Yes, AI Will Make Us Rich and META Just Proved It

We love our job: writing about cutting-edge technology developments for the Gilder’s Technology Report. And we are super-bullish on coming growth in the tech sector, especially semiconductors and information tools.

BUT, as we keep reminding our readers, though tech is the top tool of wealth creation, the real wealth machine is the souped-up economy that uses the tools. When the world begins to realize how powerful a fount of wealth creation AI truly is, we expect to see a multi-decade bull market that could easily match what we have seen in the 42 years from 1982-2024. (Though we would not be surprised to see a longish sideways market embedded somewhere in the Bull, as we did from 2000 to 2009).

The return on stocks will come not just from the deep-tech companies we love, but from the broader market. Deep-tech will continue to see revenues and earnings rise at a stellar rate. But for the return on any stock, multiple expansion (e.g. rising Price/Earnings or Price/Sales ratios) is as important as earnings growth. Multiples in the tech sector can hardly get much higher. The current Price/Sales ratio of the NASDAQ 100 is an astounding 6.5, more than twice the already above average P/S of the S&P 500 at 2.8.

For explosive returns, look to companies that have figured out—and demonstrated—how to use AI to make real money.

We expect the first tranche of these successes to be closely “tech adjacent.”

For Exhibit A, we present Meta Platforms (NASDAQ: META), aka Facebook, and its just-issued Q4 2024 earnings release. That document should put to rest the whining that AI is too expensive, and we don’t know what to do with it.

We hear the objection: Isn’t Meta a tech company? It builds its own AI models, commissions its own AI chips, and builds vast data centers.

All true, but it’s still a good test case for conventional businesses making money on AI. Meta makes its money in a very traditional way: ad sales account for circa 98% of its revenue.

Meta’s Q4 earnings announcement made clear that AI is super-charging Meta’s ad sales. Q4 2024 revenue reached $48.4 billion, up 21% over the year-ago quarter. Earnings did even better, surging to $20.84 billion, a 64% boost. Earnings per share came in at $8.02, significantly outperforming previous estimates and keeping the stock’s PE down to a robust but not ridiculous 28.4, a shade below the S&P 500’s.

Source: Data from Meta, table by authors.

What happened? AI happened. Meta’s AI-driven ad targeting drove the revenue increase while keeping costs down. The company’s AI investments have been huge but have added more in efficiency than cost.

One key contributor, introduced in the second half of 2024, is Andromeda, a new machine learning system developed with Nvidia (NASDAQ: NVDA). During the earnings call, CFO Susan Li claimed Andromeda “enabled a 10,000x increase in the complexity of models we use for ads retrieval, which is the part of the ranking process where we narrow down a pool of tens of millions of ads to the few thousand we consider showing someone.” That “complexity” word always makes us shudder, but Li said it had driven “an 8% increase in the quality of ads” according to tested objectives.

The company’s AI-powered recommendation engines significantly improved ad engagement, boosting revenue per user. Advanced large language models optimized ad placements, boosting advertiser return on investment.

On costs, CEO Zuckerberg said he expects “2025 will be the year when it becomes possible to build an AI engineering agent that has coding and problem-solving abilities of around a good mid-level engineer.” He emphasized the result of being able to build better AIs. He did not so much as mention how much cheaper robot engineers are compared to humans. That’s okay Mark, we heard it anyway. But robots are only useful if they enhance human productivity, and people do not get employed because they are unproductive. AI will succeed to the extent it makes humans more productive and thus more employable.

The Reality Labs division, responsible for AR/VR and metaverse development, including the recently abandoned “Quest” headset, which failed at a price point of $1,500, continues to be a financial burden. In Q4, Reality Labs generated $1.07 billion in revenue, up from $0.27 billion in Q3, (aided by those RayBan smart glasses). Still the division posted an operating loss of $4.64 billion, bringing total cumulative losses since 2020 to over $62 billion.

DeepSeek or no (and doubts grow about some of that company’s cost saving claims) META is all in on heavy metal AI, its commitment strengthened by its own AI-driven ad revenue and efficiency growth. Full-year 2024 CapEx summed to $41.54 billion, a 48% YoY increase. That included buying some 1.3 million GPUs, nearly all from NVDA, and an additional 1GW of data center capacity coming online this year. On the horizon: the biggest AI datacenter yet, which Zuckerberg said will be “so big that it’ll cover a significant part of Manhattan if it were placed there.”

Next year’s expected Cap-Ex bill: $60-65 billion.

Source: Data from Meta, table by authors.

Competitive AI Market Positioning

Meta’s $17 billion AI investment in 2024 surpassed those of Alphabet (NASDAQ: GOOG), Amazon (NASDAQ: AMZN), and Microsoft (NASDAQ: MSFT), as shown in Table 3.

Responding to concerns that cheaper AIs from, e.g. DeepSeek, would lead to buyer’s remorse for all this hardware spending, Zuckerberg minimized the potential impact. DeepSeek’s efficiency is primarily relevant for inference rather than the deep training that happens in Meta’s massive data centers. Meta’s AI models, including Llama 3 and the upcoming Llama 4, are deeply integrated into META’s own advertising ecosystem and social media platforms and already making money. The firm expects Llama 4 to be even more profitable and enhance the business with broader rollout of AI-powered features across Facebook, Instagram, and WhatsApp.

Investor Takeaway

We have never added META to our Gilder’s Technology Report recommended portfolios because we are not experts in the ad business. The more Meta’s ad business becomes an AI business, the more we think we need to reconsider. We will work that through in an upcoming issue of the Gilder’s Technology Report.

In the meantime, don’t let anyone tell you AI isn’t making anyone rich (besides Nvidia). It’s going to make us all a lot richer a lot sooner than the experts predict.