Startups

Readers in a hurry may wish to put this article aside for later. It is an important long-form exploration, not a quick read. – Editors

---

We are in the early days of a seismic shift in the global AI industry. DeepSeek, a previously little-known Chinese artificial intelligence company, has produced a “game changing”“ large language model that promises to reshape the AI landscape almost overnight.

But DeepSeek’s breakthrough also has wider implications for the technological arms race between the US and China, having apparently caught even the best-known US tech firms off guard. Its launch has been predicted to start a “slow unwinding of the AI bet” in the West, amid a new era of “AI efficiency wars.”

In fact, industry experts have been speculating for years about China’s rapid advancements in AI. While the supposedly free-market US has often prioritized proprietary models, China has built a thriving AI ecosystem by leveraging open-source technology, fostering collaboration between government-backed research institutions and major tech firms.

This strategy has enabled China to scale its AI innovation rapidly while the US – despite all the tub-thumping from Silicon Valley – remains limited by restrictive corporate structures. Companies such as Google and Meta, despite promoting open-source initiatives, still rely heavily on closed-source strategies that limit broader access and collaboration.

What makes DeepSeek particularly disruptive is its ability to achieve cutting-edge performance while reducing computing costs – an area where US firms have struggled due to their dependence on training models that demand very expensive processing hardware.

Where once Silicon Valley was the epicentre of global digital innovation, its corporate behemoths now appear vulnerable to more innovative, “scrappy” startup competitors – albeit ones enabled by major state investment in AI infrastructure. By leveraging China’s industrial approach to AI, DeepSeek has crystalized a reality that many in Silicon Valley have long ignored: AI’s center of power is shifting away from the US and the west.

It highlights the failure of US attempts to preserve its technological hegemony through tight export controls on cutting-edge AI chips to China. According to research fellow Dean Ball: “You can keep [computing resources] away from China, but you can’t export-control the ideas that everyone in the world is hunting for.”

DeepSeek’s success has forced Silicon Valley and large Western tech companies to “take stock,” realizing that their once-unquestioned dominance is suddenly at risk. Even the US president, Donald Trump, has proclaimed that this should be a “wake-up call for our industries that we need to be laser-focused on competing.”

But this story is not just about technological prowess – it could mark an important shift in global power. Former US secretary of state Mike Pompeo has framed DeepSeek’s emergence as a “shot across America’s bow,” urging US policymakers and tech executives to take immediate action.

DeepSeek’s rapid rise underscores a growing realization: Globally, we are entering a potentially new AI paradigm, one in which China’s model of open-source innovation and state-backed development is proving more effective than Silicon Valley’s corporate-driven approach.

I’ve spent much of my career analyzing the transformative role of AI on the global digital landscape – examining how AI shapes governance, market structures and public discourse while exploring its geopolitical and ethical dimensions, now and far into the future.

I also have personal connections with China, having lived there while teaching at Jiangsu University and then written my PhD thesis on the country’s state-led marketization program. Over the years I have studied China’s evolving tech landscape, observing firsthand how its unique blend of state-driven industrial policy and private-sector innovation has fueled rapid AI development.

I believe this moment may come to be seen as a turning point not just for AI but for the geopolitical order. If China’s AI dominance continues, what could this mean for the future of digital governance, democracy, and the global balance of power?

China’s open-source AI takeover

Even in the early days of China’s digital transformation, analysts predicted the country’s open-source focus could lead to a major AI breakthrough. In 2018, China was integrating open-source collaboration into its broader digitization strategy, recognizing that fostering shared development efforts could accelerate its AI capabilities.

Unlike the US, where proprietary AI models dominated, China embraced open-source ecosystems to bypass Western gatekeeping, to scale innovation faster and to embed itself in global AI collaboration.

China’s open-source activity surged dramatically in 2020, laying the foundation for the kind of innovation seen today. By actively fostering an open-source culture, China ensured that a broad range of developers had access to AI tools, rather than restricting them to a handful of dominant companies.

The trend has continued in recent years, with China even launching its own state-backed open-source operating systems and platforms, in 2023, to further reduce its dependence on western technology. This move was widely seen as an effort to cement its AI leadership and create an independent, self-sustaining digital ecosystem. .

While China has been steadily positioning itself as a leader in open-source AI, Silicon Valley firms remained focused on closed, proprietary models – allowing China to catch up fast. While companies like Google and Meta promoted open-source initiatives in name, they still locked key AI capabilities behind paywalls and restrictive licenses.

In contrast, China’s government-backed initiatives have treated open-source AI as a national resource, rather than a corporate asset. This has resulted in China becoming one of the world’s largest contributors to open-source AI development, surpassing many western firms in collaborative projects. Chinese tech giants such as Huawei, Alibaba and Tencent are driving open-source AI forward with frameworks like PaddlePaddle, X-Deep Learning (X-DL) and MindSpore — all now core to China’s machine learning ecosystem.

But they’re also making major contributions to global AI projects, from Alibaba’s Dragonfly, which streamlines large-scale data distribution, to Baidu’s Apollo, an open-source platform accelerating autonomous vehicle development. These efforts don’t just strengthen China’s AI industry, they embed it deeper into the global AI landscape.

This shift had been years in the making, as Chinese firms (with state backing) pushed open-source AI forward and made their models publicly available, creating a feedback loop that western companies have also – quietly – tapped into.

A year ago, for example, US firm Abicus.AI released Smaug-72B, an AI model designed for enterprises that built directly upon Alibaba’s Qwen-72B and outperformed proprietary models like OpenAI’s GPT-3.5 and Mistral’s Medium.

But the potential for US companies to further build on Chinese open-source technology may be limited by political as well as corporate barriers.

In 2023, US lawmakers highlighted growing concerns that China’s aggressive investment in open-source AI and semiconductor technologies would eventually erode western leadership in AI. Some policymakers called for bans on certain open-source chip technologies, due to fears they could further accelerate China’s AI advancements.

By then, however, China’s AI horse had already bolted.

AI with Chinese characteristics

DeepSeek’s rise should have been obvious to anyone familiar with management theory and the history of technological breakthroughs linked to “disruptive innovation.” Latecomers to an industry rarely compete by playing the same game as incumbents – they have to be disruptive.

China, facing restrictions on cutting-edge western AI chips and lagging behind in proprietary AI infrastructure, had no choice but to innovate differently. Open-source AI provided the perfect vehicle: a way to scale innovation rapidly, lower costs and tap into global research while bypassing Silicon Valley’s resource-heavy, closed-source model.

From a Western and traditional human rights perspective, China’s embrace of open-source AI may appear paradoxical, given the country’s strict information controls. Its AI development strategy prioritizes both technological advancement and strict alignment with the Chinese Communist party’s ideological framework, ensuring AI models adhere to “core socialist values” and state-approved narratives.

AI research in China has thrived not only despite these constraints but, in many ways, because of them.

China’s success goes beyond traditional authoritarianism; it embodies what Harvard economist David Yang calls “Autocracy 2.0.” Rather than relying solely on fear-based control, it uses economic incentives, bureaucratic efficiency and technology to manage information and maintain regime stability.

The Chinese government has strategically encouraged open-source development while maintaining tight control over AI’s domestic applications, particularly in surveillance and censorship.

Indeed, authoritarian regimes may have a significant advantage in developing facial-recognition technology due to their extensive surveillance systems. The vast amounts of data collected through these networks enable private AI companies to create advanced algorithms, which can then be adapted for commercial uses, potentially accelerating economic growth.

China’s AI strategy is built on a dual foundation of state-led initiatives and private-sector innovation. The country’s AI roadmap, first outlined in the 2017 new generation artificial intelligence development plan, follows a three-phase timeline: achieving global competitiveness by 2020, making major AI breakthroughs by 2025, and securing world leadership in AI by 2030. In parallel, the government has emphasised data governance, regulatory frameworks and ethical oversight to guide AI development “responsibly.”

A defining feature of China’s AI expansion has been the massive infusion of state-backed investment. Over the past decade, government venture capital funds have injected approximately US$912 billion into early-stage firms, with 23% of that funding directed toward AI-related companies. A significant portion has targeted China’s less-developed regions, following local investment mandates.

Compared with private venture capital, government-backed firms often lag in software development but demonstrate rapid growth post-investment. Moreover, state funding often serves as a signal for subsequent private-sector investment, reinforcing the country’s AI ecosystem.

China’s AI strategy represents a departure from its traditional industrial policies, which historically emphasized self-sufficiency, support for a handful of national champions and military-driven research.

Instead, the government has embraced a more flexible and collaborative approach that encourages open-source software adoption, a diverse network of AI firms and public-private partnerships to accelerate innovation. This model prioritizes research funding, state-backed AI laboratories, and AI integration across key industries including security, healthcare and infrastructure.

Despite strong state involvement, China’s AI boom is equally driven by private-sector innovation. The country is home to an estimated 4,500 AI companies, accounting for 15% of the world’s total.

As economist Liu Gang told the Chinese Communist Party’s Global Times newspaper: “The development of AI is fast in China – for example, for AI-empowered large language models. Aided with government spending, private capital is flowing to the new sector. Increased capital inflow is anticipated to further enhance the sector in 2025.”

China’s tech giants including Baidu, Alibaba, Tencent and SenseTime have all benefited from substantial government support while remaining competitive on the global stage. But unlike in the US, China’s AI ecosystem thrives on a complex interplay between state support, corporate investment and academic collaboration.

Recognizing the potential of open-source AI early on, Tsinghua University in Beijing has emerged as a key innovation hub, producing leading AI startups such as Zhipu AI, Baichuan AI, Moonshot AI and MiniMax — all founded by its faculty and alumni.

The Chinese Academy of Sciences has similarly played a crucial role in advancing research in deep learning and natural language processing.

Unlike the West, where companies like Google and Meta promote open-source models for strategic business gains, China sees them as a means of national technological self-sufficiency. To this end, the National AI Team, composed of 23 leading private enterprises, has developed the National AI Open Innovation Platform, which provides open access to AI datasets, toolkits, libraries and other computing resources.

DeepSeek is a prime example of China’s AI strategy in action. The company’s rise embodies the government’s push for open-source collaboration while remaining deeply embedded within a state-guided AI ecosystem. Chinese developers have long been major contributors to open-source platforms, ranking as the second-largest group on GitHub by 2021.

Founded by Chinese entrepreneur Liang Wenfeng in 2023, DeepSeek has positioned itself as an AI leader while benefiting from China’s state-driven AI ecosystem. Liang, who also established the hedge fund High-Flyer, has maintained full ownership of DeepSeek and avoided external venture capital funding.

Though there is no direct evidence of government financial backing, DeepSeek has reaped the rewards of China’s AI talent pipeline, state-sponsored education programs and research funding. Liang has engaged with top government officials including China’s premier, Li Qiang, reflecting the company’s strategic importance to the country’s broader AI ambitions.

In this way, DeepSeek perfectly encapsulates “AI with Chinese characteristics” – a fusion of state guidance, private-sector ingenuity and open-source collaboration, all carefully managed to serve the country’s long-term technological and geopolitical objectives.

Recognizing the strategic value of open-source innovation, the government has actively promoted domestic open-source code platforms like Gitee to foster self-reliance and insulate China’s AI ecosystem from external disruptions. However, this also exposes the limits of China’s open-source ambitions. The government pushes collaboration, but only within a tightly controlled system where state-backed firms and tech giants call the shots.

Reports of censorship on Gitee reveal how Beijing carefully manages innovation, ensuring AI advances stay in line with national priorities. Independent developers can contribute, but the real power remains concentrated in companies that operate within the government’s strategic framework.

The conflicted reactions of US big tech

DeepSeek’s emergence has sparked intense debate across the AI industry, drawing a range of reactions from leading Silicon Valley executives, policymakers and researchers. While some view it as an expected evolution of open-source AI, others see it as a direct challenge to western AI leadership.

Microsoft’s CEO, Satya Nadella, emphasized its technical efficiency. “It’s super-impressive in terms of both how they have really effectively done an open-source model that does this inference-time compute, and is super-compute efficient,” Nadella told CNBC. “We should take the developments out of China very, very seriously.”

Silicon Valley venture capitalist Marc Andreessen, a prominent advisor to Trump, was similarly effusive. “DeepSeek R1 is one of the most amazing and impressive breakthroughs I’ve ever seen – and, as open source, a profound gift to the world,” he wrote on X.

For Yann LeCun, Meta’s chief AI scientist, DeepSeek is less about China’s AI capabilities and more about the broader power of open-source innovation. He argued that the situation should be read not as China’s AI surpassing the US, but rather as open-source models surpassing proprietary ones. “DeepSeek has profited from open research and open source (e.g. PyTorch and Llama from Meta),” he wrote on Threads. “They came up with new ideas and built them on top of other people’s work. Because their work is published and open source, everyone can profit from it. That is the power of open research and open source.”

Not all responses were so measured. Alexander Wang, CEO of Scale AI – a US firm specializing in AI data labeling and model training – framed DeepSeek as a competitive threat that demands an aggressive response. He wrote on X: “DeepSeek is a wake-up call for America, but it doesn’t change the strategy: USA must out-innovate & race faster, as we have done in the entire history of AI. Tighten export controls on chips so that we can maintain future leads. Every major breakthrough in AI has been American.”

Elon Musk added fuel to speculation about DeepSeek’s hardware access when he responded with a simple “obviously” to Wang’s earlier claims on CNBC that DeepSeek had secretly acquired 50,000 Nvidia H100 GPUs, despite US export restrictions.

Beyond the tech world, US policymakers have taken a more adversarial stance. House speaker Mike Johnson accused China of leveraging DeepSeek to erode American AI leadership. “They abuse the system, they steal our intellectual property. They’re now trying to get a leg up on us in AI.”

For his part, Trump took a more pragmatic view, seeing DeepSeek’s efficiency as a validation of cost-cutting approaches. “I view that as a positive, as an asset …. You won’t be spending as much, and you’ll get the same result, hopefully.”

The rise of DeepSeek may have helped jolt the Trump administration into action, leading to sweeping policy shifts aimed at securing US dominance in AI. In his first week back in the White House, the US president announced a series of aggressive measures, including massive federal investments in AI research, closer partnerships between the government and private tech firms and the rollback of regulations seen as slowing US innovation.

The administration’s framing of AI as a critical national interest reflects a broader urgency sparked by China’s rapid advancements, particularly DeepSeek’s ability to produce cutting-edge models at a fraction of the cost traditionally associated with AI development. But this response is not just about national competitiveness – it is also deeply entangled with private industry.

Musk’s growing closeness to Trump, for example, can be viewed as a calculated move to protect his own dominance at home and abroad. By aligning with the administration, Musk ensures that US policy tilts in favour of his AI ventures, securing access to government backing, computing power,and regulatory control over AI exports.

At the same time, Musk’s public criticism of Trump’s US$500 billion AI infrastructure plan – claiming the companies involved lack the necessary funding – was as much a warning as a dismissal, signaling his intent to shape policy in a way that benefits his empire while keeping potential challengers at bay.

Not unrelated, Musk and a group of investors have just launched a US$97.4 billion bid for OpenAI’s nonprofit arm, a move that escalates his feud with OpenAI CEO Sam Altman and seeks to strengthen his grip on the AI industry. Altman has dismissed the bid as a “desperate power grab”, insisting that OpenAI will not be swayed by Musk’s attempts to reclaim control. The spat reflects how DeepSeek’s emergence has thrown US tech giants into what could be all-out war, fuelling bitter corporate rivalries and reshaping the fight for AI dominance.

And while the US and China escalate their AI competition, other global leaders are pushing for a coordinated response. The Paris AI Action Summit, held on February 10 and 11, has become a focal point for efforts to prevent AI from descending into an uncontrolled power struggle.

France’s president, Emmanuel Macron, warned delegates that without international oversight, AI risks becoming “the wild west,” where unchecked technological development creates instability rather than progress.

But at the end of the two-day summit, the UK and US refused to sign an international commitment to “ensuring AI is open, inclusive, transparent, ethical, safe, secure and trustworthy … making AI sustainable for people and the planet.” China was among the 61 countries to sign this declaration.

Concerns have also been raised at the summit about how AI-powered surveillance and control are enabling authoritarian regimes to strengthen repression and reshape the citizen-state relationship. This highlights the fast-growing global industry of digital repression, driven by an emerging “authoritarian-financial complex” that may exacerbate China’s strategic advancement in AI.

Equally, DeepSeek’s cost-effective AI solutions have created an opening for European firms to challenge the traditional AI hierarchy. As AI development shifts from being solely about compute power to strategic efficiency and accessibility, European firms now have an opportunity to compete more aggressively against their US and Chinese counterparts.

Whether this marks a true rebalancing of the AI landscape remains to be seen. But DeepSeek’s emergence has certainly upended traditional assumptions about who will lead the next wave of AI innovation – and how global powers will respond to it.

End of the ‘Silicon Valley effect’?

DeepSeek’s emergence has forced US tech leaders to confront an uncomfortable reality: They underestimated China’s AI capabilities. Confident in their perceived lead, companies like Google, Meta, and OpenAI prioritized incremental improvements over anticipating disruptive competition, leaving them vulnerable to a rapidly evolving global AI landscape.

In response, the US tech giants are now scrambling to defend their dominance, pledging over US$400 billion in AI investment. DeepSeek’s rise, fuelled by open-source collaboration, has reignited fierce debates over innovation versus security, while its energy-efficient model has intensified scrutiny on AI’s sustainability.

Yet Silicon Valley continues to cling to what many view as outdated economic theories such as the Jevons paradox to downplay China’s AI surge, insisting that greater efficiency will only fuel demand for computing power and reinforce their dominance. Companies like Meta, OpenAI and Microsoft remain fixated on scaling computational power, betting that expensive hardware will secure their lead. But this assumption blinds them to a shifting reality.

DeepSeek’s rise as the potential “Walmart of AI” is shaking Silicon Valley’s foundation, proving that high-quality AI models can be built at a fraction of the cost. By prioritizing efficiency over brute-force computing power, DeepSeek is challenging the US tech industry’s reliance on expensive hardware like Nvidia’s high-end chips.

This shift has already rattled markets, driving down the stock prices of major US firms and forcing a reassessment of AI dominance. Nvidia, whose business depends on supplying high-performance processors, appears particularly vulnerable as DeepSeek’s cost-effective approach threatens to reduce demand for premium chips.

The growing divide between the US and China in AI, however, is more than just competition – it’s a clash of governance models. While US firms remain fixated on protecting market dominance, China is accelerating AI innovation with a model that is proving more adaptable to global competition.

If Silicon Valley resists structural change, it risks falling farther behind. We may witness the unraveling of the “Silicon Valley effect”, through which tech giants have long manipulated AI regulations to entrench their dominance. For years, Google, Meta,and OpenAI shaped policies that favored proprietary models and costly infrastructure, ensuring AI development remained under their control.

DeepSeek is redefining AI with breakthroughs in code intelligence, vision-language models and efficient architectures that challenge Silicon Valley’s dominance. By optimizing computation and embracing open-source collaboration, DeepSeek shows the potential of China to deliver cutting-edge models at a fraction of the cost, outperforming proprietary alternatives in programming, reasoning and real-world applications.

More than a policy-driven rise, China’s AI surge reflects a fundamentally different innovation model – fast, collaborative and market-driven – while Silicon Valley holds on to expensive infrastructure and rigid proprietary control. If US firms refuse to adapt, they risk losing the future of AI to a more agile and cost-efficient competitor.

A new era of geotechnopolitics

But China is not just disrupting Silicon Valley. It is expanding “geotechnopolitics”, where AI is a battleground for global power. With AI projected to add US$15.7 trillion to the global economy by 2030, China and the US are racing to control the technology that will define economic, military and political dominance.

DeepSeek’s advancement has raised national security concerns in the US. Trump’s government is considering stricter export controls on AI-related technologies to prevent them from bolstering China’s military and intelligence capabilities.

As AI-driven defence systems, intelligence operations and cyber warfare redefine national security, governments must confront a new reality: AI leadership is not just about technological superiority, but about who controls the intelligence that will shape the next era of global power.

China’s AI ambitions extend beyond technology, driving a broader strategy for economic and geopolitical dominance. But with over 50 state-backed companies developing large-scale AI models, its rapid expansion faces growing challenges, including soaring energy demands and US semiconductor restrictions.

