In early 2025, a quiet evolution emerged out of Hangzhou—a city of thirteen million famed for its historic West Lake, misty mountains, and poets who depicted it as "Heaven on Earth." Yet it was here that a Chinese company—under-resourced by Silicon Valley standards—released a reasoning-focused large language model in January 2025 a few steps behind that era's ChatGPT 4-class models. DeepSeek had been trained, its parent company claimed, for only $6 million—pocket change compared to OpenAI or Google expenditures.

For a global industry convinced that US large language models (LLMs) were unassailable—built by companies with oceans of compute, elite talent, large capital infusions, and company valuations the size of small-nation GDP—DeepSeek upended assumptions. The Chinese model emerged at a time that China's AI sector seemed beleaguered by chip bans and a slowing economy and against a backdrop in which private sector investment into the US sector was twelve times China's and twenty-four times the UK's. Analysts found that DeepSeek's parent, High-Flyer, hadn't spent its way into contention; instead, it had been crafty: focusing on targeted domains, lean training, and aggressive energy efficiency. While critics whispered about secret Nvidia chips, covert ChatGPT usage, or hidden costs, the signal was clear: US companies wouldn't hold a monopoly on advanced LLMs.

This "DeepSeek moment" ignited a debate: "Are Chinese companies catching up in AI?" That framing misses the point. The AI stack is broad—spanning energy, infrastructure, chips, foundational research, and application deployment—and across that landscape the US and China possess different strengths. The real picture is that the world's two dominant AI markets aren't competing in the same race; they're running parallel ones shaped by different economic constraints, cost structures, and market demands—and that divergence will shape global AI over the next decade.

"If you look at the future of deep tech, it's clear that the US and China are like the two sides of [the traditional Chinese martial arts practice] tai chi, with its black-and-white yin and yang symbol—each with unique strengths, each pushing the other forward," says Soul Capital's Herry Han, "It's not just competition. It's a dynamic balance."

In the US, AI development follows a capital-intensive path: massive data centers, soaring talent costs, and billion-dollar frontier science bets. Well-funded US leaders will keep pushing toward artificial general intelligence (AGI) and increasingly sophisticated agentic systems designed to pursue goals through multi-step planning and self-directed action.

Chinese companies face a different competitive reality: relatively tighter capital, more limited access to high-end computing power, and a smaller domestic profit pool. Its companies respond by leaning hard into open-source, cost efficiency, fast application layer innovation, and global markets for monetization.

China's AI story isn't about catching up or winning the frontier model race but rather by industrializing the use of AI. Chinese AI firms are building an alternative ecosystem that's leaner by necessity, more open by design, and working to wire into the economy. This is prompting its adoption across start-ups, universities, and mid-market firms worldwide. As global industry leaders continue a preference for keeping their most capable models "closed-source"—which keeps source code proprietary or secret—Chinese models are diffusing through easy access and utility, quietly shaping global AI practices and creating a new, technical form of soft power.

The Chinese spatial design firm Manycore Tech offers a close-up view of how China's AI dynamics play out inside a single company.

When Hang Chen and two friends at the University of Illinois Urbana-Champaign first started tinkering in the early 2010s with graphics processing units (GPUs) that excel at parallel computing, they explored gaming, film, and other applications before broadening out. "Almost no one realized they could be used for super-computing," Hang tells us. That hunch became the foundation for Manycore, a company built around rethinking how space is computed.

After returning to Hangzhou, the three engineers pursued the idea of moving multidimensional rendering entirely to the cloud. Instead of installing heavy software on powerful local machines, users could simply open a browser, sketch a layout, and watch it materialize in 3D. That vision became Kujiale in China and Coohom abroad, cheap to adopt and scalable, and now among the world's most widely used spatial design platforms.

Manycore's timing has been fortuitous. As China's home decoration and furniture industries went digital, its platform became their operating layer. Each object—every sofa, tile, or window frame—could be converted into production parameters sent directly to factories, enabling a whole-house customization boom that spawned companies big enough to go public.

