Does Walnut recommend which AI stocks to buy?

No. Walnut is not a registered investment adviser and does not tell you what to buy. It lets you build a thematic basket from AI stocks you choose, set target weights, see how the basket would track against the S&P 500, and place trades you approve yourself at your own broker. Every page here is descriptive and informational, not a recommendation.

Best AI Stocks

Last updated July 2026

Short answer

There is no single list of best AI stocks, because the right holdings depend on your goals and no one can predict prices. What dominates AI portfolios is a spread across three layers. The chipmakers build the silicon the models run on: NVDA, AMD, AVGO, TSM, MRVL, and MU. The hyperscalers and software platforms monetize AI on top of huge existing businesses: MSFT, GOOGL, AMZN, ORCL, and PLTR. Around the buildout sit the hardware and networking names like ANET and DELL. A record wave of data-center spending is the tailwind people cite, but a lot of growth is already priced in and the chip layer is cyclical. The useful move is to treat a list like this as research and build a diversified portfolio from it, not to buy one name. Walnut, an AI investing app, can compare these names against your existing holdings. This page is descriptive and informational, not investment advice.

AI has been the single biggest driver of the market as hyperscalers poured capital into data centers and the chips that fill them. That backdrop produces endless headlines about the top AI stocks to buy, which read like predictions, and predictions about individual stock prices are the one thing no one does reliably. So this guide does something more honest. It groups the AI stocks people most widely hold and discuss in 2026 by their layer in the stack, explains what each one actually does and the risks it carries, links each to a fuller page, and then shows how to turn a list like this into a portfolio instead of a single bet. Nothing here is a recommendation to buy or sell, and Walnut is not an investment adviser.

What is the AI buildout, honestly?

The reason AI stocks get so much attention is a real surge in spending. The largest cloud companies have committed hundreds of billions of dollars to build AI data centers, and most of that money flows down the same supply chain: accelerators, memory, networking, servers, and power. That is the mechanism behind the theme, and it is genuine.

But honesty cuts both ways, and a spending wave is not a guarantee.

Valuations price in the future. Many AI names already trade on years of expected growth, so a disappointing quarter or any sign the spending is slowing can trigger sharp drops.
The chip layer is cyclical. Semiconductors and especially memory have a long history of boom-and-bust. Demand that looks endless can turn quickly.
Concentration is real. A handful of hyperscaler customers drive much of the demand, and the AI names tend to move together, which reduces the diversification of owning several of them.

None of this is a recommendation. It is the context you need to read the list below as research rather than as a set of hot tips riding a spending headline.

What AI stocks are most widely held in 2026?

Below are the AI names most widely held and discussed in 2026, grouped by the layer each one occupies in the AI stack. For each, the note explains what the business does and why it is commonly held, not whether you should own it. Every name links to its own page with the deeper detail.

The chipmakers (AI infrastructure)

The most direct exposure to AI is the silicon that trains and runs the models. These are the companies that design the accelerators, memory, and networking chips the buildout depends on, plus the foundry that fabricates most of them. They anchor most AI portfolios because demand has been visible in their order books, with the standing caveat that chip demand is cyclical and expectations here are high.

Nvidia (NVDA). Nvidia designs the GPUs that train and run most large AI models, and its data-center segment became the single biggest driver of the AI trade. It is the most widely held AI name and the one the whole theme is benchmarked against, though its valuation prices in years of continued growth.
AMD (AMD). AMD is the main challenger in AI accelerators with its MI-series data-center GPUs, alongside a strong CPU business. It is commonly held as the number-two AI-chip story and a way to express the view that the accelerator market has room for more than one winner.
Broadcom (AVGO). Broadcom builds custom AI chips (ASICs) for hyperscalers and dominates the networking silicon that ties data centers together, plus enterprise software. It is widely held as the pick-and-shovel AI name that wins regardless of which model provider leads.
Taiwan Semiconductor (TSM). TSMC is the foundry that actually manufactures the leading-edge chips Nvidia, AMD, and Apple design. It is commonly held as the indispensable node in the whole supply chain, with the well-known overhang of Taiwan-related geopolitical risk.
Marvell (MRVL). Marvell supplies custom silicon and optical and networking components for AI data centers. It is held as a smaller, more leveraged play on the interconnect layer of the buildout, which also makes it more volatile.
Micron (MU). Micron makes the high-bandwidth memory (HBM) that AI accelerators need to feed data to the compute. It is commonly held as the memory-cycle expression of AI demand, with the caveat that memory is historically the most boom-and-bust part of semis.

The hyperscalers and AI-software platforms

The other main way to own AI is through the companies building the models, the cloud that serves them, and the software layer on top. These are larger, more diversified businesses where AI is a growth driver rather than the whole company, which is why they are often held as the lower-volatility way into the theme, though the amount they are spending on AI is itself a debated risk.

