Feature
In previous articles, we looked at how leading Western AI companies build defensible moats. Their strategies differ, but all pursue advanced intelligence through a Western free market lens. By contrast, Chinese AI firms now match Anthropic and OpenAI at a fraction of the cost — while competing under CCP-defined market conditions. As OpenAI CEO Sam Altman recently warned, the US may be “underestimating” Chinese AI.
While DeepSeek has dominated Western coverage, it is only one of several major players in the emerging cohort of "Chinese tigers": Moonshot, Z.AI and 01.AI.
China’s AI Paradox: Control, Coordination and Competition
China’s AI industry operates under a paradox: tight political control paired with extraordinary coordination. On the control side, Beijing requires models to register training processes and certify outputs as “true and accurate” (e.g., refusing to discuss Tiananmen Square), with the CCP often fast-tracking favored companies. They must still compete in the market, but only under close oversight.
After passing regulatory hurdles, ecosystem coordination begins. State support is extensive, with direct subsidies, government investment, national compute hubs and expansive, large-scale power plant build-outs. Some experts argue that China may surpass the US because it does not face the same energy constraints.
International challenges are another obstacle. US export controls on Nvidia’s most advanced GPUs have forced Chinese labs to innovate with older hardware. This paradox — rigid oversight combined with fluid coordination — may actually be a not-so-hidden advantage.
- In Western markets, competition often delays collaboration, as labs race to protect proprietary advantages or outbid each other for infrastructure projects.
- In contrast, Chinese AI companies can benefit from state-orchestrated resource sharing, standardized compliance processes and prioritized access to infrastructure.
For AI strategic observers, it raises a critical question: Can a centrally coordinated model outpace decentralized innovation, especially in an era where compute, energy and distribution have become strategic bottlenecks? This question is at the heart of Nvidia's soaring valuation, as it supports Western firms. However, for Chinese firms, without access to high-end Nvidia GPU clusters, efficiency has become the differentiator, most clearly demonstrated by DeepSeek’s R1 release at the start of 2025.
Related Article: No Moat, No Problem? Predictions on Future AI Competition
DeepSeek: Efficiency as Strategy, Openness as Distribution
Founded in Hangzhou out of the hedge fund High-Flyer, DeepSeek applied quantitative trading methods to AI training. It gained international visibility in late 2024 and early 2025 with its V3 model, which delivered GPT-4–class performance at a fraction of the cost of its American rivals.
The contrast was stark: just months earlier, OpenAI had raised $6.6 billion to scale its systems, while DeepSeek claimed it trained V3 for less than $6 million (in reality, it may have been closer to $500M). Since then, DeepSeek has continued to iterate.
In August 2025, it launched V3.1, a 685-billion-parameter hybrid model featuring a 128k-token context window — larger than ChatGPT’s, except for the Pro plan. The model unified chat, reasoning and coding into a single flagship system and was fully open-sourced. Benchmarks showed performance at parity or better compared to models from OpenAI, Anthropic and others.
DeepSeek, however, is just one of many.
Moonshot: The 'Idealistic' AGI Company
Founded in 2023 by Yang Zhilin, Moonshot launched with an ambitious stance centered on AGI, prioritizing AGI over profits. Yang’s academic pedigree — co-author of influential papers on transformers — shaped the company’s early focus on long context windows.
By 2024, the Kimi chatbot had gained notoriety for its ability to handle two-million-token contexts, enabling students and researchers to upload entire textbooks and query them directly. Kimi was briefly a breakout consumer success, but by early 2025, after DeepSeek’s disruptive open-weight release, it had faded from the spotlight.
Key 2025 Developments
- Kimi K2 open-sourced: In July, Moonshot released Kimi K2, a trillion-parameter Mixture-of-Experts model with 128k-token context windows and industry-leading benchmarks in reasoning and coding. K2 introduced a new training method (MuonClip) that lets the team run on 15.5 trillion tokens without instability — a feat not previously demonstrated at this size. The company also released detailed academic papers outlining its process, with further analysis here.
- Developer-first pivot: Kimi began as a consumer chatbot, but Moonshot has since pivoted to a developer-centric platform, similar to OpenAI and Anthropic. It slashed API pricing and expanded hosting across multiple inference providers, such as Hugging Face, marking a shift from a China-first strategy to global provider ambitions.
Even as Moonshot looks to regain momentum, older companies like Z.AI are also staging comebacks with the Deepseek approach.
Related Article: xAI vs Mistral vs Meta: Who Has the Strongest AI Moat?