E-commerce soon followed. Merchants needed product photos and videos: furniture staged in a Paris apartment, a kitchen gleaming under American Christmas lights. Manycore's rendering engine could generate thousands of virtual rooms and studios for merchants selling on Amazon, Shopify, and Shopee—a quiet global diffusion that emerges from practical AI tools.

Manycore's real advantage isn't software—it's data; after a decade of rendering, the company has accumulated a dataset of hundreds of millions of fully-labeled "3D scenes," or interiors with geometry, textures, lighting, and materials. These structured datasets help convert text or sketches into fully-realized 3D layouts, scenes, and videos, a foundation not just for design but also for research for robotics and AI agent developers. Tasks once requiring costly software can now be done by anyone.

The company relies on standard computing GPUs, which are sufficient for rendering, Hang says. And pragmatism, adaptation, and technical improvisation have kept the company scaling up even as competitors emerged. "Entrepreneurship is a 'nine-out-of-ten-ways-to-die process,'" he says. "If you're not stopped by one factor, you'll be stopped by another. Entrepreneurs operate in a highly dynamic state, constantly searching for a way."

The overseas business remains small, but Hang sees global expansion as essential for scale and monetization. Korea is now one of Manycore's fastest-growing markets. "It is easy for us given how GenAI (generative artificial intelligence) can help with localization," says Hang.

To understand how Manycore fits into China's broader AI landscape, we need to zoom out.

If you ask the Chinese when the country's "AI awakening" began, many will chuckle. AI? They've been living inside a digital and big-data society for years. Ping An Group digitized insurance claims and call centers a decade ago. McDonald's uses data analytics to run customer relationship programs. Alibaba and ByteDance rebuilt shopping and communication with recommendation engines, predictive logistics, and automated payments. Chinese firms learned early that surviving meant mastering data; by the time GenAI arrived, it didn't feel revolutionary—just the next logical step.

China's AI ecosystem has been expanding—pragmatic, open, and competitive—"letting a hundred flowers bloom," as the old metaphor goes. This was deliberate. Nearly a decade ago, Chinese leaders declared an ambition to lead the world in AI by 2030 and created scaffolding for research and commercialization; by 2021, the country was producing about one-third of global AI papers and drawing a fifth of global AI start-up investment.

When ChatGPT appeared in late 2022, Chinese entrepreneurs reacted instantly. As is often the case, the initial breakthrough came from Silicon Valley. In China, hundreds of teams piled into the space at once. Every major tech company—Baidu, Tencent, Alibaba, ByteDance—launched LLM programs, joined by an explosion of start-ups.

If this sounds chaotic, that's because it was. Entrepreneurs defaulted to the traditional digital playbook: copying, rapid iteration, free trials, and breakthrough or collapse. Out of the pileup, a few leaders emerged: Z.ai spun out of Tsinghua University labs; MiniMax and Moonshot AI raised hundreds of millions of RMB in their first year; Baichuan rivaled global models by mid-2024; iFlyTek's SparkDesk spread through classrooms and call centers.

By mid-2025, Chinese firms had released more than 1,500 large models—though most never gained traction—and built a full AI stack spanning chips, data centers, models, applications, and deployment. While a handful of players remain locked in a sprint at the model level, most have shifted into applications, building businesses on top of the models, from agents for design, commerce, logistics, medical imaging, and education to a wide range of applied systems in between. The ecosystem now includes over five thousand AI firms, seventy-one unicorns, and more than three hundred listed companies generating roughly 70 percent of China's AI revenue. The intensity of experimentation has been unforgiving; several well-funded AI start-ups that launched competitive models in 2023–2024 failed to convert usage into revenue and shut down within eighteen months, their teams quietly absorbed into larger platforms or overseas labs.

Government support continues to accelerate in this strategic sector; several ministries have created plans to facilitate AI adoption and have even come up with key performance indicators (KPIs) to track progress. "AI is one of the ways the Chinese government intends to increase economic output—by making existing labor and capital more efficient," notes our former colleague Gordon Orr. The result is a nationwide race to build AI infrastructure, talent, and commercialization pathways.

US GPU export controls from 2022 introduced a new constraint, but instead of stalling progress, they forced further efficiencies and also compelled homegrown chip players such as Huawei to announce a road map to develop advanced chips. The government launched the third phase of its semiconductor fund in 2024—344 billion RMB (about $50 billion), the largest ever—to accelerate domestic chip capabilities.

The Chinese public is among the world's most optimistic about AI. Stanford University's 2025 AI Index Report found that 83 percent of Chinese respondents believe AI will improve their lives—more than double the share in the US or the Netherlands. That's helped bend adoption curves upward: platform companies like Alibaba, JD.com, and Pinduoduo have embedded AI end to end, while 80 percent of sellers on JD.com use AI-generated content tools. McDonald's China built "RGM Boss," a platform that codifies operations for two hundred thousand employees. It also offers a training agent: "Our restaurant managers can actually practice with an AI customer, so they feel more confident the next time they face a demanding guest," says Phyllis Cheung, McDonald's China CEO. "When you have the right infrastructure, a lot of new possibilities open up." According to the latest national policy plans, most state-owned enterprises have incorporated AI as an integral part of their upcoming strategy.

Chinese entrepreneurs are poised to lead in business-to-consumer (B2C) innovation, says GSR Ventures founder Allen Zhu, as competitive advantage shifts toward "AI+ scenarios"—such as gaming, education, and personalized content. "These are areas where Chinese teams excel," he says. He expects a major boom in AI applications with standout products emerging from platforms like TikTok alternative Kuaishou and RedNote.

According to a McKinsey survey of more than a thousand global companies, GenAI adoption across at least one business function in mainland China reached 84 percent in 2025, slightly ahead of the global average.

We'd like to highlight a few key dynamics that define China's AI sector.

An equivalent dollar in China goes farther—a fact that reframes the AI investment picture. While the US dominates total AI funding by a wide margin, Chinese firms often produce more output per unit of capital.

This efficiency begins with input costs. AI engineers earn about 400,000 RMB (about $57,000) per year, far below US salary norms. China also trains one-and-a-half to two times as many AI-relevant PhDs as the US, and many trained researchers are returning home, creating a large, affordable talent pipeline. Data centers benefit from cheaper electricity, discounted land, and aggressive local subsidies—reinforced by national policies that treat large-scale computing as strategic infrastructure. In some provinces, electricity costs are halved for facilities using chips.

These advantages are further sharpened by the constraints that characterize China's AI ecosystem. Foreign direct investment into China has fallen more than two-thirds since 2019, access to Western markets has tightened, high-end domestic GPUs lag behind imported ones, and regional grids are straining under data center loads. Moreover, China's business-to-business (B2B) market is characterized by a low willingness to pay for services; firms routinely build software tools in-house rather than buy them. The result is a domestic software market roughly one-quarter the size of the US's $237 billion industry, making it difficult for many AI start-ups to reach scale or charge premium prices. In addition, large corporation culture—especially state-owned enterprises—prizes headcount and incremental change, resists automation, and blunts AI's potential impact.

Angel investor Jun Xu captures a core monetization constraint: AI's total addressable market—the revenues available for a company's product or service—tracks the cost of white-collar labor. "And that pool is simply much larger in the US and other developed markets than in China because salaries are much higher," Jun says. "China's AI problem isn't chips or models or supply—it's demand. Demand is cheaper and smaller." These pressures have pushed Chinese firms toward efficiency and ingenuity.

Nvidia's Jensen Huang said in 2025 that restricting US chip sales to China would only accelerate China's domestic push. "Local companies are very, very talented and very determined," he said, "and the export control gave them the spirit, the energy, and the government support to accelerate their development." Altogether, the ecosystem has created a set of Chinese open-source LLMs operating along the "efficient frontier," delivering leaner architectures and stronger reasoning at modest compute levels. By late 2025, DeepSeek's parent announced that one million units of output could be had for about 3 RMB—that's about fifty cents and a twentieth the cost of ChatGPT at the time.

This combination of high efficiency, thin margins, and a smaller domestic monetization pool shapes how Chinese AI companies scale. Many are becoming what we call "skinny athletes"—lean, fast, and relentlessly efficient—which not only affects how they compete at home but also how and where they grow. For many, that increasingly means serving customers abroad, and the use of generative AI means that language, localization, and customer support are no longer decisive obstacles to global expansion.

"All AI start-ups have an international strategy from day one in China, and Southeast Asia is top of the list," says Cindy Chow, CEO of the Alibaba Hong Kong Entrepreneurs Fund, which recently launched a new fund dedicated to AI application start-ups. "In fact, many Southeast Asian conglomerates and financial institutions are keen to invest with us, as they believe that the region won't catch up in AI development on their own, and that accessing advancements from China is key to staying competitive."

The Chinese AI unicorn 01.AI shows what global expansion looks like when delivering value is the organizing principle. Founded by Kai-Fu Lee, an ex-Google China chief and president of leading tech VC Sinovation Ventures, the company started as a pure LLM developer in 2023 but quickly shifted toward enterprise AI agents, or what they call "super-employees" designed for functions such as insurance brokering, procurement, and logistics optimization. "Monetization challenges force AI to accelerate faster," says Ning Ning, 01.AI's vice president of international business and AI consulting. "The future of enterprise AI is not about selling technology but rather making AI accountable for business outcomes."

In practice, delivering value means embedding "24/7 digital specialists" directly into enterprise operations. In one deployment with a Perth, Australia-based mining company, 01.AI's engineers positioned its AI agents as "teammates" alongside employees: a logistics scheduler to optimize rail and port traffic; a procurement agent that reads vendor emails and generates purchase orders; an operations planner to juggle trucks and crews. Each agent retrains itself as conditions change, says Ning, continuously improving its performance.

China's open-source culture has become an accelerant. Hundreds of teams now openly release their model architectures and weights—the blueprints and parameters that a model acquires through training—creating a shared infrastructure anyone can examine, build upon, and improve.

Chinese AI firms are pragmatic, says Chloe Fang, a founder in the text-to-image/video space. They aim for AI that delivers value and often use open-source "as a global hook to attract global users," she says. "They start building brand equity and word of mouth—and then release better closed-source models later on." Chinese models now lead in several key generative media categories, accounting for five of the top ten image-to-video models globally and three of the top ten text-to-video and image-editing models.

This open-source approach aligns with national priorities emphasizing "open ecosystems," deep integration with all industries and sectors of the economy and society, and "increased global cooperation." Chinese open-source LLMs—led by Qwen, MiniMax, and DeepSeek—now account for one-third of global LLM usage, up from virtually nothing in late 2024.

Start-ups and companies from Silicon Valley to Africa and Southeast Asia now run Chinese models because they're accessible, transparent, and far cheaper to run than many US alternatives. "You can remove anything you want, add anything you need. That flexibility and openness actually gains trust," says Sinovation Ventures founder Kai-Fu Lee, an ex-Google China chief and an influential AI voice. He points out many US institutions, students, and researchers are using Chinese models "not because they're Chinese but because they're open."

Jian Wang, founder of Alibaba Cloud, argues that this moment in AI echoes the late 1990s: Netscape made its browser free and its code publicly available—an open-source "watershed" that helped catalyze the commercial internet. Today's AI parallel is similar, he says: foundational tools are now open and collectively improvable, so the constraint is no longer whether source code itself is accessible. Rather, the industry is shifting toward what Jian calls "'open resources,' especially model weights, data, and computing resources, which are indispensable to advancing this industry." As more of these resources become available, the more developers can skip over the massive costs of reproducing work already done by others—a key variable in accelerating the spread of AI, he says.

Open-source cuts both ways: it not only quickens diffusion, but also brings on key features of involution: too many players chasing little to no margins.

We've seen multiple generations piling into China's AI race at once: industrialists in their sixties retrofitting legacy businesses and funding new start-ups; first-generation internet giants such as Alibaba and Tencent pivoting; second-generation digital natives such as ByteDance deploying AI across content and commerce; and a wave of AI-native start-ups alongside consumer electronics and even EV makers now calling themselves AI companies.

That deep bench matters. In the US, AI salaries remain sky high, but as we often say, California may attract the ecosystem's valedictorian and salutatorian while China might grab most of numbers three through one hundred. China's frenzied ecosystem can attempt many more low-cost prototypes, model variants, and application-driven experimentation.

Across China, large numbers of teams experiment simultaneously, open-sourcing models and chasing down practical applications. China's data-rich industries and dense digital infrastructure allow AI agents to plug into WeChat mini-apps, enterprise systems, and factory IoT networks with little friction. As one Shenzhen entrepreneur puts it, "We don't call them agents. We call them workers who don't need to eat or sleep."

AI literacy among young entrepreneurs is ubiquitous, fueling a competition that's ultimately Chinese versus Chinese. "Four generations of Chinese entrepreneurs are now thinking about AI, ten times as many engineers as in the West, working twice the hours—you can imagine the future," says Yibing Wu, China CEO of investment company Temasek.

We expect China's AI trajectory will unfold similarly to earlier digital revolutions in e-commerce, fintech, and mobile internet: application first, fiercely competitive with periods of involution, and propelled by rapid iteration rather than orderly planning.

Finally, while global commentary often frames US-China AI as a Sputnik-style race, this isn't a winner-take-all contest. The American approach remains the global benchmark for frontier model ambition, built on hyper-scale cloud computing, advanced semiconductors, enterprise software depth, and foundation models built on billion-dollar training runs. Chinese companies aren't trying to replicate that playbook; they couldn't afford to. Instead, political friction and firewalls are driving them to optimize for a different endgame, resulting in two parallel systems evolving side by side rather than together.

From inside the Chinese ecosystem, the dynamic feels even less like a geopolitical race. "Chinese companies aren't trying to beat the US—they're trying to outcompete each other and survive in a global market that's skeptical of them," says Hang of Manycore. "We didn't even call ourselves an AI company when we started—it wasn't in fashion yet."

If people like a space metaphor, a better one is two programs exploring different planets. They share some tools, occasionally glance sideways to check pacing and possibility, but operate in atmospheres shaped by different economics, regulatory pressures, markets, and industrial bases. The "competition" will feel fierce especially in model uptake, but their trajectories aren't directly comparable. At the same time, they're interdependent: "Each side pushes the other forward faster than it could move alone—unidentical twins with different strengths and headed for complementary futures," says Jason Chiu, co-founder of the spatial intelligence start-up Collectiv.

Despite geopolitical tensions, the US and Chinese AI ecosystem remains deeply intertwined. Nearly half of the world's AI researchers earned their undergraduate degrees in China, and many built their careers in US labs and companies before returning or founding new ventures. The lineage is braided: Kai-Fu Lee spent his formative years at Apple, Microsoft, and Google; Baidu founder Robin Li worked at Infoseek; ByteDance's Yiming Zhang had a stint at Microsoft. As Tsinghua University professor Yi Wu told a thousand-plus attendees and millions more watching online at a Shanghai AI conference, "I was mainly influenced by two places: Berkeley, where I did my PhD, and OpenAI… when the team was only a few dozen people."

That shared intellectual DNA shows up in daily practice. We've met founders in Shenzhen training models with open-source code from California. We know of start-ups in Boston fine-tuning on Chinese models. Every week, foreign groups arrive in Shenzhen and Hangzhou for immersion tours. Ideas, code, and talent continue to cross borders. In late 2025, Meta underscored that reality by announcing plans to acquire the Chinese-founded start-up Manus for more than $2 billion as part of its push into advanced AI agents. The Manus deal distilled the entire cross-border AI reality into a single transaction: Chinese entrepreneurs building the application layer of AI; funding from both Chinese and US venture capital; international monetization abroad; re-incorporation in Singapore to tap global customers; and an eventual exit to a US tech giant. And geopolitics shaped the transaction at every stage: the US Treasury reviewed the venture capital firm Benchmark's role, while Chinese regulators questioned Manus's overseas move and the loss of talent and technology to a foreign company. Meanwhile, some customers balked at Meta's data-privacy record. Though the deal closed in January 2026, uncertainty hangs overhead as regulatory bodies examine whether the acquisition violates export or national security rules.

Geopolitics draws borders more easily around companies than around ideas. For years, language was a barrier for Chinese companies expanding into overseas markets. That dynamic may be quietly changing. Joe Tsai, chairman of Alibaba, recounts seeing a social media post from a Meta employee that much of the informal idea exchange on his AI team now happens in Chinese—a language that employee didn't understand. "This is the first time… knowing Chinese has become an advantage in the AI world—that's very, very interesting," says Joe.

Channeled well, this intellectual cross-pollination could create a more resilient global ecosystem. As Nobel laureate Richard Sutton posited in Shanghai, centralized control is rooted in fear—an "us versus them" mindset. "Regarding the political aspects of AI, my conclusion is: the development of AI and the progress of human society can and should stem from decentralized cooperation."

It's an exciting dynamic, says Soul Capital's Herry, who argues that China's participation in the ecosystem "opens the door wider." "Without AI we're in kindergarten, and with AI we gain doctorate-level understanding. Previously only the US could provide it as a premium service, reaching a small percentage of humanity, and now the door's open wider with both providing this new electricity of AI."

We should expect more globally relevant AI companies to emerge from China, offering the global ecosystem a genuine alternative: more open-source models, lower training and inference costs, and an increasingly self-sufficient tech stack from hardware to tooling to LLMs.

The economic upside is substantial. McKinsey estimates AI could add $600 billion to China's GDP by 2030, while Goldman Sachs projects generative AI will lift annual growth by 0.2 to 0.3 percentage points by 2030—two to three times earlier estimates—driven by transport, autos, logistics, manufacturing, enterprise software, and healthcare.

Broadening out, China's AI trajectory fits a familiar historical pattern. Chinese firms entered the PC era nearly a decade behind Silicon Valley; by the internet era the gap had narrowed to a few years. With AI, the lag is now measured in months. As in every major industrial or technological wave, it will be competitive intensity—rather than outright dominance by either side—that will drive progress.

China will need that momentum. A shrinking population, low productivity relative to OECD standards, and swaths of low-margin, low-digitization industries raise the stakes for the country. That said, the barriers to AI adoption look the same everywhere. Whether in Hangzhou, Houston, or Hamburg, companies struggle to turn AI enthusiasm into real business impact—we often say that AI is everywhere except the bottom line.

The most significant barriers to adoption aren't technical but rather organizational: leadership, incentives, and re-designed workflow. Our colleagues see them all over the world, in developed and developing markets; we see them daily in China. Corporate structures anywhere can feel slow, hierarchical, and allergic to risk. And adoption, how quickly societies absorb and normalize a technology, is what drives impact. Frankly, in that context, "good enough" computing power matters more than absolute best. Ultimately, the biggest factor in AI-driven productivity won't be who taps the better models but which CEOs are willing to reinvent how their organizations operate. "At the end of the day, you don't keep score by looking at how good these large language models are. The score is being kept by the adoption rate," says Joe of Alibaba.

For now, AI's financial impact inside companies remains modest, even as its broader economic impact in China is likely to be enormous. As Unitree Robotics CEO Xingxing Wang—whose company is known for robot dogs and agile humanoids—humbly puts it, the industry is still in a "desert stage. Just a few blades of grass sprouting. The eve of large-scale explosive growth hasn't arrived yet."

We expect the next wave of momentum to emerge at the application layer, with gains to society realized only as organizations and individuals adapt how they work. The potential of AI is universal; the challenge of capturing it remains distinctly human.