Kimi K3, Qwen 3.8, and Anthropic's potential unravelling

The AI model market is fragmenting faster than anyone predicted, and if you're a small business owner trying to figure out which AI tools to trust with your workflows, the ground is shifting beneath your feet right now.

The Quiet Power Shift Nobody's Talking About

While most small business owners are still getting comfortable with ChatGPT, the underlying competitive landscape for AI models has become genuinely chaotic. New capable models from Chinese labs like Moonshot AI (Kimi K3) and Alibaba (Qwen 3.8) are arriving with performance that genuinely challenges the established players, at a fraction of the cost to run. These aren't novelty releases; they're production-grade models that businesses and developers are actually switching to.

At the same time, there are serious questions circulating in the AI industry about the long-term financial sustainability of some of the most prominent Western AI labs. Building and maintaining frontier AI models costs an extraordinary amount of money, and not all of the companies you've come to rely on are on solid ground. Anthropic, the maker of Claude (one of the most popular AI assistants used in business tools), is reportedly facing significant financial pressure. Nothing has collapsed, but the signals are there for anyone paying attention.

The Infrastructure Scramble Underneath It All

There's a related development worth knowing about. Together AI has just partnered with Y Combinator to create a dedicated GPU cluster for YC startups. GPUs are the specialised hardware that runs AI models, and access to them has been a genuine bottleneck, often requiring expensive two-year contracts that only well-funded companies could stomach.

This matters to you even if you've never heard of Together AI. When startups get easier access to AI infrastructure, more AI-powered tools get built, faster. The software tools you use for your business will likely incorporate more capable AI features over the next twelve months than they have in the last three years combined.

What This Means If You Run a Business

The tools you rely on today might look very different by the end of the year. If your project management software, your email platform, or your customer support tool is powered by a specific AI model under the hood, and that lab runs into financial difficulty or gets undercut on price by a Chinese competitor, your provider may swap the model out entirely. Sometimes this is an upgrade. Sometimes it isn't.

The more immediate practical implication is about cost. As more capable models become cheaper to run (thanks to competition from these newer entrants), AI features that were previously expensive to offer should become more accessible in the tools aimed at small businesses. That's genuinely good news if you're not on an enterprise budget.

The risk is building too deep a dependency on any single AI provider before the market has settled. Automation workflows that rely entirely on one specific AI tool are worth auditing now, not after something breaks.

What To Do About It

1. 1.Audit your AI dependencies. List every tool in your stack that uses AI. Note which company provides the underlying model if you can find out. This gives you a clear picture of where you're exposed if something changes.

1. 1.Don't build automation on sand. If you're using tools like Zapier or Make to build business workflows, prioritise ones that let you swap AI providers without rebuilding from scratch. Flexibility here is worth more than marginal performance gains right now.

1. 1.Watch your vendors' communications. If the AI tools you use start quietly changing their model providers or pricing, that's a signal worth noticing. Sign up for product update emails you'd normally ignore.

1. 1.Keep one eye on the open-source and multi-model platforms. Tools that give you access to multiple AI models (rather than locking you into one) are becoming increasingly practical. We're seeing clients benefit from this approach already.

1. 1.Don't panic, but don't sleepwalk either. This market will settle eventually. In the meantime, stay curious and stay flexible.