Meta released a Q1 earnings report that exceeded expectations, but provided a Q2 outlook below the median expectation. The primary reason is that higher capital expenditure expectations will affect profit performance.
Due to this slightly lower-than-expected outlook, Meta's stock price fell by more than 15% after hours. This also reinforced market concerns that tech giants are accelerating capital expenditures for AI while returns are slow to arrive.
Will Meta repeat the mistakes of the Metaverse? Massive investments continue to mount, yet commercial returns remain unseen.
In fact, many people—perhaps an increasing number—still believe the Metaverse is not an illusion, myself included. In reality, the accelerated development of applications like digital twins and intelligent driving has benefited from rapid breakthroughs in Metaverse-related technologies. Although only a few are currently paying for it, it must be acknowledged that AR/VR/MR devices are changing the way some people work and live, myself included.
Technological evolution always follows a relatively stable curve—even if it changes from quadratic to log-linear, the trend is always upward. However, when projected onto commercial considerations, it behaves more like a sine pulse; the midpoint moves up with the technology curve, while the amplitude scales with the slope.
We know clearly from Mark Zuckerberg's interview a few days ago: if they hadn't increased investment in computing infrastructure due to the growing competitive pressure from TikTok, Meta would not have had the capacity to release the LLaMA-3 model and make it open for use.
Even though I have always believed that generative AI is a tool for a minority to improve efficiency, and its greatest benefits lie in the B2B sector rather than B2C, it will impact almost everyone very quickly. If that's the case, it doesn't matter if it doesn't generate massive revenue in the short term, because the allure of potential AGI is simply too great. Moreover, generative AI is not failing to generate revenue; it's just that before the inflection point arrives, the growth rate may not keep pace with the investment.
Every technological revolution has been this way.
Memories go back 30 years to the dawn of the web portal era, or for gaming, the MUD (Multi-User Dungeon) era. As a user then—perhaps because of youth or other reasons—I never worried about where the revenue for websites and MUD games came from.
The earliest internet and games changed my generation; they were the strongest mark of the era and the source of our behavioral models. Then came Gen Y and Gen Z. Today, AI is becoming the strongest mark on my daughter's generation, changing how they view the world, their education, and gradually becoming the source of their behavioral patterns.
Yes, every great era is full of minor interludes. However, I have begun to believe, and will continue to believe, that technology is the ultimate remedy for the problems of every era, even though technology simultaneously brings new problems.
This may sound like a grand narrative.
But returning to AI itself, we see that the acceleration of computing power growth has not slowed. The scaling laws brought by data and model size are still working. The work efficiency of more and more people is improving at an unprecedented rate, and the number of practical problems models can solve is growing rapidly.
Of course, we still need to solve the problems of better planning and reasoning in models. That path does not seem entirely clear yet, but our understanding of this problem is much more comprehensive than before, and our ability and methods to solve it are much greater. By comparison, the problem of revenue growth is a minor one.
Returning to the question posed in the title: the Metaverse seems more ephemeral, while generative AI provides more tangible leverage. However, if we compare AGI with the Metaverse, though the former is more shocking and has more believers, Meta is not repeating past mistakes; it's simply placing a larger "bet." If the certainty of this competitive path is not low, then the preemptive investment in infrastructure will likely be proven correct in hindsight.
Written after the release of Meta's quarterly report, and after realizing that I had inevitably allowed my perspective to narrow due to unnecessary emotional influences recently.