U.S. enterprises are increasingly looking beyond America for their AI needs. As Chinese open-weight models close the gap with their closed-source U.S. counterparts, a growing number of businesses are warming to Chinese alternatives — and new spending data suggests the shift, while still modest, is underway.

The appeal is straightforward. "Open weight" means a model's underlying parameters are published for anyone to download, so businesses can run the systems on their own infrastructure or through third-party platforms rather than depending on a single provider's API. That gives companies more opportunities to fine-tune AI systems, makes deployment much cheaper, and generally hands enterprises more control — including greater assurance that their data will not be used to train potentially competing products. While Anthropic and OpenAI still dominate how American businesses buy AI, the figures point to change at the margins.

According to Ramp's latest AI Index — the spend-management firm tracks token and subscription spending across its business customer base, offering an unusually direct window into how companies actually pay for AI — the share of businesses paying for model serving platforms, which give companies access to open-source and Chinese-developed models, rose to 6.1% of total AI-spending businesses in July, up from 4.5% in January 2026.

The shift may reflect rising enterprise interest in open-weight models such as Moonshot's Kimi K3, which made waves on its release for being both an unusually large open-weight model and one that showed coding and agentic performance close to leading proprietary systems.

Z.AI, another Chinese AI lab, this week confirmed that it built Ox Alpha, a previously anonymous model that had been performing well on AI benchmarks and drew enthusiastic reviews from AI developers over the weekend. The company has renamed it GLM-5.3-Flash and said it will charge $0.15 per million input tokens and $0.50 per million output tokens — another aggressively priced Chinese offering in a market where DeepSeek and Moonshot have already put pressure on rivals' pricing. Because API access to AI models is sold by the token, those rates translate directly into running costs, and sub-dollar pricing makes high-volume workloads far cheaper to operate than frontier alternatives.

Z.ai is also thought to be close to releasing a larger version of GLM-5.3 that many believe will rival some of the best models from Anthropic and OpenAI in cyber capabilities — potentially a watershed event that many fear will usher in a new era of AI-powered cyber attacks for which industry is woefully unprepared (New York Times). So far, that security debate has coexisted with, rather than curbed, the commercial shift Ramp's data captures.

China is now clearly ahead in the open-model race. Hugging Face found that in almost every month of 2026, the largest and most capable open model came from a Chinese lab (State of Open Models). The most notable recent U.S. challengers have come from Thinking Machines Lab and, more recently, Meta — whose earlier Llama releases helped kick off the open-weight movement among major American labs — but those American releases have generally not matched Kimi K3's scale or developer pull.

Alex Brunicki, a partner at Backed VC, said he is already seeing a shift in how enterprises approach open-source models.

"We've seen a lot of companies…developing industry-specific foundation models using open-source models that are then fine tuned on very particular data sets," he said. "They're not necessarily using the frontier models for all of the work that they're doing. They're actually using these open-source models which are free to use."

At least two established organizations have recently and publicly switched to open-source for parts of their business. Thomson Reuters, the legal-information and media company, said this week that it has built an in-house model, called Thomson-1, based on Snowdon, a system the company developed by adapting Alibaba's open-source Qwen model (Computing). The model will handle document-review tasks that previously ran on Claude. As part of the announcement, CTO Joel Hron said companies do not need ever-larger, more expensive models to get useful results, and that starting from a strong open foundation and specializing it deeply can produce capable AI at lower cost.

Harvey, the legal tech firm backed by OpenAI, Sequoia, and Andreessen Horowitz, also recently announced that its new model, Harvey Tenet, was post-trained on top of Moonshot AI's open-weight Kimi K3 (SCMP). The firm said Harvey Tenet outperformed both its base model and U.S. frontier systems — including Fable 5 and GPT-5.6 Sol — on complex legal agentic tasks. Harvey had previously built its product by customizing closed models from Anthropic, OpenAI, and Google — a reminder that even companies funded by the leading U.S. labs are willing to change foundations when the economics favor it.

Ramp's lead economist, Ara Kharazian, wrote that the recent growth of open-source has not yet dented spending on OpenAI or Anthropic directly. New AI buyers are still choosing the established American labs: Anthropic gained the most ground among businesses in July, rising 1.1 percentage points to 43.5% market share, while OpenAI climbed 0.23 points to 39.7%.

However, Anthropic's Fable 5 — thought to be the most advanced model on the market, so much so that the U.S. government briefly suspended foreign access to it for national security purposes — accounted for just 6% of the tokens businesses purchase from Anthropic and 11.4% of the dollars spent on Anthropic models overall. That share comes despite Fable 5 being priced at roughly $10 per million tokens, twice the cost of OpenAI's GPT-5.6 Sol — and many times the sub-dollar rates Chinese open-weight vendors are quoting. GPT-5.6 Sol, by comparison, makes up 25% of OpenAI's tokens and 23% of its spend.

Kharazian argues that Fable 5 has effectively found the market's ceiling: businesses are not willing to pay a premium for the best model on the market when a cheaper one is good enough.

Taken together, the data may suggest that American frontier labs have found the limit of what customers will pay for the newest, most expensive model — and open-source may simply be stepping in to fill that gap. The clearest upcoming signals are the next readings of Ramp's index, which will show whether the model-serving share keeps climbing, and Z.AI's larger GLM-5.3, which will test both how far Chinese open weights can push on capability and the security concerns that come with them.

Elsewhere in AI news:

• Google DeepMind is losing its grip on elite AI talent, new data shows.
• DeepSeek seeks $7.4 billion at a $74 billion valuation.
• OpenAI's data-center head leaves amid a broader exodus.
• OpenAI released more details on the Hugging Face hack.

Reported by Beatrice Nolan for Fortune (beatrice.nolan@fortune.com, @beafreyanolan). Originally published on Fortune.com.