OpenAI set off the artificial intelligence race when it released ChatGPT 3.5 four years ago. Since then, AI has morphed from a niche portion of the modern technology movement to the ubiquitous driving force behind most of the industry. While OpenAI and Anthropic have been the cream that has risen to the top so far, there are three major players that are spending hundreds of billions of dollars to catch up with them: Meta, Google, and SpaceX. John Belton, portfolio manager at Gabelli Funds, spoke with TheStreet about the current AI landscape and the “other” companies that most investors don’t think about first when they think about AI. Meta learned its lesson from Llama The developments coming from Meta Superintelligence Labs (MSL), its internal AI lab, have been more forward-looking and big-picture, according to Belton.This team is different than the one that developed the ill-fated Llama portfolio of large language model products, which “had pretty clearly fallen behind the frontier,” he said. But have those changes been enough for Meta to overtake market leaders OpenAI and Anthropic?Not quite, but MSL’s work has been enough to reposition Meta in a better position than it was six months ago. “Where we stand today, my view is that there’s still a gap between OpenAI and Anthropic and everyone else, but Meta has definitely shown a lot of progress,” Belton said. “They’ve identified a few areas they’re prioritizing, including personal agents, business agents, and LLM technology for their ad models and content-recognition systems for ad targeting. In other words, they’re prioritizing models that are useful for the things their existing business does best, particularly their core advertising business,” Belton added.While Meta hasn’t caught its rivals, Belton does believe the company is executing a more focused and targeted strategy than Anthropic and OpenAI, whose approach is “much more horizontal and centered on building pure intelligence.””I don’t know if the long-term goal there is some form of AGI or superintelligence. Meta’s strategy seems more targeted toward specific use cases, applications, and markets,” Belton said. “And I think Google is now sort of pivoting toward the same approach.”Google loses top-3 AI statusGoogle has never reached the levels OpenAI and Anthropic have with its artificial intelligence ambitions, but it has carved out a nice spot as the third-best with Gemini. But muted comments during Alphabet’s earnings call leads Belton to believe that Google’s AI efforts may be losing a bit of steam. “The message coming off earnings, to me, felt like: we’re still committed to the frontier, or to remaining at the frontier, and to investing in this effort, but for whatever reason, we’ve lost some ground and some focus,” Belton said. It’s not like Google is abandoning developing AI, but the feeling now is that it has more so shifted its focus to applications. “I think Demis Hassabis (chairman and co-founder of Google DeepMind), at I/O, painted a picture where Google had been focused on real-world models—multimodal models, rather than just text-based models. In focusing on that, I think they also deprioritized a major commercial application for these models, which was coding,” Belton said.
SOPA Images / Getty Images
SpaceX near-term estimates need revisionSpaceX analysts may have the company’s story wrong, according to John Belton, portfolio manager at Gabelli Funds, since the space exploration company is “aggressively building out infrastructure” and has a path towards a very lucrative “neocloud” leasing program. Neocloud is a leasing program where providers like SpaceX rent out AI computing infrastructure to tech companies like Anthropic and Google. The revenue stream is so lucrative for SpaceX that the company made more money doing that than from space launches.SpaceX reported second-quarter neocloud revenue of $1.6 billion, all of which came from Anthropic. Deutsche Bank analysts expect that deal alone to ramp up to $3.75 billion in the third quarter.”What’s interesting to me about SpaceX is that they’re very clearly aggressively building out infrastructure and have a line of sight to a big business in leasing infrastructure—I guess you could call it a neocloud business,” Belton said. “Given that, I think estimates need to be revised significantly higher in the near term.”Belton says the current environment is great for neocloud companies because supply and demand are “way out of balance” as there isn’t enough infrastructure supply to satisfy all the demand, “which is benefiting incumbent computing infrastructure platforms,” like SpaceX is becoming. But the current environment won’t last forever, so Belton’s SpaceX comments are only relevant for now.”Eventually, that situation will normalize, and when it does, not every platform will have the same business model. There are very different services being provided, and some are much more commoditized. So, for SpaceX right now, there’s clear upward pressure on estimates,” Belton said. “But I think there will come a time when some of the sources of this acceleration, and the quality of those revenues, get called into question.”Related: SpaceX analyst plots path to bold $100 billion claim