Claude Won the Enterprise. The Competition Is Now Chinese.

“Chat est mort. Nos clients utilisent Claude.” That was a Geneva wealth manager’s verdict at a recent panel on AI in Swiss finance. He is mostly right.

Claude has won corporate AI, for now. Ramp, which tracks real corporate card spend, put Anthropic at 34.4% of US business adoption in May 2026, ahead of OpenAI’s 32.3% for the first time. Menlo Ventures puts Anthropic at 40% of enterprise model spend. Google now offers Claude as a partner model on its own Gemini Enterprise platform. When a direct competitor resells your model, the lead is real.

There is one problem everyone running Claude at scale knows. Opus is brilliant and it writes the invoice. LinkedIn is full of posts comparing monthly token bills, only half in jest.

The cheaper answer is not coming from OpenAI. It is coming from China. Moonshot released Kimi K2.6 in April: open-weight, matching GPT-5 and Claude Opus on several coding and agentic benchmarks, at roughly 80% lower token cost. MiniMax shipped M3 on June 1, announced to be also open-weight, one-million-token context, topping the open-weight field on SWE-Bench Pro. Moonshot raised around $2 billion in May at a $20 billion valuation. Investors are placing a specific bet: that cheap, capable, open-weight AI built in China takes real enterprise share.

I tested both on a real build: an autonomous multi-asset dashboard running on an agent harness that learns and improves itself across a run. Build with Kimi, let MiniMax review and correct, and you land close to where you get building with Claude and having GPT-5 review. Not identical. Close enough that for a cost-sensitive workflow, the difference becomes a budget decision rather than a quality one.

That raises an uncomfortable question for European firms. US frontier models are proprietary and run on US hyperscaler infrastructure. Chinese open-weight models can be downloaded and run entirely on Swiss or EU servers, with no client data leaving your jurisdiction. For a firm under FINMA and the EU AI Act, which carries the lower risk: a capable US model on American infrastructure, or a capable Chinese model on your own servers in Zürich or Geneva? There is no clean answer. There is only a documented, deliberate choice, which is exactly what a regulator will ask you to show.

The map has changed. A year ago the conversation was ChatGPT against DeepSeek. Today it is Claude against Kimi and MiniMax, with OpenAI and Google fighting for enterprise share in between.

If you are weighing which models belong in your firm’s stack and how to run them in line with FINMA and the EU AI Act, connect with me on LinkedIn or book a conversation at gerevest.ai.

Sources:

1. Ramp AI Index, May 2026
2. Menlo Ventures, State of Generative AI in the Enterprise, December 2025
3. Moonshot AI, Kimi K2.6 release, April 20, 2026
4. MiniMax, M3 release, June 1, 2026
5. TechCrunch, Moonshot AI raises $2B at $20B valuation, May 7, 2026
6. TechCrunch, Anthropic overtakes OpenAI in business adoption, May 13, 2026

About the Author: Dr. Andreas K. Janoschek specializes in AI applications for Asset & Wealth Management. Based in Geneva, he helps industry professionals navigate at the intersection of finance and technology.

This newsletter aims to inform and does not constitute investment or legal advice. Always consult with qualified professionals for specific circumstances.