In 2016, the future looked bright for tech companies outside of Silicon Valley.
American investors were suddenly waking up to opportunities overseas, and allocating money accordingly. Venture firms that had once focused almost exclusively on Silicon Valley were now chasing deals in Beijing, Bengaluru, Jakarta, and São Paolo — kicking off a nearly decade-long golden age for global startups. That year, for the first time, private companies outside the U.S. raised more than American companies.
In the summer of 2016, Japanese social networking app Line went public in the biggest global tech IPO of the year. Indonesia’s Gojek was fast transforming from a motorbike ride-hailing service into a sprawling “super app,” becoming one of Southeast Asia’s only unicorns and later, a “decacorn,” valued at roughly $10 billion.
The spread of mobile technology fueled a fintech boom in places like Brazil, where Nubank was building what would become one of the world’s largest digital banks. In India, Walmart’s $16 billion investment in the online store Flipkart in 2018 proved the value of the Indian market for global investors.
Venture-backed companies outside the U.S. raised a little over $100 billion in 2016. By 2021, that figure stood at more than $300 billion. The surge marked a brief window when, at least by one metric, it looked like technology’s long-promised equalizing power was one step closer to becoming real.
But in 2026, that window appears to have closed. Investment trends have flipped back: In 2024, the U.S. once again seized the top spot for startup funding, outpacing all other countries on the planet combined. Last year, the gap grew even wider. Much of those investment trends have to do with a single, increasingly all-encompassing technology: artificial intelligence.
In mid-February, Anthropic announced it had raised an eye-watering $30 billion at a $380 billion valuation. Two weeks later, OpenAI raised the stakes dramatically, announcing a $110 billion round at a $840 billion valuation. All in, according to an analysis by the Organisation for Economic Co-operation and Development, U.S. AI firms attracted 75% of all AI investment last year — $194 billion. That’s nearly half of all of global venture funding, across every industry.
This funding has given American companies a massive strategic advantage, providing capital that these companies can use not only to recruit top talent but also to invest in the physical infrastructure that the AI industry uniquely requires: big, powerful data centers stacked with expensive and hard-to-source chips. These data centers also require massive amounts of water and electricity to run, and many billions of dollars to build and operate — resources that much of the world simply lacks.
“This particular market is rigged in ways that always cut against the global majority,” said Amba Kak, co-executive director of the AI Now Institute.
The one counterweight to the United States’ AI dominance, of course, is China, whose AI industry has been heavily subsidized by the government. Even so, U.S. export controls on the most advanced chips have thwarted even China’s ability to match the raw computing power of U.S.-based rivals.
The U.S. — and to some extent China — now has a potentially insurmountable lead in owning the world’s foundational AI models. That doesn’t mean other countries can’t develop their own robust AI ecosystems with fine-tuned technology built on top of those models. It does mean, however, that countries that do so will be increasingly dependent on a small number of American and Chinese firms. That leaves them vulnerable to shifting geopolitical winds and the risk that these companies might one day swallow their global competitors whole.
Even as global leaders and entrepreneurs outside the West scramble for some measure of self-determination by rushing to build their own “sovereign AI” ecosystems from scratch, their fate may be sealed.
“What we’re seeing [is] this kind of grandstanding bluster, like, ‘We can compete. We can build our own AI startup ecosystem,’ which doesn’t feel like it’s fully calling out the elephants in the room,” Kak said.
Just over three years on from the public release of ChatGPT, nearly every data point available about our AI era tells a startling story of geographic resource concentration.
Since 2023, more than 4,000 venture-backed AI companies have been founded in the U.S., according to Crunchbase — roughly 800 more than the entire rest of the world combined.
The investment gap is even starker: The top 10 global AI investors led $96 billion in funding rounds for U.S. AI companies last year, compared with just $1.9 billion across all other countries combined.
Those dollar amounts, of course, are skewed by a few colossal investments. But the sheer number of deals in the U.S. versus in the rest of the world tells a similar story. Last year, the top 10 global investors by deal count made 1,261 investments in AI in the U.S. and just 271 everywhere else.
“This is unprecedented,” said Gené Teare, Crunchbase News’ senior data editor, noting that investments in AI now account for 50% of all private funding globally.
