Open Source vs. Closed Source: Who Will Win the US-China AI Battle?

Recently, Chinese open-source large language models have collectively surged forward. From DeepSeek and Kimi K3 to Zhipu GLM and MiniMax, a wave of domestic models has swept the globe with an "open source + free" strategy, directly threatening the commercial foundations of American closed-source companies like OpenAI and Anthropic. On the other side, NVIDIA CEO Jensen Huang has publicly championed open source, Elon Musk retweeted in support, and Microsoft's CEO has also cast a vote in favor. This debate over AI technology roadmaps is becoming a new focal point of US-China tech competition.

This article is based on discussions from the Taiwanese political commentary show Headline Talk, analyzing the business logic and strategic dynamics behind the open-source and closed-source approaches.

Open Source vs. Closed Source: A Battle of Interests

Before diving deeper, it's worth clarifying the basic concepts of these two approaches. Open-source models publicly release model weights, architecture, and even training code, allowing anyone to download, modify, and deploy them. Closed-source models only provide services through API interfaces—users cannot access the underlying weights or code. The fundamental difference lies in control: open source hands control to the community and users, while closed source retains it with the developing company. This divide directly determines business models—closed-source vendors charge per API call based on usage, while open-source vendors profit indirectly through ecosystem services, cloud computing resource sales, and other means.

In this debate, a noteworthy phenomenon is that even within the American tech community, clear divisions have emerged. OpenAI and Anthropic are reportedly in closed-door discussions with US regulators, hoping to tighten oversight of Chinese open-source models, arguing that "once model capabilities approach or exceed human levels, full open-sourcing increases the risk of misuse."

However, Jensen Huang offered a starkly different perspective. He emphasized on social media: "AI will reshape every industry, power every company, and be built by every nation. Open-source models strengthen security and cybersecurity, accelerate the diffusion of innovation, and enable sovereignty." His conclusion: "The world needs both frontier closed models and frontier open-source models."

Show guests pointed out incisively that these two camps represent different interest calculations. For OpenAI and Anthropic, the greatest pressure is monetization and profitability—closed source means "if you use my tools, you pay." For NVIDIA, whether open or closed source, everyone needs to buy GPUs and computing power, so Huang naturally "profits from both sides" and welcomes everyone in the world using AI.

The Market Shock Triggered by Kimi K3

What truly alarmed the American AI community was Kimi K3, released by Moonshot AI. According to the show's analysis, Kimi K3 boasts 2.8 trillion parameters, making it the world's largest open-weight model, and developers can download it for free.

More critically, Kimi K3's challenge is no longer in chatbots but in coding capabilities—precisely Anthropic's core strength. Guests noted that Kimi K3's coding ability can now rival Claude's top-tier models, directly shaking the established belief that "you must use American closed-source models for programming."

After its debut, Kimi K3's benchmark performance boosted Moonshot AI's daily sales by at least 6x, while its competitors' AI stocks triggered a sell-off. The show mentioned that the Philadelphia Semiconductor Index has been on a roller-coaster ride recently, with Kimi K3 being one of the key drivers.

Regarding external claims that Kimi K3 "distilled" other models, guests explicitly rebutted this. Here's an explanation of knowledge distillation: Knowledge Distillation is a model compression technique proposed by Geoffrey Hinton and others in 2015. Its basic principle uses an already-trained large model (called the "teacher model") to guide the training of a smaller model (called the "student model"). The teacher model's output probability distribution contains rich "dark knowledge"—it not only tells the student what the correct answer is but also conveys the relative relationships between incorrect answers. By mimicking the teacher's output distribution rather than merely learning labels, the student model can achieve performance close to the teacher's with a much smaller footprint. Because distillation's logic is "big teaches small," Kimi K3 itself has 2.8 trillion parameters—virtually the largest model globally—and there's simply no larger model that could serve as its "teacher." This is the core argument guests used to refute the distillation allegations.

This technical breakthrough stems from its proprietary Latent Mixture of Experts (Latent MOE) algorithm. Mixture of Experts (MOE) is a sparsely-activated neural network architecture whose core idea is decomposing a model into multiple specialized "expert" sub-networks, activating only a small subset of experts for each inference to handle specific inputs. The benefit: the model's total parameter count can be enormous (like Kimi K3's 2.8 trillion), but the parameters actually involved in computation per inference are far fewer, maintaining powerful capabilities while controlling computational costs. Kimi K3's "Latent MOE" is a further evolution that introduces latent space representations in the expert routing mechanism, making expert selection and combination more flexible and efficient—considered the key technical innovation enabling the model to achieve top-tier performance at relatively low inference costs.

The Strategic Value of Open Source: Sovereign AI and Ecosystem Lock-in

China's open-source free strategy is far from simple altruism. Guests analyzed that this approach dramatically lowers the barrier to AI entry, allowing any country or company—even those without the capability to develop their own models or with limited budgets—to modify and deploy based on open-source models.

