliers, model developers and software firms. There is no official definition, so always check what a company actually sells.

How do I start investing in AI stocks as a beginner?

Check how much AI exposure you already have through index or pension funds, decide how much of your portfolio may go into one theme, then choose a broad fund, a thematic fund or individual shares. Use a registered broker and read the prospectus or filings first.

Is it too late to invest in AI?

Nobody can answer that reliably. The outcome depends on the price paid relative to future results, which are unknown. You can control diversification, position size, costs and time horizon.

Are AI stocks in a bubble?

It is not possible to know in advance, and even central banks find bubbles hard to identify. A real technology can coincide with over-investment and over-valuation, so diversify and avoid leverage instead of trying to time it.

Should I buy individual AI stocks or an AI ETF?

Individual shares give control and no fund fee but concentrate risk and require research. A thematic ETF spreads risk across more companies but charges an expense ratio and may duplicate funds you already own.

How do I know if a company's AI claims are real?

Compare the marketing with the filings. Look for AI revenue in segment reporting and specific product descriptions in the Business section. The SEC has brought cases against firms that overstated their use of AI.

Why do AI stocks fall after good earnings?

Because prices reflect expectations, not just results. If guidance for future growth, margins or spending is below what investors expected, the share can fall even after a strong quarter. Both the earnings estimate and the multiple investors pay can shrink at the same time.

Can I invest in private AI companies before they go public?

Usually not directly. In the US, most private offerings are limited to accredited investors who meet income, net worth or professional criteria. Private shares are illiquid and come with limited disclosure. Be very wary of unsolicited "pre-IPO" offers.

Are AI trading bots legitimate?

Automated trading tools exist, but the CFTC warns that AI cannot predict the future or sudden market changes, and that promises of high or guaranteed returns from bots are a hallmark of fraud. Check registration and never rely on testimonials or screenshots.

How much of my portfolio should be in AI stocks?

There is no official figure and this guide cannot give personal advice. Set a limit you could tolerate falling substantially, count the AI exposure inside your existing funds, and review it at least once a year.

Does SIPC protect me if my AI stocks lose value?

No. SIPC protects customers when a member brokerage firm fails and assets are missing, up to its limits. It does not cover falls in the value of investments. The UK's FSCS likewise does not pay out for poor investment performance.

How are gains on AI stocks taxed?

In the US, gains on shares held for more than a year are long-term and generally taxed at lower rates than short-term gains, according to the IRS. Accounts such as ISAs in the UK and TFSAs in Canada can shelter gains, subject to their rules.

Bottom line

"AI stocks" is a label covering very different businesses, from cyclical chipmakers to diversified cloud giants to companies that merely mention AI in their marketing. The label tells you little; the value chain, the filings and the valuation tell you a lot.

Know how much AI exposure you already hold. Decide your limits before you buy. Read the annual report instead of the headline. Remember that a price built on high growth can fall hard on a small disappointment. Keep fees low, avoid borrowed money, verify anyone who asks for your money and ignore every promise of a guaranteed return. None of this tells you which company wins, but it means being wrong about one of them need not derail your finances.

This article is general information, not investment advice.

Official sources referenced in this guide

Labels
Previous Post Next Post