The clustering of investment is even more stark because of the language around AI’s promise of being inclusionary. OpenAI CEO Sam Altman, for one, is fond of describing AI — like the internet before it — as “an equalizing force” for the world. Nvidia CEO Jensen Huang has called it “the great equalizer.”
AI tools, that argument goes, can be used by anyone anywhere. In theory, that should lower barriers to entry to build the next big thing, because now, everyone has access to the same technical capabilities. “You could be sitting in a mountain in Lesotho in southern Africa with a Starlink backpack using Claude Code,” said Shu Nyatta, managing partner of the growth equity firm Bicycle Capital, which invests in Latin America. “Everyone on Earth with a connection to the internet, which is effectively like almost everyone at this point, can use AI to do stuff.”
And yet, in reality, at least so far, the AI era appears to be entrenching power where it already exists.
If any country among emerging markets were poised to buck that trend, it would be India. The country has one of the world’s largest pools of tech workers, a massive digital economy, and a government that has thrown its full weight behind becoming a global AI leader.
“My vision is that India should be among the top three AI superpowers globally, not just in the consumption of AI but in creation,” Prime Minister Narendra Modi recently declared at the start of the AI Impact Summit, which was hosted in New Delhi in February. The government’s flagship AI program has devoted more than $1 billion to AI investments, and the country is reportedly preparing to unveil another $11 billion fund, focused on chipmaking.
But already, stories of once-promising Indian AI companies going bust are piling up. Mad Street Den, a computer vision and AI firm was acquired in a distress sale last year. CodeParrot, a Y Combinator-backed startup launched by two Indian founders, also closed up shop in 2025 after failing to “break through,” its co-founder Vedant Agarwala wrote on LinkedIn. And the Hyderabad-based enterprise AI startup Subtl.ai similarly closed down last year due to funding shortages and an inability to attract paying clients.
“Some Investors flirt ALOT with founders, but it doesn’t mean shit until they give you a term sheet,” Subtl.ai co-founder Vishnu Ramesh wrote in his own LinkedIn post. “I had made decisions assuming that funding budgets would come in based on conversations I had with few good funds out of Hyderabad. This became a big reality check for me.”
Other AI firms that have been core to India’s sovereign AI ambitions, meanwhile, have struggled out of the gate. Krutrim, one of India’s first AI unicorns, which was started by Indian billionaire and Ola Cabs founder Bhavish Aggarwal, has spread itself thin attempting to take on every aspect of the AI ecosystem. That includes building its own AI agent, its own chips, and marketing access to Nvidia GPUs as a service to other businesses. It has, in turn, conducted several rounds of layoffs in its less than three-year history and is far from achieving the scale of its American competitors, which are increasingly expanding in the region.
Even Sarvam AI, one of the most advanced Indian AI firms, faced criticism after launching a much-hyped model in 2025, with an investor at Menlo Ventures calling the rollout “embarrassing.” Sarvam had sought to offer an alternative to U.S. and Chinese large language models by training its model on Indic languages, but saw weak developer downloads within the first few days of its highly anticipated launch. “Definitely not all, but much of the Indian AI scene seems like more ‘I want to do cool AI things that cool AI people do,’ not ‘let’s solve important hard problems,’” Deedy Das, a partner at Menlo Ventures, wrote on X at the time.
Sarvam has since unveiled two new India-specific frontier models (and Das later said he was “wrong about Sarvam”), but the ordeal highlighted the challenges global AI startups face in being constantly compared to their far larger and better-funded American counterparts. “There’s a lot of antagonistic, ‘us versus the West’ mentality” in Indian tech, even though no one in Silicon Valley is really thinking about India as a threat, said Ria Mirchandani, an early product manager at Sarvam who is now building her own data infrastructure firm GoldMind in San Francisco.
Part of the problem, as she sees it, is this drive to build the OpenAI of India, rather than focusing on discrete local problems that don’t require replicating Silicon Valley’s playbook.
In India, in particular, one of the biggest challenges facing AI startups is the country’s legacy as the world’s back office, said Goutham Ramkumar, a former Uber and Amazon Web Services executive based in Bengaluru, who was most recently director of evaluation and AI data at Krutrim. “We’re sort of caught in that self-fulfilling prophecy,” Ramkumar said.
