AI reasoning models create new $1B+ startup opportunities as investment hits $110B

Andreessen Horowitz partner Anjney Midha predicts a new “golden age” of AI investment opportunities will emerge from specialized startups using reasoning models and reinforcement learning to solve industry-specific problems. Speaking at Fortune’s Global Forum in Riyadh, Midha challenged the prevailing view that only a few major AI labs would dominate the market, arguing that new AI paradigms are creating space for vertical-focused companies to build multibillion-dollar businesses.

What’s driving the shift: Reasoning models have fundamentally changed the competitive landscape by enabling more sophisticated problem-solving approaches than traditional language models.

• These AI systems can evaluate their own outputs, break complex tasks into sub-tasks, and learn from feedback through a process called reinforcement learning—essentially teaching AI to improve by rewarding good decisions.
• “Reinforcement learning as a new paradigm is working so extraordinarily well, especially on mission-critical problems,” Midha said.
• Startups can leverage this technology by embedding themselves deeply within specific industries and understanding customer problems end-to-end.

The investment surge continues: Venture capital funding in generative AI has reached unprecedented levels in 2024, defying concerns about an industry bubble.

• Total funding is on pace to more than double from last year, with investors pouring more than $73.6 billion into GenAI application startups in the first three quarters.
• Overall investment across the GenAI and broader AI ecosystem reached $110.17 billion this year, representing an eightfold increase since 2019.
• Major foundation model providers continue commanding massive rounds, with OpenAI’s $40 billion funding remaining the largest deal, followed by Anthropic’s $13 billion round and Mistral’s €1.7 billion Series C.

Geopolitical concerns: Midha expressed worry about China’s growing dominance in open-source AI development, calling it “China’s game right now.”

• This trend poses challenges for the U.S. and its allies in maintaining competitive advantage in AI development.
• Western labs are scrambling to catch up, which Midha predicts will result in a wave of open-weight models from American companies.

What they’re saying: “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,” Midha explained.

• “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.”