Microsoft (MSFT). Microsoft is the enterprise AI leader through its OpenAI partnership, Azure cloud, and Copilot products embedded across Office and Windows. It is one of the most widely held AI names precisely because AI rides on top of a huge, profitable existing business.
Alphabet (GOOGL). Alphabet builds its own frontier models (Gemini), designs its own AI chips (TPUs), and runs Google Cloud, all funded by the search-and-ads engine. It is commonly held as a full-stack AI owner that is less dependent on any one supplier.
Amazon (AMZN). Amazon runs AWS, the largest cloud provider, builds its own AI chips (Trainium, Inferentia), and has invested in Anthropic. It is widely held as the cloud-infrastructure way to own AI demand at scale.
Oracle (ORCL). Oracle re-rated on huge AI-cloud capacity deals as it became a place hyperscalers and model labs rent compute. It is commonly held as the enterprise-database incumbent turned AI-infrastructure story, though its capital-spending ramp is a real risk.
Palantir (PLTR). Palantir sells the AI software platform (AIP) that companies and governments use to put models to work on their own data. It is widely held as the marquee applied-AI-software name, though it trades at one of the richest valuations in the market.

The hardware and networking around the buildout

Beyond chips and clouds, the data-center buildout needs servers, switches, and the plumbing that connects it all. These names are held as second-order exposure to the same capital spending, with the caveat that they depend heavily on a handful of hyperscaler customers.

Arista Networks (ANET). Arista makes the high-speed switches that connect the GPUs inside AI data centers. It is commonly held as a direct networking beneficiary of the buildout, concentrated in a few large cloud customers.
Dell Technologies (DELL). Dell builds the AI-optimized servers that ship the accelerators to enterprise and cloud buyers. It is held as the hardware-integration way to own AI demand, at thinner margins than the chip designers.

At a glance

The same names, grouped by layer, so you can scan the breadth across the list rather than read it as a ranking.

Ticker	Company	What it does
NVDA	Nvidia	Designs the GPUs that train and run most large AI models.
AMD	AMD	Number-two AI accelerator maker plus data-center CPUs.
AVGO	Broadcom	Custom AI chips for hyperscalers and data-center networking.
TSM	Taiwan Semiconductor	The foundry that fabricates most leading-edge AI chips.
MRVL	Marvell	Custom silicon and networking for AI data centers.
MU	Micron	High-bandwidth memory that AI accelerators depend on.
MSFT	Microsoft	Azure cloud, OpenAI partnership, and Copilot across its software.
GOOGL	Alphabet	Gemini models, TPUs, Google Cloud, and search.
AMZN	Amazon	AWS cloud, custom AI chips, and an Anthropic stake.
ORCL	Oracle	Enterprise database and fast-growing AI-cloud capacity.
PLTR	Palantir	Applied-AI software platform for enterprises and government.
ANET	Arista Networks	High-speed networking switches for AI data centers.
DELL	Dell Technologies	AI-optimized servers that ship accelerators to buyers.

How do you build a portfolio from these instead of buying one?

A list of stocks is an input, not a portfolio. The difference between the two is structure: which layers you want exposure to, how much weight each name gets, and the discipline to keep no single position from dominating. The repeatable way to do it looks like this.

Pick a thesis. Decide what view you are expressing. Owning the chipmakers for direct demand is a very different portfolio from leaning on the hyperscalers for lower volatility.
Spread across layers, not just names. Holding Nvidia, AMD, and Broadcom is still one bet on AI silicon. Mixing in the software and hyperscaler layer, or pairing AI with unrelated themes, spreads risk so a single chip-cycle shock does not sink everything.
Set target weights. Assign each name a percentage that sums to 100, so concentration is a choice you made rather than an accident of which stock ran up.
Compare against the S&P 500. Check how the mix would have tracked the benchmark, because a sector tilt should earn its keep versus just holding a broad index (the mega-cap AI names are already a large part of that index).
Place the trades and review. Buy to your targets, then revisit periodically as weights drift or as the spending story shifts.

This is exactly what Walnut is built for. You create a thematic basket from the stocks you choose, set a target weight for each, see how the basket would track against the S&P 500, and place trades you approve yourself at your own broker. Walnut frames each holding against the S&P 500 and shows how the mix is concentrated, so the portfolio is a deliberate structure rather than a pile of separate bets. Walnut does not tell you which stocks to buy.

If you would rather own the theme in one holding instead of picking names, see our guide to the best AI ETFs, or browse the AI infrastructure theme and the AI agents theme for ready-made baskets.

How we chose what to feature

To be clear about method, since framing matters on a page like this: this is not a prediction and not a ranking. We did not forecast which AI stocks will rise, score them, or order them by expected return, because no one can do that reliably. We featured names on three descriptive criteria instead.

