Cathie Wood Says Software Is the Next Big AI Opportunity — 1 Super Stock You’ll Regret Not Buying in 2025 If She’s Right

Morgan Stanley estimates that just four companies — Microsoft, Amazon, Alphabetand Meta Platforms (META 3.51%) — will spend a combined $300 billion on data center infrastructure and chips during 2025 alone, to support their ambitions in the artificial intelligence (AI) space.

Cathie Wood is the founder of Ark Investment Management, which operates several exchange-traded funds (ETFs) focused on innovative technology stocks. She believes software companies will be the next big opportunity in AI, predicting they could generate $8 in revenue for every $1 they invest in chips from suppliers like Nvidia.

Those four tech giants could earn a mind-boggling payoff from their AI infrastructure spending if she’s right, but here’s why investors might regret not buying Meta Platforms stock, in particular, in 2025.

Meta’s Llama 4 could be the most powerful AI model in the world next year

Meta is the parent company of social networks Facebook, Instagram, WhatsApp, and Messenger. Together, those apps serve almost 3.3 billion people around the world every day, and AI is playing an important role in shaping their experience.

Meta developed an AI-powered content recommendation engine that learns what each user likes to see, and uses that information to curate their Facebook and Instagram feeds. CEO Mark Zuckerberg says this strategy drove an 8% increase in the amount of time users spend on Facebook this year, and a 6% increase for Instagram. That means each user sees more ads, which translates into more revenue for Meta.

Meta is also launching new AI-powered features like Meta AI, a virtual assistant accessible through its apps. It can generate images and text, and even join group chats to settle debates or offer recommendations for fun activities. The quality of an AI assistant depends on the large language model (LLM) upon which it is built, and since Meta has an enormous pool of data from billions of posts on its social networks, it was able to create an advanced family of LLMs called Llama.

Most popular LLMs (like those developed by OpenAI and Anthropic) are closed source, whereas Llama is open source. That means millions of developers are regularly digging through the code, which helps Meta rapidly identify bugs and improve its functionality.

Llama 3.2 is the latest version of Meta’s flagship LLM, but Zuckerberg says the company is on track to launch Llama 4 in 2025. It’s improving so quickly that he believes it will be the most advanced in the entire industry. That would be an impressive accomplishment, considering start-ups like OpenAI were initially several years ahead.

Meta knows how to monetize new technologies

Meta generated a record $40.6 billion in total revenue during the recent third quarter of 2024 (ended Sept. 30). Most of that revenue comes from selling advertising slots to businesses, and the company has an incredible track record of launching new features and monetizing them in that way.

Meta originally placed ads in its content feeds to catch users’ attention while they were scrolling through posts from friends and family members. But video content formats like Stories and Reels became significant contributors to the company’s revenue, even though rival platforms like Snap‘s Snapchat and TikTok were first to market with similar features.

Meta AI already has over 500 million monthly active users, and while it’s free to use, the company says there will be opportunities to monetize it in the future. Personally, I think advertisers would pay money to link their websites or products within Meta AI’s responses when a user asks a relevant question.

But Meta AI also paves the way for new features like Business AI. Zuckerberg believes every business using Meta’s apps will eventually have its own AI-powered virtual agent, which can handle incoming inquiries from customers, and potentially even process sales. Businesses could customize their agent to suit their own needs by training it on their product inventory, or by teaching it to respond to customers in a way that fits their brand.

Meta will almost certainly charge money for a tool like Business AI, potentially on a per-message basis or even on a subscription basis, which would unlock entirely new revenue streams.

Meta stock is cheap compared to its peers

Meta stock soared 532% from its 2022 low point of around $90. However, based on its trailing 12-month earnings per share (EPS) of $21.23, it trades at a price-to-earnings (P/E) ratio of just 27.2, which is still cheap.

It’s a 15% discount to the Nasdaq 100which trades at a P/E ratio of 32.2. It also makes Meta cheaper than every other AI stock with a valuation of $1 trillion or more, except for Alphabet, which is currently fighting several regulatory battles against the U.S. government:

PE Ratio data by YCharts

Meta stock looks even cheaper on a forward basis. Wall Street’s consensus estimate (according to Yahoo!) suggests the company will generate $25.33 in EPS during 2025, placing its stock at a forward P/E ratio of 22.7. In other words, Meta stock will have to soar by 42% next year just to trade in line with the P/E ratio of the Nasdaq 100.

Llama will be the key to Meta’s success in the AI race from here, because as mentioned, quality AI software starts with a quality LLM — but developing LLMs is not cheap. It took around 16,000 of Nvidia’s H100 graphics processors (GPUs) to train Llama 3, but Llama 4 is going to need over 100,000.

That’s why Meta is on track to spend up to $40 billion on AI infrastructure in 2024, and it plans to spend even more in 2025. In fact, Morgan Stanley thinks the company could spend a whopping $52 billion next year.

If Wood is right, Meta could eventually reap hundreds of billions of dollars in revenue from that AI spending, which will translate into spectacular returns for investors. Given its attractive valuation, Meta stock might be the best pick out of the AI giants to buy next year.