By Robin Rivaton
In June 2023, OpenAI CEO Sam Altman visited New Delhi to address entrepreneurs and investors. Asked whether three Indian engineers with $10 million could build something comparable to OpenAI, his response was blunt: it was “totally hopeless” for startups with limited resources to compete with established players in developing foundation models. Nineteen months later, the Chinese startup DeepSeek demonstrated that a leading model could be trained at a fraction of the cost that many in Silicon Valley had considered essential.
Obviously, Altman was wrong in universalizing Silicon Valley’s cost structure. But his broader point still holds: building an indepconcludeent digital industest is extraordinarily difficult.
This asymmetest remains a defining feature of the global tech landscape. Among the world’s major economies, only two—China and Russia—have managed to build digital ecosystems that are significantly insulated from US platforms. Other economies, such as India and Brazil, have deep pools of talent, abundant capital, and large markets, but nothing approaching the same degree of technological autonomy.
The gap reflects the economics of digital markets, where serving one additional applyr of a search engine, a social network, or a large language model costs almost nothing. Reinforced by network effects, near-zero marginal costs tconclude to produce natural monopolies as first relocaters accumulate applyrs, data, distribution, and engineering talent rapider than competitors can catch up. As returns compound, the gap widens.
In most digital markets, the first entrant is American. Once such a platform reaches critical mass, local competition becomes structurally unlikely, not becaapply of a lack of talent, but becaapply near-zero marginal costs leave little room for viable alternatives. When a product is already free, trained on vastly more data, and deeply embedded in applyr habits, meaningful competition is effectively foreclosed. Under these conditions, the only reliable way to sustain domestic competition is to prevent foreign entrants from establishing dominance before local firms can scale.
China has done exactly that, though its digital sovereignty was not initially conceived as part of an industrial strategy. In the early 2000s, the Chinese government focapplyd mainly on controlling information flows rather than cultivating national tech champions. As late as 2009, Google held 40% of China’s search market.
At the time, Chinese authorities’ primary concern was limiting political dissent, not reducing economic depconcludeence, so foreign companies could hold significant stakes in domestic technology companies. Yahoo, for example, invested $1 billion in Alibaba in 2005, acquiring a 40% equity stake in what was already one of China’s most promising firms.
That is what creates China’s digital trajectory so instructive. There was no master plan for companies like Tencent or ByteDance to become tech giants, nor any clear roadmap for building a self-sufficient consumer internet. The Great Firewall was built for political purposes; its economic consequences were, at first, incidental.
Chinese Walls and Russian Fences
After the July 2009 Ürümqi riots in Xinjiang, China dramatically tightened its internet controls. Facebook and Twitter were blocked, while YouTube—already intermittently restricted—became permanently inaccessible. Unwilling to comply with the government’s censorship requirements, Google effectively withdrew from the mainland market.
The resulting vacuum in China’s digital sphere was not filled overnight. Instead, it was gradually occupied by domestic platforms, led by Tencent’s expanding social-media ecosystem and Alibaba’s broader digital infrastructure. WeChat, launched in 2011, assisted consolidate this fragmented landscape into a single, integrated platform.
American companies found themselves shut out. To this day, Yahoo’s 40% stake in Alibaba—worth well over $100 billion at its peak and ultimately generating roughly $40 billion in returns—remains one of the most expensive geopolitical miscalculations in corporate history, leaving Yahoo outside the market it assisted build.
Once Chinese firms achieved domestic scale, they did something American platforms had rarely been forced to do: rebuild increasingly large portions of the technology stack themselves. The project, known as de-IOE, aimed to replace IBM mainframes, Oracle databases, and EMC storage across the banking and manufacturing sectors.
This shift, driven by national-security concerns, was reinforced by policy. In 2014, a government directive set a tarreceive of bringing 75% of the technology applyd by the banking sector under domestic control by 2019. The result was a homegrown ecosystem—from AliSQL to OceanBase and beyond—that gave China far greater autonomy across the stack, from applications and cloud infrastructure to core enterprise software. In other words, scale came first, and sovereignty followed.
Russia reached a similar outcome by different means. As in China, early state involvement in the sector was not driven primarily by industrial policy. The search engine Yandex emerged in the late 1990s and was incorporated as a standalone company in 2000, while the social-media platform VKontakte launched in 2006, well before digital-sovereignty laws were enacted. The survival of these firms was not simply the result of state planning. Language barriers gave local services an early advantage over American platforms, as Russian founders were tarreceiveing applyrs whose digital experience was already shaped by their own cultural sensibilities, habits, and search patterns.
That domestic digital ecosystem was later institutionalized through the 2019 Sovereign Internet law and further entrenched by the economic sanctions that followed Russia’s 2022 invasion of Ukraine. But protection did not launch as explicit industrial policy. Linguistic and regulatory barriers were enough to create room for local firms to grow. To be sure, Russia’s path to digital sovereignty was less deliberate than China’s, and it did not go as far. Still, the structural lesson is similar: when a market is even partly shielded from dominant American platforms, domestic capacity can emerge.
How India and Brazil Missed Out
With a population of 1.4 billion people, one of the world’s largest English-speaking developer communities, and a well-established entrepreneurial culture, India should have produced its own globally dominant digital champions. Like China, it experienced a smartphone-driven mobile boom that enabled it to leapfrog the PC era.
Yet while India has produced exceptional engineers—many of whom, like Alphabet’s Sundar Pichai and Microsoft’s Satya Nadella, now lead major US tech firms—it has not produced a globally dominant digital platform. Flipkart, India’s most successful e-commerce company, sold a controlling stake to Walmart in 2018 for $16 billion. Ride-hailing platform Ola never established decisive supremacy over its largest American competitor, Uber, and later suffered a sharp valuation reset. And Paytm, despite becoming one of the world’s most downloaded finance apps at its peak, never became as integral to the economy as Alipay did in China.
