Generative AI is getting harder to read, not easier. Elad Gil, a veteran tech investor who predicted in 2021 that frontier LLMs would consolidate into an oligopoly, now admits the market is outpacing his models. His prediction is largely holding: OpenAI, Google, Anthropic, Meta, and Mistral are the remaining credible frontier players, and the cost to train at that level keeps climbing while commodity model prices drop. Training a GPT-3.5 equivalent costs roughly 5x less today than two years ago.
The funding structure is the most important dynamic to understand. Microsoft put $10 billion into OpenAI. Amazon and Google together put $7 billion into Anthropic. Venture capital is a rounding error by comparison. Azure grew 6 percentage points from AI in Q2 2024 alone, representing an annualized revenue increase of $5 to $6 billion, roughly 50 percent of Microsoft's total OpenAI investment, per year. Meta announced a $20 billion compute budget. The entities financing frontier AI are cloud providers with direct revenue incentives and nation-states backing local champions like the UAE with Falcon. That concentration is selecting winners now.
The original piece goes further than the funding layer. Gil works through open questions at every level of the AI stack, from LLMs and compute to applications and business models, and he is explicit that he does not have answers. That intellectual honesty is what makes it worth reading in full. The questions he cannot answer are more useful than the consensus views everyone else is publishing.
[READ ORIGINAL →]