Wells Fargo is no longer telling investors to stay away from the AI infrastructure trade. In a new note, analysts led by Ohsung Kwon raised the firm’s view on US AI infrastructure to neutral from negative.
The change follows a deep selloff. Semiconductor shares trade about 38% below their June highs, and the multiples investors pay for them have compressed. At current prices, the analysts said, valuations imply earnings growth of only about 8% a year.
A compressed multiple means investors now pay less for each dollar of expected profit. That can happen even while earnings keep rising, which is why a stock can fall while its business improves.
Politics Adds a New Worry to the AI Spending Debate
Democratic wins in the midterm elections, including in states that oppose new data centers, fueled concerns about regulation. A stricter permitting climate could slow construction and raise its cost, which would cut into the spending that has fed AI company revenue.
The selloff has erased about $3 trillion from the S&P 500. Readers following the pullback can find the names analysts like most in our list of oversold AI stocks with 50%-plus upside. Kwon and his team said the risk-reward is now reasonable after the drop.
Why 2028 Worries Analysts More Than 2027
AI spending added about 8 percentage points to S&P 500 earnings growth over the past year. The Wells Fargo analysts wrote that “The tailwind is expected to continue in 2027,” and that it largely disappears in 2028 as spending growth slows and depreciation costs rise.
Depreciation explains much of the concern. Companies spread the cost of data centers and chips over several years, so those charges keep landing on income statements even after the buying slows. Those charges are real expenses even though the cash left the company earlier. Profit growth then shrinks, just as the spending that lifted it fades. Analysts build their forecasts from spending levels, so a plateau in budgets changes the math quickly.
The bigger risk, the analysts said, is a downturn in earnings per share once capital spending peaks. They also pointed to extra supply arriving in 2028 that prices do not yet reflect. For investors, the question is how much of that is already in share prices after the drop.
Companies still growing quickly through the selloff are the subject of our roundup of the fastest-growing AI stocks. The Wells Fargo shift is a move from caution to a neutral stance, not a call that the selloff is over.
Photo by Arturo Añez on Unsplash
Our Methodology
For this article, we used Finviz to screen for AI stocks with an average upside potential of at least 30%. We have ranked these stocks in ascending order of their upside potential, as of November 12, 2026. We also list each stock’s number of hedge fund holders as of Q2 2026, from our database of 1,006 hedge funds.
At Insider Monkey, we are obsessed with what hedge funds do. The reason is simple: our research has shown that we can outperform the market by imitating the top stock picks of the best hedge funds. Insider Monkey’s quarterly newsletter strategy selects 14 small-cap and large-cap stocks every quarter and has returned 612% since May 2014, beating its benchmark by 375 percentage points (see more details here).
10 Best AI Stocks to Buy After Wells Fargo’s Outlook Upgrade
10. Lam Research (NASDAQ:LRCX)
Average Upside Potential: 55%
Number of Hedge Fund Holders: 139
Bernstein analyst Stacy Rasgon kept a Buy rating and a $350 target on Lam Research on November 2, after the stock lost about 34% in a month.
Lam makes the machines that chipmakers use to etch and deposit the thin layers inside a chip, and memory makers are among its largest customers. As chips stack more layers, each factory needs more of those steps. The selloff came after a record September quarter. Guidance for the December quarter only matched expectations, and worries about AI spending pulled down the whole chip equipment group. China made up 35% of September-quarter revenue and is expected to fall sharply in December, which is the main near-term drag.
Other firms trimmed their targets after the drop, and the Street’s average target slipped from its October peak, though it still sits above Bernstein’s. Part of Lam’s cushion comes from service work on machines already installed, which we examined in our look at its fast-growing support revenue. Memory makers’ spending plans decide how quickly orders come back. The next test comes when the company reports December-quarter results in late January.
9. ServiceNow (NYSE:NOW)
Average Upside Potential: 55%
Number of Hedge Fund Holders: 115
Argus reiterated a Buy rating and a $170 target on ServiceNow on November 3. The stock trades 42% below its high from a year ago.
The drop traces to weak guidance for fourth-quarter contracted backlog. ServiceNow sells subscription software that routes work requests and automates tasks for IT, HR and customer service teams, so backlog is the best early signal of future sales. Management blamed the weak figure on the timing of deals, and Argus’s call rests on that explanation holding up. The target works out to 37 times expected earnings, below the stock’s five-year average of 53 times. The company is also absorbing its purchase of Armis, a security company, which adds costs. Bears point to analysts trimming their 2026 and 2027 earnings estimates over the past month.
