Realtime AI News
US Firms Send Nearly 90% of Open-Source Model Spending to Chinese Companies, Kimi Takes 61%
BigGo Finance reports that nearly 90% of US enterprise spending on open-source models goes to Chinese companies, with Kimi alone accounting for 61%. The shares describe a highly concentrated purchasing pattern rather than dispersed price competition.
BigGo Finance reports that nearly 90% of US companies' spending on open-source models flows to Chinese firms, with Kimi taking 61%. Those two figures are the core of the story.
The report presents the market as shares: close to 90% for Chinese vendors as a group, and 61% for a single model family. The second number means Kimi is not one participant among many in that spending but the majority holder.
It is worth noting what the headline and summary leave out. There is no stated methodology, no time window, no sample of companies and no indication of the underlying data source. Whether the count covers API calls and hosted services only, or also folds in the engineering cost of self-hosted deployments, would change the conclusion considerably.
Even so, the distribution is worth recording. Open-source model purchasing has long been described as fragmented, price-sensitive and cheap to switch away from, and concentration at this level suggests buyers have converged more than expected.
Put together, the near-90% share and the 61% figure suggest this is not merely about Chinese models selling well, but about purchasing in open models having become highly concentrated. Concentration has two consequences: bargaining power moves toward a few suppliers, and tooling ecosystems coalesce around the leading models.
For Chinese model makers this is evidence of market position. For US enterprises it is a question about supply dependency, especially as data governance and compliance review tighten. The same dataset reads differently from each side.
What to watch is whether the full report publishes its methodology, and whether the shares keep moving in the same direction in the next period. A single reading can show a trend without proving that it has settled.
Why it matters
If the methodology holds, competition in open models has shifted from scattered trials to a concentrated head, and bargaining power moves with it. For enterprise buyers, supply dependency and compliance exposure rise together.
Nearby Updates
All11/06, 21:02
Underdog demos a 32MB speech model that also generates music
Underdog has shown off a new demo of a speech model that is only 32MB in size, and despite the tiny footprint it can generate both speech and music. The report covers the demo alone, leaving sample rate, audio quality, supported languages and latency undisclosed.
11/06, 20:06
Tencent Cloud Enterprise launches full-stack private cloud for sovereign agentic AI
Tencent Cloud Enterprise has announced a full-stack intelligent private cloud aimed at sovereign agentic AI, disclosed via press release. The offering pairs private deployment with integrated delivery, positioning the stack around enterprise agent workloads that need to stay under customer control.
11/06, 18:37
Singapore's IMDA issues safety guidelines for AI agent use
Singapore's Infocomm Media Development Authority has published safety guidelines for the use of AI agents, according to a Lianhe Zaobao report, targeting systems that call tools and execute multi-step tasks autonomously. The guidelines' provisions, scope and whether they are mandatory still need to be confirmed from the official document.
11/06, 17:54
ByteDance Seed posts Cola2, a text model tagged for diffusion and continuous latents
ByteDance's Seed team added a Cola2 entry to its official Hugging Face page, listing a text-generation pipeline with safetensors weights loaded through transformers. The tags point to diffusion, continuous latents and a base-model checkpoint, a combination rarely seen on standard language model releases.