Key Takeaways
- Chinese AI startups such as Moonshot (Kimi K3), Z.ai (GLM‑5.2) and DeepSeek are gaining rapid traction among U.S. developers and companies because they deliver performance close to frontier U.S. models at a fraction of the cost.
- Affordability is a major driver: users report paying “a handful of cents per million output tokens” versus $30‑$50 for comparable U.S. services, making the Chinese models “good enough” for most everyday tasks.
- The open‑source nature of many Chinese models facilitates broader adoption and allows users to inspect, modify, and build on the technology, contrasting with the closed‑source stance of firms like Anthropic and OpenAI.
- U.S. policymakers have responded with export controls and accusations of illicit “distillation,” but short of an outright ban, Chinese AI continues to appeal to independent developers seeking cost‑effective solutions.
- While Chinese models are narrowing the performance gap, they still lag behind U.S. leaders in full‑range capabilities, and the rapid growth of these startups raises questions about long‑term financial sustainability.
- Chinese tech giants and the government are pushing AI abroad, positioning open‑source AI as a tool for global equity and seeking to expand their ecosystems beyond domestic markets.
Overview of the Shifting AI Landscape
The United States’ long‑standing dominance in artificial intelligence is being challenged by a wave of capable, low‑cost models emerging from China. As reported by the Associated Press, San Francisco‑based Mozilla CTO Raffi Krikorian switched to Moonshot’s Kimi K3 shortly after its launch, noting that it “just seems snappier” compared with Anthropic’s Claude Fable. His experience mirrors a broader trend: American developers and firms are increasingly turning to Chinese AI for routine work such as calendar management, document handling, and email processing.
Chinese AI Adoption in the United States
Krikorian is not an isolated case. U.S. companies like cryptocurrency exchange Coinbase have openly stated they are migrating to Chinese models to trim expenses. The appeal lies in the combination of competitive performance and markedly lower pricing. According to the AP story, Kimi K3 saw more than 930,000 downloads in the week after its July release—a 200 % increase from the prior week—with roughly 86,000 of those downloads occurring in the United States, a 387 % jump. The surge was so intense that Moonshot had to temporarily halt new subscriptions as its servers approached capacity.
U.S. Policy Responses and Industry Pushback
The rapid uptake has unsettled some U.S. tech giants and prompted a policy reaction. The Trump administration accused Moonshot of using “covert” but not necessarily illegal methods to build K3 off the back of Anthropic’s Fable, while several U.S. politicians and AI firms, including Anthropic, have alleged illicit “distillation” of their models by Chinese startups—a claim Beijing dismisses as “groundless.” In parallel, U.S. Treasury Secretary Scott Bessent warned that additional sanctions could be forthcoming to protect American intellectual property, reflecting ongoing efforts to restrict China’s access to cutting‑edge AI chips.
Cost and Performance Comparison
A central factor driving adoption is price. AI services are typically priced per million “tokens” (units of data). As technology executive Curt Meinhold of Greensboro, N.C., explained, “If I can pay a handful of cents per million output tokens versus 30 bucks or 40 bucks or 50 bucks, then it’s good enough.” Meinhold, who founded the digital‑legacy platform LilyList, added that “the vast majority of us, 90‑plus percent, don’t need [Anthropic’s] Mythos or Fable.” He emphasized that for tasks such as generating sales leads or conducting research, Chinese models like DeepSeek’s V4 deliver results that are “pretty close on code and research” to those of U.S. frontier models while costing dramatically less.
Model Examples and Their Capabilities
The AP report highlights several Chinese models that have closed the performance gap:
- Z.ai’s GLM‑5.2 (released mid‑June) – used by Krikorian for calendar and email tasks.
- Moonshot’s Kimi K3 (launched early July) – praised for its responsiveness and quickly overwhelmed Moonshot’s subscription capacity.
- DeepSeek’s V4 (previewed in April) – positioned as a challenger to OpenAI’s GPT and Google’s Gemini.
- Alibaba’s Qwen3.8 Max (previewed in July) – another large‑scale model aiming to match frontier capabilities.
Experts quoted in the article note that these models are “nearly as intelligent as the frontier models from OpenAI and Anthropic,” especially when evaluated on standard benchmarks.
The Open‑Source Advantage
A distinguishing feature of many Chinese AI offerings is their open‑source licensing. Unlike the closed‑source models of Anthropic and OpenAI, Chinese startups such as Z.ai and Moonshot make their code publicly available, allowing anyone to examine, modify, and build upon it. Mozilla’s Krikorian observed that “the open frontier is becoming increasingly Chinese‑built.” This openness aligns with a broader industry push: a group of U.S. tech firms including Microsoft, Meta, and Nvidia recently signed an open letter endorsing “open” AI models, signaling recognition that transparency can spur innovation and wider adoption.
Limits and Remaining Gaps
Despite their advances, Chinese models still trail U.S. leaders in overall, full‑range capabilities. Anastasios Angelopoulos, co‑founder and CEO of AI‑evaluation platform Arena, pointed out that while Chinese systems are becoming serious competitors, they “lag American AI leaders across their overall, full‑range capabilities.” Additionally, U.S. firms continue to seek cheaper alternatives to their own models, as noted by Yasir Atalan of the Center for Strategic and International Studies. These limitations suggest that, for highly specialized or cutting‑edge research, U.S. models may still hold an edge.
Global Expansion Ambitions
China’s AI push is not confined to its domestic market. At a flagship technology summit in Shanghai, President Xi Jinping championed open‑source AI models and stressed the importance of promoting global equity, pledging to raise AI capabilities especially in developing nations. Chinese tech giants such as Huawei and Tencent are already embedding AI into consumer devices—smartphones, AI glasses, and humanoid robots—while leading startups are raising additional funding, including through public share offerings, to support overseas expansion. Morningstar analyst Chelsey Tam summarized the dual goal: both China and the United States seek to “encourage widespread adoption of their AI ecosystems, while safeguarding technologies that could materially strengthen strategic rivals.”
Sustainability and Financial Pressures
The rapid growth of Chinese AI startups brings financial challenges. The AP cites Z.ai’s 2023 results: revenue surged 132 % to $107 million, yet net loss widened 60 % to $694 million, illustrating the heavy investment required to compete in the AI arena. Such losses raise questions about the long‑term viability of these companies, especially as they scale globally and face intense domestic competition. Angelopoulos noted that “the Chinese labs are also putting a lot of pressure on one another,” indicating that internal rivalry could further strain resources.
Conclusion
The rise of affordable, high‑performing Chinese AI models is reshaping the global AI market. U.S. developers like Mozilla’s Krikorian and executives such as Curt Meinhold are finding that Chinese systems offer a “good enough” alternative at a fraction of the cost, prompting a shift that even major U.S. firms are beginning to emulate. While policy measures and accusations of illicit distillation aim to curb this trend, the open‑source nature of many Chinese models and their competitive pricing continue to drive adoption. Looking ahead, the balance between cost, performance, openness, and sustainability will determine whether Chinese AI can maintain its momentum and truly challenge the long‑standing U.S. leadership in artificial intelligence.
https://www.latimes.com/world-nation/story/2026-07-26/cheaper-open-intelligent-chinese-ai-models-gain-ground-as-they-make-inroads-in-u-s