This aligns perfectly with Jensen Huang's repeatedly emphasized concept of "Sovereign AI": every country can have its own large model. Sovereign AI is a concept that has rapidly gained traction in international technology policy circles over the past two years. Its core proposition is that every nation should possess the capability to build AI infrastructure using its own data, language, and culture, rather than depending entirely on AI services provided by foreign companies. The background is that AI models are becoming infrastructure akin to electricity and the internet—if a country's AI capabilities rely entirely on overseas closed-source models, it faces not only data sovereignty risks (sensitive data must be uploaded to foreign servers) but also potential "supply cuts" during geopolitical conflicts. Open-source models provide a low-cost path to sovereign AI—countries can download open-source models, deploy and fine-tune them on local infrastructure, achieving autonomous control over AI capabilities.

Data shows that India has the highest download volume for DeepSeek, for a simple reason—open source means no need to train a large model from scratch. This also explains why the Middle East, Southeast Asia, and other regions have shown strong interest in Chinese open-source models.

From a corporate cost perspective, the differences between the two models are stark:

• Closed-source models: Charged per input/output token, costs determined by the vendor
• Open-source models: Users can freely choose computing sources (e.g., Alibaba Cloud, Baidu Cloud), with far greater cost controllability

Here's an explanation of token-based billing. A token is the basic unit through which large language models process text—one token corresponds roughly to one English word or one to two Chinese characters. Closed-source model APIs typically charge separately for input and output tokens. For example, GPT-4o is priced at $2.5 per million input tokens and $10 per million output tokens. This means every API call incurs costs, and as usage grows, costs increase linearly or even super-linearly. In contrast, open-source models allow enterprises to deploy on their own servers or third-party cloud platforms, converting per-call fees into fixed computing costs—offering extremely significant cost advantages for high-frequency usage scenarios.

This also explains why the top-ranked large models by global usage are almost all products from mainland China.

The deeper strategic significance lies in ecosystem lock-in: once enterprises adopt open-source models like DeepSeek or Kimi K3, they are to some degree drawn into China's computing ecosystem. This has also driven surging prices for domestic chips and HBM (High Bandwidth Memory).

Chip Ecosystem: The CUDA Moat Is Being Filled In

The rise of open-source models has triggered chain reactions. According to the show's analysis, starting from DeepSeek V4, eight Chinese domestic chip companies can all connect to these models, no longer dependent on NVIDIA's CUDA ecosystem.

CUDA (Compute Unified Device Architecture) is a parallel computing platform and programming model launched by NVIDIA in 2006, allowing developers to leverage GPUs' massive parallel processing capabilities to accelerate general-purpose computing tasks. Over the past decade-plus, virtually all mainstream deep learning frameworks (PyTorch, TensorFlow, etc.) have been deeply bound to CUDA, and hundreds of millions of lines of AI developers' code worldwide are built on CUDA, creating extremely high migration costs and ecosystem lock-in—this is the so-called "CUDA moat."

Guests pointed out that the CUDA moat is essentially a computing paradigm that uses graphics card architecture for AI computation. But today, more and more AI chip companies are directly building AI-specific chips (ASICs), bypassing the graphics processing layer entirely. Meanwhile, the maturation of open-source compilation frameworks (such as Triton and MLIR) is also reducing CUDA dependency, with more AI workloads able to run directly on non-NVIDIA hardware bypassing CUDA. The CUDA moat's effectiveness has thus been significantly weakened. The recent surge in China's chip industry is fundamentally driven by massive expansion at the application layer.

On the other side, memory chip prices, especially HBM (High Bandwidth Memory), are skyrocketing. HBM is an advanced packaged memory that vertically stacks multiple DRAM chip layers using Through-Silicon Via (TSV) technology. Compared to traditional GDDR memory, HBM offers bandwidth several to ten times higher (the latest HBM3E exceeds 1.2TB/s), with lower power consumption and smaller footprint. For large language model inference, bandwidth is often more critical than raw compute—model parameters need to be rapidly read from memory to processing units, and HBM's ultra-high bandwidth directly determines inference speed and efficiency. The global HBM market is currently highly concentrated among SK Hynix, Samsung, and Micron, with SK Hynix holding approximately 50% market share.

Guests used the analogy that "computing centers are like power plants": when the whole world is building computing centers, memory becomes a scarce resource in short supply. As the global AI computing center construction boom progresses, HBM has become one of the most critically scarce semiconductor products, with prices continuously surging through 2024-2025. This also explains why, when Apple reportedly considered using Chinese memory chips, Micron would be deeply concerned—because it could mean losing a crucial major order.

Who Will Win: Open Source or Closed Source?

On the question of "which will ultimately prevail, open source or closed source," show guests generally leaned toward open source. The reasoning: the closed-source model is extremely capital-intensive, and US fiscal pressures are enormous. Meanwhile, if the US and China each block the other's AI, it amounts to "cutting off their own path"—harming both innovation and markets.

Guests also mentioned that recent internal system attack issues at OpenAI have exposed AI security concerns, a phenomenon not yet observed in Chinese open-source models. Of course, this represents a single-source perspective that requires more information to verify.

You might not have noticed, but the essence of this battle is a deep contest between business models and technology ecosystems. Open source spreads rapidly through low barriers and low costs, capturing global markets and developer mindshare; closed source maintains high premiums through technological leadership. In the short term, both will likely coexist for an extended period, as Huang put it, "both are needed." But the trend suggests that open source is rewriting the global AI landscape at a faster pace.

For industry practitioners following AI, this isn't merely a choice between technology roadmaps—it's a critical entry point for understanding the future trajectory of global AI competition.

Key Takeaways