Bulls lean on the AI product line, where ServiceNow raised its 2026 annual contract value goal for AI products to $1.7 billion. Investors are watching how fast those new AI sales convert. Whether cash flow can keep up with that spending is the question we raised in our piece on ServiceNow’s AI expansion.
8. Teradyne (NASDAQ:TER)
Average Upside Potential: 58%
Number of Hedge Fund Holders: 82
Jefferies analyst Dushyant Ailani started coverage of Teradyne on November 11 with a Buy rating and a $284 target, writing that “AI has made TER a compute test company.” The stock jumped that day.
Teradyne makes the automated equipment that checks chips for defects before they ship. Jefferies expects the market for that equipment to grow from about $9 billion in 2025 to about $25 billion by 2030. It also sees Teradyne’s share of system-on-chip testing rising from 27% to 34%, as Nvidia and large cloud companies buy test systems from more than one supplier instead of one. That opens room for a challenger like Teradyne, and dual sourcing at the largest chip buyers is the central pillar of the Jefferies thesis.
Teradyne also reported better-than-expected results in its latest quarter. Jefferies’ revenue forecasts for the next two years run slightly above consensus. The target itself sits well below the average of other analysts who cover the stock, and shares still trade far under their summer high. Readers weighing the valuation can find a fuller discussion in our piece on whether Teradyne’s AI testing wins are already priced in.
7. SpaceX (NASDAQ:SPCX)
Average Upside Potential: 82%
Number of Hedge Fund Holders: 119
Stifel analyst Jonathan Siegmann reiterated a Buy rating and a $190 target on SpaceX on November 9, days after Democratic midterm wins in states that oppose new data centers raised fears of tougher rules. He argues that SpaceX could step in as a buyer for stalled U.S. projects that already have power secured, which links its launch and power know-how to a real bottleneck for AI builders.
SpaceX launches rockets and runs the Starlink satellite internet network, and its AI computing business now brings in more than half of its sales. Third-quarter revenue rose 161% to $14.6 billion, and its first revenue-generating Starship mission delivered a batch of new Starlink satellites. Chief Financial Officer Bret Johnsen said new computing capacity is paying for itself in less than a year at current prices. The company guided first-quarter 2027 revenue to $33 billion to $37 billion.
The spending behind that growth is heavy, and most of it goes to AI computing. How Starlink fits into funding it is the subject of our look at whether the satellite business can carry the AI bill.
6. Nvidia (NASDAQ:NVDA)
Average Upside Potential: 83%
Number of Hedge Fund Holders: 285
Baird downgraded Nvidia to Neutral from Outperform on November 11 and cut its target to $200 from $500, pointing to the outlook for AI spending and risks tied to China. Evercore ISI followed on November 12, moving the stock to In Line from Outperform with a $186 target. Oppenheimer also lowered its rating that day, to Perform from Outperform.
Nvidia designs the processors that train and run most AI models, so any doubt about customer spending lands on it first. Its chips sit in the data centers of the largest cloud companies, the same buyers whose budgets Baird questions. Investors get a fresh read next Tuesday after the close, when the company reports quarterly results. Wall Street expects earnings of $2.47 per share for the quarter that ended in October.
Bulls argue that the stock’s slide has left it cheap against its own growth record, while bears, Baird among them, want proof that spending stays strong into next year. Our piece on whether Nvidia’s valuation already prices in a growth stall covers the other side of that debate. The report will show whether customers are still placing orders at the same pace.
5. Alibaba (NYSE:BABA)
Average Upside Potential: 84%
Number of Hedge Fund Holders: 97
Mizuho analyst Wei Fang raised his Alibaba target to $200 from $195 on November 3 and kept an Outperform rating, ahead of the company’s earnings report next Thursday.
Alibaba runs the Taobao and Tmall shopping sites, the AliExpress and Lazada marketplaces, and Alibaba Cloud, and the cloud arm is where analysts see the AI payoff. Its Qwen family of AI models feeds that business, giving Alibaba a product to sell alongside computing power. In the quarter ended in June, external cloud revenue grew 45%, and AI-related products made up 35% of it. Analysts also see the QwenWork enterprise agent adding recurring demand.
The cost of that growth worries bears. Adjusted EBITA fell 30% in the same quarter as heavier AI infrastructure spending and higher costs for the Qwen app weighed on earnings, and free cash flow was a larger outflow than a year earlier. Alibaba reports September-quarter results on November 18, and analysts expect earnings of $1.65 per share. The spending side of the story is the focus of our piece on what Alibaba’s AI bill means for its cash.
