SAN FRANCISCO — Three data points arrived Thursday that, taken together, suggest the artificial intelligence industry is entering a more turbulent competitive phase than its public valuations have priced in.
First: China's Moonshot AI released Kimi K3, a freely available model that topped Anthropic's Claude Opus 4.8 on at least one major coding benchmark. The release spooked markets. Anthropic, which closed a $3.5 billion Series E at a $61.5 billion valuation earlier this year, is now watching a Chinese open-weight model compete directly with its premium tier. Moonshot is not a household name in the West. It is now.
Second, and almost simultaneously: Meta is reportedly in talks to lease compute capacity to Anthropic in a deal that could reach $10 billion. The irony is structural. Meta and Anthropic are direct competitors in the foundation model market. Yet Anthropic apparently needs GPU access badly enough to buy it from a rival. Meta, which has spent tens of billions building data center infrastructure, apparently sees a new line of business in reselling that capacity. Compute scarcity is forcing strange alliances.
The subtext in both stories is the same: the cost of staying at the frontier is compressing margins and warping competitive logic. A company can release a state-of-the-art model and watch a freely distributed Chinese alternative match it within months. It can raise billions and still need to lease infrastructure from an adversary.
The third data point is less strategic but culturally telling. AI-generated books — unauthorized biographies, generic how-to titles, bulk content — are proliferating on Amazon at a scale that human reviewers cannot track. One journalist discovered an AI-written biography of herself she never authorized. The books are largely unreadable. They are also effectively free to produce and cheap to list. The economics select for volume over quality.
For Trilogy portfolio companies building on top of AI infrastructure — from DevFactory's engineering pipelines to Klair's financial analytics — the Kimi K3 release matters. Benchmark parity from open-weight models compresses the premium that closed frontier models can charge. That changes the build-vs.-buy calculus for every enterprise software stack in the ESW Capital portfolio.
