What was the big story with Meta Platforms‘ (META +0.28%) first-quarter update? Some might say that it was the company’s increased capital expenditures on artificial intelligence (AI) infrastructure. Others could vote for Meta’s strong revenue and earnings growth.

I think, though, that the most important story from Meta’s latest investor update with a long-term perspective is something CEO Mark Zuckerberg briefly mentioned. Zuckerberg just hinted at the next big thing in AI. And investors who pay attention to what’s happening could profit tremendously.

A massive technological shift

In Meta’s Q1 earnings call, Barclays analyst Ross Sandler asked Zuckerberg about the company’s strategy for AI coding and “the recursive self-improvement loop.” Zuckerberg replied that he thinks that “people conflate coding with self-improvement more than they should.” He noted, “Coding is one ingredient for the model self-improving. It’s not the only thing.”

Zuckerberg is right. AI model self-improvement is much bigger than coding. It represents a massive technological shift. But what exactly is AI self-improvement?

It’s first important to understand how AI models have been improved so far. They’re retrained manually. This process takes some time. However, self-improving AI models can refine themselves. They learn continuously through user interactions and real-time data ingestion. AI takes control of its own evolution.

As you might expect, self-improvement should dramatically accelerate advances in AI models. AI that can figure out how to perform tasks better on its own can achieve breakthroughs exponentially faster than human engineers could. What’s more, these self-improving AI models could even get better at getting better.

Zuckerberg alluded to the endgame for self-improving AI during Meta’s Q1 earnings call. He stated that his company is “focused on all of the parts that are going to be necessary for self-improvement in service of the personal superintelligence vision that we have for people and businesses.”

Sure, “personal superintelligence” is one of those corporate phrases that tech CEOs like to throw around. However, AI with capabilities beyond human levels is at the heart of Meta’s goal. Self-improving AI models could make superintelligence a reality sooner than anyone expects.

Three big winners

Zuckerberg understandably wants Meta to be a major player in developing self-improving AI. He said in the Q1 earnings call, “I think the companies that don’t do that [developing self-improving AI] are not going to be leading labs; they’re not going to produce leading products.” While I suspect that Meta will be in the race, I think three other AI stocks will be the biggest winners from self-improving AI.

Nvidia (NVDA +0.04%) appears to be a no-brainer to profit from advances in self-improving AI models. The continuous training involved in self-improving AI is highly GPU-intensive. Despite a flurry of activity by other companies to develop AI chips, Nvidia’s GPUs remain the standard for training AI models.

Nvidia

Today’s Change

(0.04%) $0.07

Current Price

$198.52

Self-improving AI models will also blur the line between training and inference. The internal processing required for the level of inference will drive demand for more powerful GPUs. Again, Nvidia is in the best position to benefit.

Broadcom (AVGO 1.15%) is another obvious winner. Self-improving AI models don’t use just one chip; they use tens of thousands in a cluster. Moving data between the GPUs requires super-fast networking. Broadcom is the leader in this technology.

Broadcom

Today’s Change

(-1.15%) $-4.85

Current Price

$416.43

In addition, the race to develop self-improving AI will almost certainly prompt big tech companies to accelerate their efforts to improve their own AI chips, reducing their reliance on Nvidia. Broadcom’s custom silicon is a key ingredient.

Last, but not least, I think that Alphabet (GOOG 0.84%) (GOOGL 0.59%) is in an excellent position to profit from self-improving AI. Google’s parent is arguably the leader in the pack in “Self-Taught Reasoner” (STar) methods. Google DeepMind is also a pioneer in reinforcement learning, the primary engine behind self-improving AI.

Alphabet

Today’s Change

(-0.59%) $-2.29

Current Price

$383.40

Additionally, Alphabet owns the full stack involved with self-improving AI, from its own AI chips to its Vertex AI platform. Google Cloud is the fastest-growing major cloud services provider. As organizations adopt AI to take advantage of its rapidly growing capabilities through self-improvement, Google Cloud will likely see even more growth.

The next big thing

Zuckerberg’s comments weren’t idle speculation or mere wishful thinking. He correctly pinpointed exactly where AI is heading next.

Self-improving AI really is a huge deal. We’re moving from a world where humans improve AI models to a world where the models improve themselves. I suspect that investors who buy Nvidia, Broadcom, and Alphabet will be glad they did in a few years when the full impact of self-improving AI becomes clear.