AI Capital Raise Prediction Came True
Also, have LLMs maxxed out on "intelligence?
Back in January I wrote “Hyperscaler Equity Issuance in 2026?” given that AI is CapEx hungry. Yesterday Google announced its $80 billion capital raise. Google is busy building its “world model”. At some point I will bring Google’s Gemini and open source models into Kilby as model options (we are exclusive to Anthropic models at the moment).
Google (GOOGL) has the better business model as compared to Anthropic and OpenAI, even if Gemini significantly lags in coding performance, and despite the changes being forced upon Google Search.
Software code automation is an important space for LLM vendors to win given that coding represents a massive recurring revenue stream that will soon exceed a $trillion in revenue if it does not already, especially when you consider ongoing code maintenance and cybersecurity elements.
Changing gears, the cynic in me suspects that Anthropic and OpenAI are rushing to IPO to both capitalize on ridiculous valuation multiples, and to hit the bid just as Revenue growth stalls. After heavily using Opus 4.8 for the past 4 days, I am disappointed in the model’s performance, it feels flat with Opus 4.6 (4.7 was a downtick from 4.6). I doubt Mythos will be much better - in which case we are probably done with model improvements until 2027.
The big question for 2027 is - how much will the models improve? Or, has model “intelligence” peaked, and now the name of the game is to continue to add tools to the harness (build and acquire) and build the ecosystem around the harness - more third party connections, perhaps an MCP partner store (similar to the App Store) - the type of activities one does when in harvesting mode that were impossible to perform during hypergrowth mode.
It may be incremental model performance improvements from here on out, in which case Anthropic and OpenAI have managed to top tick their respective growth rates (OpenAI is actually a bit off its peak growth).