Widely held. Each is a large, broadly owned company central to the AI trade, appearing across the major AI funds and mainstream portfolios, so the page reflects what people actually hold rather than obscure tips.
Liquid and established. We featured large, liquid, well-covered companies rather than speculative microcaps, so the descriptions lean on durable business facts rather than hype.
Layer-representative. Each name illustrates a layer of the AI stack (chips, cloud and software, or supporting hardware) so the list teaches how an AI portfolio is built, not which single stock to chase.

The result is a map of what tends to anchor AI portfolios in 2026 and how to think about it, not a buy list. Treat every name as a starting point for your own research. Company facts, spending plans, and valuations change; verify current details before you act.

The bottom line on the best AI stocks

The honest answer to “what are the best AI stocks” is that there is no single list, because the right holdings depend on your goals and no one can predict prices. What tends to anchor AI portfolios is a spread across three layers: the chipmakers like Nvidia, AMD, Broadcom, TSMC, Marvell, and Micron; the hyperscalers and software platforms like Microsoft, Alphabet, Amazon, Oracle, and Palantir; and the hardware and networking names like Arista and Dell. A record wave of data-center spending is the tailwind people cite, but a lot of growth is already priced in, the chip layer is cyclical, and the names move together. The useful move is to treat a list like this as research and build a diversified, weighted portfolio from it rather than buying a single name. Walnut helps you turn that into a thematic basket you control. It is not an investment adviser, and nothing here is a recommendation.

Try Walnut on top of your broker

Walnut connects any major US broker so you can see how AI names fit your portfolio by chatting through Claude, ChatGPT, or built-in AI. Read-only by default until you choose to trade; Walnut is not an investment adviser and does not tell you what to buy.

Create a free Walnut account

FAQ

What are the best AI stocks to buy in 2026?

There is no single list of best AI stocks, because the right holdings depend on your goals, time horizon, and risk tolerance, and no one can predict prices. What this page shows instead is the AI names most widely held and discussed in 2026, grouped by layer: the chipmakers (NVDA, AMD, AVGO, TSM, MRVL, MU), the hyperscalers and AI-software platforms (MSFT, GOOGL, AMZN, ORCL, PLTR), and the hardware around the buildout (ANET, DELL). Treat them as a research starting point, not recommendations. Walnut is not an investment adviser.

What is driving AI stocks higher?

The main driver is a wave of capital spending on AI data centers. Hyperscalers like Microsoft, Alphabet, Amazon, and Oracle have committed hundreds of billions of dollars to build compute capacity, and most of that flows to chipmakers, memory suppliers, networking, and server makers. The debate, and the risk, is whether that spending produces enough return to justify the current valuations, because a lot of future growth is already priced in.

What is the difference between AI chip stocks and AI software stocks?

Chip stocks like Nvidia, AMD, and Broadcom sell the physical accelerators, memory, and networking the buildout runs on, so they move with hardware demand and are more cyclical. Software and platform names like Microsoft, Alphabet, and Palantir monetize AI through cloud services and applications on top of large existing businesses, so AI is a growth driver rather than the whole story and they tend to be less volatile. Many portfolios hold some of each.

Is Nvidia the best AI stock?

Nvidia is the most widely held and most talked-about AI stock because its GPUs train and run most large models, but most widely held is not the same as best for you. Its valuation prices in years of continued growth, so it carries real expectations risk, and concentrating in one name raises the stakes on that single company. It is a starting point for research, not a recommendation. Walnut is not an investment adviser.

Should I buy individual AI stocks or an AI ETF?

Both are common, and the choice is yours. An AI ETF spreads a single investment across the chipmakers, hyperscalers, and software names in one holding, so any one company stumbling matters less. Individual stocks let you tilt toward a specific layer or name you have a view on, at the cost of more concentration and more work. Many investors use an ETF as a base and add a few individual names. See our guide to the best AI ETFs for the fund route.

What are the risks of AI stocks?

The biggest risk is valuation: a lot of expected growth is already in the prices, so any slowdown in AI spending or a disappointing quarter can trigger sharp drops. The chip layer is cyclical and can swing hard. There is concentration risk, since a few hyperscalers drive much of the demand. And the whole theme can move together on sentiment, which reduces the diversification you might expect from owning several AI names. Spreading across layers helps but does not remove these risks.

From here you can dig into any individual stock, browse the best AI ETFs for instant diversification, or explore the AI infrastructure theme you want exposure to.

Walnut is informational and is not a registered investment adviser. This page describes AI stocks that are widely held and commonly discussed, grouped by layer; it is not a prediction, a ranking, or a recommendation to buy, sell, or hold any security. Investing involves risk, including the possible loss of principal, and past performance does not indicate future results. Company facts, spending plans, and valuations change; verify current details before making any decision. Do your own research or consult a licensed financial professional.