Good morning, tech reporter Beatrice Nolan here, filling in for Allie Garfinkle.
“It was very popular two or three years ago to say there’s only going to be three or four labs and teams that are going to do any real training…and startups will be left to pick the pieces up of tiny niche opportunities here and there,” he said.
But reasoning models have changed the game, Midha said, referring to the new generation of AI systems designed to “reason” problems step by step, mimicking logic and reflection rather than predicting the next word in a sequence. These models can evaluate their own outputs better, break complex tasks into sub-tasks, and learn from feedback, potentially bringing AI closer to complex, real-world problem-solving.
“Reinforcement learning as a new paradigm is working so extraordinarily well, especially on mission-critical problems,” Midha said. “If you can define the reward model correctly, which startups are really good at doing when they embed themselves inside an industry—they go deep, they go vertical, and they end up understanding the customer’s problem end to end—you can build entirely new, multibillion-dollar companies doing full end-to-end reinforcement learning for each industry.”
Despite some of the ongoing debate about an AI industry bubble, the investment surge doesn’t appear to be cooling off.
Much of this capital has flowed to large foundation model providers such as OpenAI, Anthropic, and Mistral AI, which continue to command multibillion-dollar rounds and soaring valuations. OpenAI’s $40 billion funding earlier this year remains the single largest deal, while Anthropic’s $13 billion round and Mistral’s €1.7 billion Series C underline the dominance of a handful of major players.
Beatrice Nolan
X: @beafreyanolan
Email: bea.nolan@fortune.com
