A major power shift just landed.

Chinese open-source AI models have overtaken US models in global downloads.

MIT and Hugging Face data shows Chinese developers taking 17 percent of global downloads over the past year, pushing past US creators at 15.8 percent. DeepSeek and Alibaba’s Qwen are scaling rapidly, while US giants continue to prioritise closed systems.

The signal is straightforward: open models are moving faster, cheaper and with broader developer pull. This changes how organisations should think about capability, cost and competitive tempo.

So what now?

Companies should treat open-source AI as a strategic asset, not an experiment. They should build the capacity to benchmark and integrate multiple open models, rather than default to a single proprietary provider. This reduces cost, avoids lock-in and increases flexibility as the frontier shifts.

They should invest in lightweight internal platforms that let teams test, fine-tune and deploy models quickly, using open foundations where practical. The pace of improvement in open ecosystems means firms that develop this muscle will ship products faster and at lower cost.

They should also upgrade talent models. Open-source strength rewards organisations that can attract engineers, data specialists and product teams able to work hands-on with model families that evolve monthly.

For government services, the same logic applies: design for interoperability, build internal capability and avoid long-term dependence on proprietary stacks that cannot keep up with open-source velocity.

The centre of gravity in AI is moving. Organisations that adapt their operating model to the new tempo will lead; those that hesitate will be overtaken.

Expect the FUD-factor to be dialled up in coming weeks from the likes of OpenAI et al.