China is breathing down OpenAI’s neck. Alibaba-backed Moonshot AI has launched a new model, Kimi K2 Thinking, which outperforms OpenAI’s GPT-5 and Anthropic’s Claude Sonnet 4.5 on several key benchmarks.
Moonshot said the model’s architecture activates 32 billion parameters per inference out of a total of one trillion, and supports context windows of up to 2,56,000 tokens. It can also perform 200 to 300 sequential tool calls without human input.
“The AI frontier is open-source!” wrote Hugging Face CEO Clement Delangue after the launch of Kimi K2 Thinking.
At the same time, NVIDIA CEO Jensen Huang warned that China could soon lead the global AI race. He said that China’s lower energy costs, lighter regulations, and government subsidies make building and running AI infrastructure far cheaper than in the West.
However, he later softened his stance, clarifying on social media that “China is nanoseconds behind America in AI,” adding that it is vital for the US to maintain and extend its lead globally by attracting more developers to American technology platforms.
As researcher Yuchen Jin aptly put it, “If you ever wonder how Chinese frontier models like Kimi, DeepSeek, and Qwen are trained on far fewer (and nerfed) Nvidia GPUs than US models — remember that NASA landed people on the moon in 1969 with just 4KB of RAM.”
It’s a reminder, Jin says, that creativity often thrives under constraints.
The estimated cost of developing the Kimi K2 Thinking AI model was around $4.6 million, according to reports from outlets such as CNBC. That price tag is strikingly modest when compared with the billions of dollars usually invested in training top-tier AI systems in the United States.
Side effect of blocking Chinese firms from buying the best NVIDIA cards: top models are now explicitly being trained to work well on older/cheaper GPUs.
The new SoTA model from @Kimi_Moonshot uses plain old BF16 ops (after dequant from INT4); no need for expensive FP4 support. https://t.co/KI1quF6iiQ pic.twitter.com/910XgbK3rR— Jeremy Howard (@jeremyphoward) November 9, 2025
That efficiency is now translating into global influence. Alibaba Group’s Qwen AI models are finding eager adopters among major Western firms like Airbnb — a sign of the growing international appeal of China’s open-source AI ecosystem.
According to a Bloomberg report, Airbnb co-founder and CEO Brian Chesky said the company relies heavily on Alibaba’s Qwen models to power its AI-driven customer support agent.
Chesky, a longtime friend of OpenAI’s Sam Altman, added that ChatGPT’s integration tools weren’t “quite ready” for Airbnb’s needs, while Qwen proved “very good” as well as “fast and cheap.”
NVIDIA’s Struggles in China
Most recently, the US government has moved to bar NVIDIA from selling its newest scaled-down AI processor, the B30A, to Chinese customers. According to reports, the White House has notified federal agencies that exporting this chip to China violates existing trade rules, as Washington continues to tighten its technology restrictions.
Meanwhile, China has introduced new subsidies that halve energy bills for primary data centres using domestic chips, aiming to strengthen its semiconductor industry and reduce reliance on US technology, the Financial Times reported, citing sources.
Local governments in provinces such as Gansu, Guizhou and Inner Mongolia are offering discounts of up to 50% on electricity for facilities that adopt chips made by Chinese firms, including Huawei and Cambricon Technologies.
Also, the top Chinese tech firms, including Alibaba and ByteDance, are ramping up spending on AI and cloud infrastructure. Alibaba has pledged more than $50 billion over the next three years to boost its cloud and AI capabilities, while ByteDance plans to invest about $20 billion in GPUs and data centres to power AI development.
Just the Beginning
Deedy Das of Menlo Ventures told AIM that China’s Kimi K2 Thinking model shows that open-source systems can now rival, and in some cases even outperform, top-tier proprietary models. “They’re definitely bringing the heat to American labs that rely on far more resources — better chips, larger compute budgets, and higher R&D costs,” he noted.
Das added that while this doesn’t mean revenue will immediately shift toward Chinese models, given that US labs have strong user bases and many enterprises are reluctant to adopt Chinese systems, their lower pricing and open approach will help them steadily gain market share.
But not everyone agrees that openness alone gives China an edge. Anthropic CEO Dario Amodei said the idea of open source in AI is often misunderstood. “I don’t think open source works the same way in AI that it has worked in other areas,” he said, explaining that while traditional open source allows anyone to inspect and build on source code, AI models only make their weights public, not their inner workings.
Amodei called the open-source debate a “red herring,” arguing that what really matters is model quality, not openness. “When I see a new model come out, I don’t care whether it’s open source or not. I ask, is it good? Is it better than us at the things that matter?”
OpenAI is Self Sufficient
While the Chinese government continues to support its tech companies, OpenAI CEO Sam Altman has clarified that the company is not seeking any government guarantees for its data centres. His statement comes after recent comments from OpenAI’s chief financial officer, Sarah Friar, sparked speculation about possible state backing.
“We do not have or want government guarantees for OpenAI data centres,” Altman said. “Governments should not pick winners or losers, and taxpayers should not bail out companies that make bad business decisions.”
Altman also revealed that the company expects to surpass an annualised revenue run rate of $20 billion by the end of 2025 and is planning infrastructure commitments of around $1.4 trillion over the next eight years, as it scales up computing power to meet growing demand for AI systems.
Responding to concerns about whether OpenAI could become “too big to fail”, Altman said, “If we screw up and can’t fix it, we should fail, and other companies will continue on doing good work. That’s how capitalism works.”
Meanwhile, David Sacks, in a post on X, said, “There will be no federal bailout for AI. The US has at least five major frontier model companies. If one fails, others will take its place.”
Notably, Microsoft is now also forming its own AI team called the MAI Superintelligence Team, led by Mustafa Suleyman, the head of Microsoft AI and co-founder of DeepMind.
Sacks said the focus should be on making power generation and permitting easier to enable faster infrastructure growth, without raising electricity costs for consumers. He also clarified that the discussion around government support for AI firms had been misunderstood. “I don’t think anyone was actually asking for a bailout — that would be ridiculous,” he wrote on X.
The AI race won’t end with a single winner, but it will showcase who’s been running the smarter marathon. OpenAI may lead in compute, yet China’s mastery of constraint suggests the next frontier of intelligence might not be born in abundance, but in adversity.
The post Capitalism Built OpenAI, Constraints Helped China Defeat It appeared first on Analytics India Magazine.