4. Marvell Technology (NASDAQ:MRVL)
Average Upside Potential: 87%
Number of Hedge Fund Holders: 96
BofA analyst Vivek Arya downgraded Marvell to Neutral from Buy and cut his target to $225 from $400 after the company agreed to buy Lumentum in an all-stock deal. TD Cowen also downgraded the stock to Hold from Buy and cut its target to $175 from $350.
Marvell designs the chips and networking gear that move data inside AI data centers, and Lumentum makes optical components that carry that data as light. The deal, valued at about $115 billion including debt, sent Marvell shares down about 13%. Because Marvell is paying in its own stock, current shareholders will own a smaller slice of a larger company.
Not everyone backed away. UBS reiterated its Buy rating and $350 target on November 3, and Cantor Fitzgerald stayed at Neutral with a lower $225 target. The split shows investors weighing whether the optical business gives Marvell a bigger share of AI spending or adds integration risk. Investors will also watch how the two companies combine their product lines. The skeptics’ case predates the deal, and we laid out what the few analysts without a Buy rating saw weeks earlier.
3. Oracle (NYSE:ORCL)
Average Upside Potential: 90%
Number of Hedge Fund Holders: 119
Stephens upgraded Oracle to Overweight from Equal Weight on November 2 and raised its target to $213 from $185, after the company’s investor day. Management said free cash flow should turn positive in the second half of fiscal 2028 and capped new financing at $20 billion through fiscal 2030, split evenly between debt and equity.
Oracle sells databases and business software, and it now rents AI computing capacity to customers building large models. Stephens argues that vendor financing, customers’ own hardware and prepayments now cover much of the buildout, which leaves Oracle needing about half the new money analysts once expected. The stock has fallen about half from its high a year ago, so the financing plan sits at the center of the debate. The $213 target sits below the average of analysts covering the stock.
The open question is concentration. S&P has said OpenAI makes up about half of Oracle’s $664 billion backlog, so much of the expected revenue depends on OpenAI paying its bills. Oracle reports quarterly results in December. A slower-moving question, how durable Oracle’s competitive edge is, gets attention in our look at its economic moat.
2. Sandisk (NASDAQ:SNDK)
Average Upside Potential: 121%
Number of Hedge Fund Holders: 128
Wells Fargo analyst Aaron Rakers cut his Sandisk target to $1,300 from $1,550 on October 30 and kept an Equal Weight rating, a day after the company reported fiscal first-quarter results. He expects the pace of earnings upside to slow. Sandisk also announced an eight-for-one stock split. The new shares arrive next Thursday and split-adjusted trading starts the next day, so the $1,300 target is a pre-split figure.
Sandisk, spun out of Western Digital last year, makes the NAND flash chips used in solid-state drives, and AI data centers have pushed demand and prices sharply higher. In the fiscal first quarter, revenue reached $11.0 billion with an 85% gross margin, and the company guided the next quarter to $12.2 billion to $12.6 billion. AI servers need large amounts of fast storage, which explains the demand.
The worry is how long prices keep rising. Kioxia, a rival flash maker, warned that price gains are fading, and Sandisk shares fell 14% in a week. Fox Advisors downgraded the stock to Equal-Weight on November 2, while Wedbush kept an Outperform rating and a $2,000 target. Whether the memory boom has already peaked is the question we put to Sandisk’s forward earnings multiple.
1. SoundHound AI (NASDAQ:SOUN)
Average Upside Potential: 186%
Number of Hedge Fund Holders: 20
Cantor Fitzgerald analyst Thomas Blakey trimmed his SoundHound target to $10 from $15 on November 6 and kept an Overweight rating, a day after the company reported third-quarter results.
SoundHound sells voice AI software that lets cars, restaurants and customer service lines understand spoken requests. Revenue hit a record, up 48% from a year earlier. The quality of that revenue slipped, though: non-GAAP gross margin fell to 50.3% from 59.3% a year earlier. Management said revenue recognized at a single point in time shrank sharply as a share of sales. Investors are watching whether margins stabilize as the revenue mix shifts.
The company is cutting costs by $120 million on an annualized basis and says it expects an adjusted EBITDA profit in the second half of 2027. The next report will show whether margins recover. The company also closed its LivePerson acquisition, which adds customer-service software and more customers to sell to. Bulls see sales growth that is still strong, while bears see a company spending heavily to chase scale. We examined whether the LivePerson acquisition fixes SoundHound’s scale problem after the deal closed.
READ NEXT: SoundHound vs. Rezolve: Which AI Growth Story Is Worth the Financing Risk? and Salesforce (CRM) vs Oracle (ORCL): Which Is a Better Stock to Buy?
