If open-weight models ultimately win the LLM market—which I believe they will—the AI industry is going to look a lot like the auto industry.
Before long, people will spend $200 to $500 a month on AI, much like a monthly car lease. After all, we use cars to extend what we can do physically; we will use AI and software intelligence to extend what we can do mentally.
Instead of paying a single frontier lab like Anthropic, OpenAI, or Kimi, consumers will likely pay application-layer agents, telecoms, or subscription bundles (think Amazon Prime or Netflix). Whether bundled or unbundled, that revenue will ultimately flow to inference providers like Near AI and Together AI that actually serve the models. As these tools become essential to everyday life, people will gladly keep paying for them.
In the short term, OpenAI and Anthropic have every incentive to pitch investors on the idea that they are building winner-take-all monopolies. However, the rapid rise of open-weight models suggests the field won’t be so easy to lock up. At the frontier, the models simply aren’t all that different from one another.
We are seeing the exact same dynamic at the agent layer, where open-source alternatives are already fiercely competitive—including WebBrain (disclosure: my own venture), OpenCode, Osaurus, and Hermes.
This competition matters. When capable models and agents are widely available, industry value doesn’t just pool at the companies training the flashiest foundation models. Instead, it flows toward the companies providing reliable inference, building specialized products, and finding practical ways to bring AI into people’s daily lives.
None of this means OpenAI and Anthropic will fail.
Both companies are trying to convince the market that they deserve trillion-dollar valuations. By historical standards, that is a tough case to make—not even legacy automakers are valued that highly, with Tesla being the lone, massive outlier. Still, just as Elon Musk persuaded public markets to price Tesla as an exception, OpenAI and Anthropic may very well pull it off for similar strategic reasons.
Their lasting advantage, however, won’t be a permanent monopoly over weights. It will be their ability to stay a half-step ahead at the frontier just long enough to compound their lead. If they can consistently field the most capable systems first, they can use them to generate high-value intellectual property—like discovering new drugs or material inventions, or even cracking cryptography to claim Satoshi’s Bitcoin—while companies like Alphabet capture value through ecosystem plays like Waymo.
Overall, I’m deeply bullish on the broader AI market. Consumers will pay for inference, and the industry will mature into a diverse ecosystem of providers and products. I’m slightly more cautious about the post-IPO trajectory for OpenAI and Anthropic. Once the public market realizes no single lab can own the entire AI economy, there will likely be a period of disillusionment. Even if investor confidence dips for a while, though, those who stay patient should eventually make their money back.
tags: