As of November 30, 2025, there isn’t enough evidence to judge Chamath’s long‑run claim about a mature AI market.

What we observe today (2024–2025) points against the prediction in the short term:

• Closed, proprietary foundation‑model companies are generating very large direct revenues from their models. OpenAI reports around $10B in annual recurring revenue by mid‑2025 from ChatGPT subscriptions and API usage, excluding Microsoft licensing deals, and is projecting much higher revenue by 2030. (thetechportal.com)
• Anthropic has reached $3B+ in annualized revenue by May 2025 and is estimated at $5B ARR later in 2025, largely from pay‑per‑token API calls for Claude models; its valuation has risen to around $183B. (cnbc.com) These are exactly the kind of high‑margin, closed foundational‑model businesses the prediction said would be unsustainable.
• A recent study on open vs. closed models finds that while open models can be up to 84% cheaper to run and perform comparably, they still account for only ~20% of usage and ~4% of revenue; most economic value is currently accruing to closed‑model providers. (itpro.com)

On the other hand, there is movement toward powerful models being broadly available for free or near‑free:

• Meta’s Llama 3 series provides strong models (up to 70B+ parameters, later 400B‑parameter variants) that are downloadable at zero license fee and usable for most commercial purposes under a community license, even if they are not truly open source. (gigazine.net)
• Other vendors (e.g., Mistral) have also released high‑quality open‑weight models, and open‑weight downloads and usage are rising quickly. (en.wikipedia.org) This partially supports the idea that capable base models may become a cheap or free “substrate.”

Why the overall forecast is still inconclusive rather than right or wrong:

• Chamath explicitly anchored his claim to the “mature” AI market. Today’s generative‑AI ecosystem is still in hyper‑growth: industry forecasts expect the total AI market to grow from about $131B in 2024 to roughly $642B by 2029, with generative AI growing at ~90% CAGR—classic signs of an early rather than mature market. (finance.yahoo.com)
• In past tech waves (e.g., operating systems, databases, cloud infrastructure), it has often taken a decade or more for commoditization and open alternatives to erode margins. We are less than two years from the February 2024 prediction date, so current high revenues and non‑free access don’t yet prove that margin compression and commoditization won’t happen later.

Bottom line:

• Today’s evidence contradicts the prediction if you interpret it as applying to the 2024–2025 market: foundational models are generating substantial direct economic value, and leading models are not “entirely free.”
• But because Chamath framed it as a statement about a mature market with no clear time horizon, it’s too early to say whether the long‑run equilibrium will match his view. The trend could still move toward commoditized, mostly‑free base models with value shifting to applications and data, or it could stabilize with durable, profitable closed‑model providers.

Therefore, given the timeline and the explicit “mature market” qualifier, the prediction is best classified as inconclusive (too early to tell) rather than clearly right or wrong.