Z.AI: Established Momentum and State Alignment
Z.AI (formerly Zhipu AI) spun out of Tsinghua University (the MIT of China) in 2019. It developed GLM-130B in 2022, one of China’s first GPT-3-scale bilingual models (i.e., pre-ChatGPT), although it required heavy subsidies to survive. By 2023, Z.AI launched ChatGLM and the Qingyan app ecosystem.
Unlike its more consumer-focused peers, it tilted toward infrastructure. Backed by billions in state and corporate investment (including Alibaba, Tencent and Meituan), Z.AI became widely viewed as Beijing’s policy-aligned AI arm. While it is no longer a startup like the others, OpenAI has named it a critical player in the Chinese ecosystem — potentially increasing its notoriety.
Key 2025 Developments
- GLM-4.5 release: In July, Z.AI launched GLM-4.5, a 3.5T-parameter Mixture-of-Experts model, which is now also open-sourced. Benchmarks placed it among the top three globally, and API pricing undercut both domestic rivals and Western labs.
- IPO and state financing: From January to July, Z.AI raised over $400 million from government funds in multiple cities, doubling its valuation to $ 5.5 billion. IPO filings positioned it to become China’s first listed AI unicorn.
- Geopolitical positioning: Blacklisted by the US for years, Z.AI has leaned into Huawei’s chips for training. OpenAI states that it is a significant competitor for AI implementations to ASEAN and Middle Eastern governments.
Z.AI is less a scrappy startup and more a state-backed implementation engine. Its defensibility comes from subsidies, regulatory alignment and geopolitical deployment.
01.AI: Application Wrappers and Open-Weight Ecosystems
Founded in 2023 by Kai-Fu Lee, 01.AI is well established in the Chinese AI market with the Yi model series, which has also temporarily topped LLM ranking leaderboards. Lee’s prestigious background — ex-Apple, Microsoft and Google — has imbued the company with cross-cultural credibility and strategic reach.
Unlike DeepSeek’s efficiency-first research focus or Moonshot’s AGI ambitions, 01.AI remains focused on usability: building not just models but an open-source app ecosystem.
Key 2025 Developments
- Yi series expansion and Wanzhi copilot: 01.AI updated its Yi model with multilingual coding and tooling, and launched Wanzhi, an open-source productivity assistant for documents, meetings and email. Like Microsoft’s CoPilot, Wanzhi signals ambitions to move beyond the Chinese market.
- Alibaba alliance: 01.AI began co-developing models through a joint venture with Alibaba, potentially extending its global distribution channels. Alibaba will supply compute and customer distribution, while 01.AI will focus on model delivery.
While all of these Chinese startups have global visions, 01.AI may be the most likely to create a product or tool used widely outside the chatbot ecosystem.
Related Article: Where the US Stands in the Global AI Race
Webinar
Sep
25
CMS Briefing: A Live Look at What’s Next in AI-Driven Platforms
Learn how leading organizations are using AI‑driven tools to publish faster, personalize smarter and stay secure.
Register
Webinar
On demand
Ready or Not: How Data-First Organizations Are Unlocking Agentforce Potential
Learn how to cut through the noise, activate Agentforce and build a Salesforce AI strategy that actually delivers.
Watch Now
Webinar
On demand
AI in Customer Service: Faster Resolutions, Happier Customers
Don’t let rising demand burn out your team. See how to build a smarter, more resilient support org.
Watch Now
Webinar
On demand
From Hype to High-Impact CX Strategies That Actually Scale
Turn buzzworthy AI and outsourcing trends into measurable CX wins with fresh 2025 data.
Watch Now
Webinar
On demand
Insights to Action Rethinking the Contact Center for Real Business Impact
Join our exclusive webinar to hear CX executives share their innovative strategies for transforming service delivery.
Watch Now
Webinar
On demand
Accelerating Healthcare Ops with AI
Sign up to see how your peers are putting AI to work — and making it stick.
Watch Now
China’s AI Strategy Is Open, Global and Coordinated
Several trends define the Chinese AI surge in 2025:
- Open-source at the frontier: Following DeepSeek’s lead, each of these firms is now releasing cutting-edge models with open weights — putting pressure on Western rivals.
- Beyond chatbots: Once focused on domestic consumer apps, China’s AI startups are now targeting global, multilingual and developer-first ecosystems.
- Coordination over competition: Despite tight controls and hardware constraints, the CCP’s strategic oversight may be creating advantages by streamlining innovation across the AI implementation stack, where Western firms face fragmentation.
While US companies still dominate in infrastructure, talent and global usage, China’s coordinated, open-source push may pose the most formidable challenge yet to Silicon Valley’s AI hegemony.