Because international investors continue to view India — and many other developing countries — as a source of low-cost labor, he said, they continue to treat it that way. That means smaller rounds of funding, increased pressure on companies to generate revenue quickly and diminished ability to hire from the small and increasingly expensive pool of top AI talent. In a resource and time-intensive field like AI, these can be crippling constraints — particularly given that in the U.S., money-losing competitors like OpenAI can pledge to spend more than a trillion dollars on infrastructure development before making a penny of profit, and still have investors banging down their doors.
This focus on cost doesn’t just change the kind of companies that get built, Ramkumar argued; it changes the way startups and the government itself market India’s value to the AI ecosystem. During a recent conversation at the World Economic Forum, a top Indian official suggested India is poised to win the next industrial age in part by driving down costs. “India is still stuck in that strategic positioning of labor arbitrage,” said Ramkumar. At a time when AI tools can increasingly do that low-wage IT work, he said, that’s a dangerous position to be in.
If building an AI ecosystem is challenging in India, it’s even more difficult in other developing countries. According to Tracxn data, fewer than 45 AI startups have been founded in Africa since 2023, and they’ve raised less than $40 million between them. That’s hardly much of a surprise given that Africa is currently home to less than 1% of the world’s data center capacity.
Africa has also seen the lowest levels of AI adoption in the world, according to a recent Microsoft report, which warns of a “widening digital divide” between the Global North and Global South. “India, Saudi Arabia, they’re on the right track,” Microsoft president Brad Smith said during a recent panel at the World Economic Forum. “It’s hard to look at Africa as a whole and be equally optimistic.”
Cassava Technologies, a London-based tech company founded in 2021 by Zimbabwean billionaire Strive Masiyiwa, has committed to building AI data centers, equipped with Nvidia GPUs, in five different countries on the continent, starting in South Africa. But ensuring that the local economy, with its limited purchasing power, can actually tap into that resource has required philanthropic groups to step up. The Rockefeller Foundation has already committed to subsidize access for its grantees in the region, and has announced support for several nonprofits and companies in the agricultural, health, and education spaces to do so.
Andrew Sweet, vice president of innovation at the Foundation, said the goal of the partnership with Cassava is to “make sure that the technology also goes to the entrepreneurs, the researchers, the nonprofits,” and not only to the multinational companies and hyperscalers trying to stake out computing capacity on the continent.
While the Cassava effort is a start, all in, the company plans to equip what it calls its “AI factories” with just 12,000 Nvidia GPUs. Compare that to OpenAI — a single company which, according to CEO Sam Altman, was on track to bring more than 1 million GPUs online last year.
In some ways, Kak from the AI Now Institute argued, imbalance is rational, given global wealth disparities and poverty levels: “It would be totally ludicrous and even more concerning to see a diversion of resources from solving more immediate material, social, and economic concerns to, say, investing in building national AI infrastructure.”
But it also means that whatever AI ecosystem does eventually blossom on the continent will be increasingly reliant on technology that is ultimately shaped, designed, and controlled by foreign powers, most notably China. According to the Microsoft report, China’s free DeepSeek model has up to four times as many users in Africa than in other regions. Kak said the players who own AI infrastructure will inevitably become the “kingmakers,” with the power to make or break the companies — and economies — that rely on them.
Nyatta from Bicycle Capital cautioned that startup formation is only one way to measure the kind of economic opportunity that AI will bring to different countries. It’s equally important, he said, to look at the way established players in different regions actually use AI tools that already exist.
“There’s a graveyard of technologies that never get diffused into the economy, and never make anyone money,” Nyatta said of the startup landscape. “The real value from AI will come from it touching the existing world.”
Today, he said, the AI era may ultimately reward countries that can adopt AI quickly rather than those that can build expensively. “You don’t need to be building the next frontier model,” he said. “What we’ve built so far is so undiffused into the real economy that it will take a couple of decades to diffuse it.”
Nyatta likened the hundreds of billions of dollars now pouring into data centers and AI companies in the U.S. and China to the uneven distribution of global military spending. Not every country needs to manufacture its own fighter jets or submarines to benefit from modern defense capabilities, he said. “As long you have access to these technologies in some way that you think is reasonably durable, you don’t need to get obsessed with trying to build your own. It’s very hard to build your own, just like a navy.”
That may be true. Then again, there’s another truth tucked into that analogy: Throughout history, it’s the countries with the biggest arsenals that tend to call the shots.
