The artificial intelligence landscape saw some controversy in early June because of a research paper from Apple Inc. (ticker: AAPL) titled “The Illusion of Thinking: Understanding the Strengths and Limitations of Reasoning Models via the Lens of Problem Complexity.”
In it, the authors challenged the growing popularity of large reasoning models (LRMs), a type of AI system that “thinks out loud” before producing an answer.
The paper found that while LRMs perform well on medium-complexity tasks, their reasoning falls apart at higher levels of difficulty, even when they have enough computing power. Apple’s research suggested that these models are inconsistent and often fail to apply logical algorithms effectively.
[Sign up for stock news with our Invested newsletter.]
While experts are still debating both the accuracy of Apple’s findings and the validity of the data, what’s gone underappreciated is what the report may really reveal: Apple is falling behind in the AI arms race.
Microsoft Corp. (MSFT) has heavily backed OpenAI, Amazon.com Inc. (AMZN) is a major investor in Anthropic, and Meta Platforms Inc. (META) has poured resources into Meta AI. In contrast, Apple’s AI contribution so far has been significantly less high-profile.
To some commentators on social media forums like Reddit, the paper read more like a warning shot from a company trying to slow down a race it’s not leading, rather than a neutral scientific critique.
All of this underscores how hard it is to invest in AI through stock picking alone. To do it well, you’d need a firm grasp of advanced technical concepts, deep insight into competitive dynamics and a close eye on evolving global regulations. For the average retail investor, that’s a tall order.
“We’re in the early stages of the AI cycle, and proper diversification is extremely important — be it across company stages or geographies — because it’s difficult to pick a winner or two this early,” says Tejas Dessai, director of thematic research at Global X ETFs. “With a thematic exchange-traded fund (ETF), you’re following an idea as opposed to a complex strategy.”
Here are six of the best AI ETFs to buy now:
| ETF | Expense ratio |
| Xtrackers Artificial Intelligence and Big Data ETF (XAIX) | 0.35% |
| Invesco AI and Next Gen Software ETF (IGPT) | 0.58% |
| Global X Robotics & Artificial Intelligence ETF (BOTZ) | 0.68% |
| Roundhill Generative AI & Technology ETF (CHAT) | 0.75% |
| KraneShares Artificial Intelligence & Technology ETF (AGIX) | 1.00% |
| VistaShares Artificial Intelligence Supercycle ETF (AIS) | 0.75% |
Xtrackers Artificial Intelligence and Big Data ETF (XAIX)
“Broadly speaking, capital expenditures for AI remain high. However, investors are increasingly scrutinizing the return on investment for these expenditures,” says Arne Noack, regional investment head of Xtrackers, Americas, at DWS Group. “In other words, it will be important to see how adoption of AI will accelerate growth, increase productivity or reduce costs across the economic value chain.”
For AI exposure, Xtrackers offers XAIX at a 0.35% expense ratio. This ETF uses a patent-focused approach to stock selection. “Many of the existing funds in the market utilize backward-looking mechanisms to determine if a company should be classified as an AI company,” Noack explains. “On the other hand, XAIX’s underlying index screens for approved patents in fields related to AI.”
Invesco AI and Next Gen Software ETF (IGPT)
“We believe AI remains a compelling long-term growth theme in the U.S., driven by innovation and rising enterprise adoption,” says Rene Reyna, head of thematic and specialty product ETF strategy at Invesco. “However, near-term sentiment is being weighed down by renewed tariff pressures, creating uncertainty for investors in the space.” Investors banking on a resolution to the current trade war can buy IGPT.
“IGPT targets about 100 companies from across the globe that generate revenue from various forms of software and AI, such as data storage, robotics, autonomous vehicles, semiconductors and web platforms,” Reyna says. The benchmark for this passive ETF is the Stoxx World AC NexGen Software Development Index. It is more expensive than XAIX, with a 0.58% expense ratio.
Global X Robotics & Artificial Intelligence ETF (BOTZ)
“When you think about smartphones, laptops or even mobile applications, lower prices and cheaper development costs didn’t shrink the market but expanded it as innovation accelerated,” Dessai explains. “AI could follow the same trajectory, embedding itself into the physical world — from factories and drones to delivery vans and buildings.” To capture this trend, Global X offers BOTZ.
Unlike the larger Global X Artificial Intelligence & Technology ETF (AIQ), BOTZ focuses less on software developers and more on the industrial and health care companies deploying AI to automate workflows. “We see BOTZ as a more niche play on applied automation,” Dessai says. The ETF charges a 0.68% expense ratio and is fairly tax efficient thanks to a low 0.1% 30-day SEC yield.
Roundhill Generative AI & Technology ETF (CHAT)
Not all AI ETFs passively track an index. Some, like CHAT, are actively managed. This ETF features a fairly concentrated portfolio of just 38 companies, selected according to the portfolio manager’s fundamental research. However, active management typically comes with higher costs. CHAT’s 0.75% expense ratio isn’t excessive, but it does clock in significantly higher than the previous index-based ETFs.
“CHAT selects stocks using a proprietary methodology that combines a transcript score and sector score to evaluate companies’ relevance to generative AI, factoring in their revenue, profit and R&D investment in AI technologies,” explains Dave Mazza, CEO at Roundhill Investments. “Companies are then scored and selected based on their exposure to AI, market capitalization and liquidity.”
KraneShares Artificial Intelligence & Technology ETF (AGIX)
AGIX is a highly unique AI ETF. It starts by allocating at least 80% of its assets in stocks representing the Solactive Etna Artificial General Intelligence Index. However, what makes AGIX special is the ETF’s ability to own privately held companies up to a 15% cap. Currently, this includes Anthropic, which is AGIX’s fifth-largest holding at a 3.9% allocation. However, this exposure does come at a higher 1% expense ratio.
“KraneShares has a fair value committee that assigns a fair value to private holdings, based on factors such as primary round valuation, secondary market transaction prices and other material information,” explains Derek Yan, senior investment strategist at KraneShares. “Moreover, the maturing secondary private transaction market could provide liquidity if we want to change the exposure to Anthropic.”
VistaShares Artificial Intelligence Supercycle ETF (AIS)
“If you run an analysis between the S&P 500 and most AI ETFs, you will see high overlap,” says Adam Patti, CEO of VistaShares ETFs. “Not so with AIS, which consists mainly of the ‘picks and shovels’ of AI infrastructure via companies that most investors may not own.” Investors won’t find Magnificent Seven names like Microsoft or Amazon in this ETF’s top holdings, but Nvidia Corp. (NVDA) is currently No. 10, at a 3.2% allocation.
“AIS leverages our proprietary ‘Bill of Materials’ methodology that was designed by AI luminaries on our investment committee,” Patti explains. “We analyze the supply chain, layer in the actual costs to build AI data centers and semiconductors, and then dig into the company financials to determine their true AI business pipelines and related revenue.” This active ETF charges a 0.75% expense ratio.
More from U.S. News
8 New and Upcoming IPOs in 2025
7 Best Drone Stocks to Buy in 2025
5 Best Nuclear Energy Stocks and Funds to Buy Now
6 of the Best AI ETFs to Buy for 2025 originally appeared on usnews.com
Update 06/10/25: This story was previously published at an earlier date and has been updated with new information.