China’s president, Xi Jinping, remains resolute, stating: “Whoever can grasp the opportunities of new economic development such as big data and artificial intelligence will have the pulse of our times.” He sees AI driving “new quality productivity” and modernizing China’s manufacturing base, calling its “head goose effect” a catalyst for broader innovation.

To counter western containment, China has embraced a “guerrilla” economic strategy, bypassing restrictions through alternative trade networks, deepening ties with the global south, and exploiting weaknesses in global supply chains. Instead of direct confrontation, this decentralized approach uses economic coercion to weaken adversaries while securing China’s own industrial base.

China is also leveraging open-source AI as an ideological tool, presenting its model as more collaborative and accessible than western alternatives. This narrative strengthens its global influence, aligning with nations seeking alternatives to western digital control. While strict state oversight remains, China’s embrace of open-source AI reinforces its claim to a future where innovation is driven not by corporate interests but through shared collaboration and global cooperation.

But while DeepSeek claims to be open access, its secrecy tells a different story. Key details on training data and fine-tuning remain hidden, and its compliance with China’s AI laws has sparked global scrutiny. Italy has banned the platform over data-transfer risks, while Belgium and Ireland launched privacy probes.

Under Chinese regulations, DeepSeek’s outputs must align with state-approved narratives, clashing with the EU’s AI Act, which demands transparency and protects political speech. Such “controlled openness” raises many red flags, casting doubt on China’s place in markets that value data security and free expression.

Many western commentators are seizing on reports of Chinese AI censorship to frame other models as freer and more politically open. The revelation that a leading Chinese chatbot actively modifies or censors responses in real time has fueled a broader narrative that western AI operates without such restrictions, reinforcing the idea that democratic systems produce more transparent and unbiased technology. This framing serves to bolster the argument that free societies will ultimately lead the global AI race.

But, at its heart, the “AI arms race” is driven by technological dominance. The US, China, and the EU are charting different paths, weighing security risks against the need for global collaboration. How this competition is framed will shape policy: lock AI behind restrictions, or push for open innovation.

DeepSeek, for all its transformational qualities, continues to exemplify a model of AI where innovation prioritizes scale, speed and efficiency over societal impact. This drive to optimize computation and expand capabilities overshadows the need to design AI as a truly public good.

In doing so, it eclipses this technology’s genuine potential to transform governance, public services and social institutions in ways that prioritize collective wellbeing, equity and sustainability over corporate and state control.

A truly global AI framework requires more than political or technological openness. It demands structured cooperation that prioritizes shared governance, equitable access, and responsible development.

Following a workshop in Shanghai hosted by the Chinese government last September, the UN’s general secretary, António Guterres, outlined his vision for AI beyond corporate or state control: “We must seize this historic opportunity to lay the foundations for inclusive governance of AI – for the benefit of all humanity. As we build AI capacity, we must also develop shared knowledge and digital public goods.”

Both the west and China frame their AI ambitions through competing notions of “openness” – aligned in both cases with their strategic interests and reinforcing existing power structures.

Western tech giants claim AI drives democratization, yet they often dominate digital infrastructure in parts of Africa, Asia and Latin America, exporting models based on “corporate imperialism” that extract value while disregarding local needs.

China, by contrast, positions itself as a technological partner for the rest of the Global South. However, its AI remains tightly controlled, reinforcing state ideology.

China’s proclaimed view on international AI collaboration emphasizes that AI should not be “a game of rich countries,”as President Xi stated during the 2024 G20 summit.

By advocating for inclusive global AI development, China positions itself as a leader in shaping international AI governance, especially via initiatives like the UN AI resolution and its AI capacity-building action plan. These efforts help promote a more balanced technological landscape while allowing China to strengthen its influence in global AI standards and frameworks.

However, beneath all these narratives, both China and the US share a strategy of AI expansion that relies on exploited human labor, from data annotation to moderation, exposing a system driven less by innovation than by economic and political control.

Peter Bloom is a professor of management at the University of Essex.

This article is republished from The Conversation under a Creative Commons license. Read the original article.

Continue Reading