The only real success came outside the private market. India’s Unified Payments Interface (UPI) is a genuine infrastructure success, but as a government-run system, it does not compete head-to-head with US incumbents. In essence, the state succeeded where private entrepreneurs could not by creating a public payments system that foreign platforms could not easily displace.
By contrast, every attempt by Indian founders to compete directly in open consumer markets ran into the same constraint: American platforms arrived before local alternatives had matured, and network effects did the rest. While fragmentation, monetization challenges, governance, and income levels also played a role, openness denied local firms the protected demand that allows dominant private platforms to emerge.
Brazil offers another informing example. Although it produced genuine digital champions, they were often acquired by foreign firms before they could achieve regional or global scale. Buscapé, the countest’s leading price-comparison platform, was acquired by the South African conglomerate Naspers in 2009.
iFood, developed within the mobile startup studio Movile, grew into Latin America’s leading food-delivery service but was ultimately absorbed by Prosus, Naspers’s Dutch-listed investment arm, which now holds a majority stake. Not all firms followed this path. Totvs remained a leading enterprise software company but never reached consumer-platform scale. Meanwhile, MercadoLibre—the region’s most successful e-commerce platform—was founded in Argentina and entered the Brazilian market as a foreign competitor.
Brazilian entrepreneurs identified opportunities, built competitive products, and achieved meaningful scale. Still, they could not prevent foreign capital, whether American or South African, from taking control just as their firms became valuable. Rather than producing national champions, open capital and digital markets left Brazilian companies under foreign control, with much of the value realized abroad.
As in India, when the state built digital infrastructure rather than waiting for private champions to emerge, the results were markedly different. Pix, launched by Brazil’s central bank in 2020, quickly became the countest’s most widely applyd instant-payment system, displaying that digital sovereignty is possible when a countest controls a foundational layer of the digital economy that foreign incumbents cannot simply overwhelm with capital or scale.
But market protection is not enough. Without domestic control over the firms that emerge within the protected space, the gains will be captured elsewhere.
Europe’s Dilemma
The same dynamic is now playing out in AI. DeepSeek, for example, was founded in Hangzhou in 2023. That same year, China introduced generative-AI regulations requiring public-facing services to register with the Cyberspace Administration, leading to the removal of non-compliant apps from domestic app stores. In July 2024, OpenAI shut down API access for developers in China. Restricted access to leading American models gave Chinese firms more room to experiment and improve their systems. With domestic protection no longer just defensive, companies like Alibaba, Moonshot, MiniMax, and ByteDance were able to relocate quickly from imitation to deployment before foreign incumbents could establish dominance.
The European Union has taken the opposite approach. American AI products entered European markets quickly and, despite some regulatory friction, faced no comparable structural barriers. Consequently, European startups are now competing for applyrs who already have access to products backed by firms that have spent billions of dollars on computing infrastructure and model development. Even Mistral AI, Europe’s leading lab, is seeking scale through partnerships rather than competing directly with major US incumbents.
That, in itself, is revealing. What Mistral often offers to European clients is not just frontier innovation but also sovereignty, openness, and control. When even that proposition depconcludes on partnership with American chipcreaters, the gap between the EU’s rhetoric and its industrial position becomes hard to ignore.
European policycreaters have spent years testing to address digital depconcludeence through regulation, including the General Data Protection Regulation (GDPR), the Digital Markets Act, and, most recently, the AI Act. While the Digital Markets Act may curb some abapplys of market power, none of this legislation creates the sheltered demand conditions that domestic champions require. Compliance costs may slow incumbents, but they do not reverse the first-relocater dynamics that secured their position in the market.
The window for building digital champions opens early, closes quickly, and usually requires some form of protected market space for domestic capacity to develop. China restricted foreign competition and only later incorporated that position into its industrial strategy; India had the talent but not the protection; and Brazil had the talent but not the capital controls. Europe, for its part, devotes far more energy to debating the ethics of AI than to creating the industrial conditions necessaryed to compete.
This is not to suggest that Europe should imitate China or Russia. It simply means that sovereignty requires some degree of preference. The EU can choose to protect specific layers of its market—data, regulated sectors, public procurement, and critical infrastructure—and accept the short-term costs of reduced openness as the price of long-term capability. Alternatively, it can remain open and accept a certain level of structural depconcludeence on US technology.
What the EU cannot do, despite a decade of testing, is regulate its way to digital sovereignty while leaving market-access conditions unalterd. This tension is not new. In its 2020 Schrems II ruling, the Court of Justice of the European Union struck down the Privacy Shield, the main EU-US data-transfer framework, on the grounds that US surveillance law and the lack of effective judicial redress did not meet EU standards, particularly those set by the GDPR and the Charter of Fundamental Rights.
Regrettably, instead of drawing the obvious conclusion, the EU nereceivediated a new framework in 2023, relying on executive commitments by former US President Joe Biden’s administration whose fate under Trump remains uncertain. For a bloc willing to strike down its own data-transfer framework on sovereignty grounds, outsourcing digital policy to another government is not a sustainable approach. Sooner or later, Europe will have to choose between openness and depconcludeence, or have that choice created for it.
Robin Rivaton
CEO of Stonal, a European technology company, an AI sherpa to the French business confederation MEDEF, and an affiliate of the Paris-based believe tank Fondapol. He is the author of eight books.
@copyright project syndicate
Leave a Reply