Solo VC investor extraordinaire Elad Gil stated on stage at TechCrunch Disrupt that AI has been one of many least predictable tech booms he’s ever seen.

Gil is on the cap desk of nearly each hit firm of the previous decade, together with lots of at the moment’s main AI firms.

Nonetheless, he thinks that during the last 12 months, sure AI markets look like almost sewn up by market leaders. Past these areas, an unlimited swath of AI stays anybody’s recreation.

“I began investing in generative AI in 2021 … on the time, not very many individuals have been paying that a lot consideration to it,” Gil stated. However he had seen the large leap in functionality between GPT 2, launched in 2019, and GPT 3,  launched in 2021. “The step between 2 and three was so giant that if you happen to simply extrapolated out the scaling legal guidelines, or the curve, then you might actually assume that this was going to be extremely vital,” he stated.

That satisfied him to begin backing early-stage startups constructing merchandise powered by giant language fashions. His bets included each foundational mannequin makers like OpenAI and Mistral, in addition to software firms like Perplexity, Harvey, Character.ai, Decagon, and Abridge. But all through 2024 and far of 2025, the capabilities of foundational fashions leaped with each launch, upending AI each few months.

“I used to say on the time that AI was the one market the place the extra I study, the much less I do know. Normally, the extra you study one thing, the higher you understand it, the better you may predict the long run, and so forth. However AI was simply hazy. There’s simply an excessive amount of uncertainty. And I believe there’s nonetheless markets like that in AI,” he stated.

Nevertheless, he’s additionally now seeing markets with clear winners. The obvious instance is with foundational fashions themselves. Regardless that a whole bunch of fashions exist, and a few nations like South Korea are nonetheless working now to develop sovereign fashions by native firms, leaders have emerged. “Google, Anthropic, OpenAI, possibly xAI, possibly Meta, possibly Mistral — it’s like a handful,” he predicts of the winners.

After fashions, he thinks AI-assisted coding has runaway winners that can make it onerous for brand new entrants to catch up. Not solely have the foundational mannequin makers moved in (Anthropic with Claude Code, OpenAI with Codex) however startup leaders like Anysphere’s Cursor and Cognition’s Devin (which acquired Windsurf) might be onerous to beat. And there are well-funded startups like Magic (whom Gil referred to as a doable “outlier”) or Poolside on their tails.

He sees medical transcription as being cornered, with Abridge a entrance runner and a handful of others like Ambiance being “vital.”

He names buyer help – which was an early goal of each conventional AI and the brand new crop of AI agent startups – as having hard-to-catch market leaders, similar to his portfolio firm Decagon. (It raised $131 million at a $1.5 billion valuation in June.) OpenAI chairman Bret Taylor’s startup, Sierra, competes on this area. That is additionally an space the place the incumbents — Salesforce, Hubspot, and plenty of others — are including AI choices.

So which markets appear vast open? Gil says monetary tooling (fintech), accounting, AI safety, and “different markets that we all know are by default very fascinating. We simply don’t know who’s going to do it.”

Sarcastically, quick development isn’t the sign it as soon as was that an organization goes to be a breakout hit. “The CEOs of each huge firm are principally telling their groups, hey, now we have an edict. We have to determine our AI technique,” Gil stated. “These large enterprises are prepared to strive issues that two years in the past they by no means would have tried, and it’s solely due to AI.”

So new AI markets can land quite a lot of income from big-name, enterprise clients shortly, “however that doesn’t imply they’re going to stay,” Gil factors out.

It is just after a market goes via its trial-phase increase cycle {that a} startup and buyers can see if this income will keep and develop. “There’s false sign, after which there’s stuff that’s simply working,” Gil stated. He calls out authorized AI startup Harvey as one of many market-leaders that’s “simply working.” It raised three large rounds in 2025, leaping from a $3 billion valuation to $5 billion to $8 billion, in only a few months.