Europe is completely out of this race
Uma corrida de proporções históricas está a remodelar o mundo tecnológico, com americanos e chineses a investir 400 mil milhões de dólares por ano na infraestrutura que sustentará a inteligência artificial do futuro. A Europa observa à margem, sem capital nem escala para competir, enquanto gigantes como a OpenAI enfrentam uma aritmética implacável: crescer para receitas de 200 mil milhões de dólares até 2030 ou arriscar o colapso. No horizonte, paira a questão que define esta era — não quem vencerá a corrida, mas quem sobreviverá a ela.
- Os Estados Unidos e a China dominam sozinhos um investimento anual de 400 mil milhões de dólares em infraestrutura de IA, criando uma barreira de entrada que poucos países ou empresas conseguem sequer imaginar transpor.
- A Europa ficou para trás de forma alarmante — a Mistral, empresa francesa frequentemente apresentada como esperança europeia, é uma presença solitária e diminuta face ao peso continental dos seus rivais.
- A OpenAI falhou o seu próprio objetivo de 30 mil milhões de dólares em receitas anuais, e os números acumulados de perdas exigem um crescimento quase sem precedentes na história empresarial para se justificarem.
- As empresas continuam a investir em escala massiva mesmo sem certeza de retorno, numa aposta coletiva de que o mercado acabará por absorver a despesa — mas a dúvida sobre quais sobreviverão paira sobre toda a indústria.
A corrida pela dominância na inteligência artificial acelera a um ritmo sem precedentes, mas é quase inteiramente um duelo entre titãs americanos e chineses. No último ano, as maiores empresas do setor investiram 400 mil milhões de dólares em infraestrutura — centros de dados, energia e hardware especializado. A Europa ficou de fora.
Miguel Godinho de Matos, responsável pelos programas de IA e machine learning na Católica Lisbon School of Business & Economics, traçou o panorama numa conferência em Lisboa. Os modelos americanos competem em regime proprietário; a China aposta no open-source. A Europa tem apenas a Mistral, empresa francesa que os seus compatriotas relutam em chamar europeia, pequena demais para alterar o equilíbrio de forças.
A adoção da IA cresce sobretudo entre utilizadores individuais; as empresas mantêm-se cautelosas. Ainda assim, os hyperscalers continuam a investir, apostando que o mercado acabará por justificar a despesa. "O que não sabemos é se todas estas empresas vão sobreviver", admitiu Godinho de Matos.
A OpenAI ilustra bem esta incerteza. As suas receitas cresceram de 3 mil milhões de dólares em 2020 para cerca de 13 mil milhões em 2025 — mas ficaram aquém do objetivo de 30 mil milhões. Para que os números façam sentido face ao capital já queimado, a empresa precisaria de atingir 200 mil milhões de dólares em receitas anuais até 2030. Seria um dos maiores feitos empresariais da história. O fosso entre o que foi investido e o que foi ganho continua a alargar-se, e a questão central permanece: quais destas empresas resistirão à pressão, e quando?
The race for artificial intelligence dominance is accelerating at a pace that would have seemed impossible just years ago, but it is almost entirely a contest between American and Chinese titans. Last year alone, the world's largest AI companies poured 400 billion dollars into the infrastructure that makes their systems run—data centers, the power plants to feed them, the specialized hardware that processes information at scale. This spending is concentrated among a handful of "hyperscalers," the massive technology firms with the capital to build at this level. Europe, by contrast, has been left behind.
Miguel Godinho de Matos, who leads artificial intelligence and machine learning programs at Católica Lisbon School of Business & Economics, laid out the problem plainly at a conference in Lisbon this week. The performance of language models has improved at an almost bewildering speed, he explained, and what we are witnessing now is a full-scale race to dominate the field. The Americans have their proprietary models—OpenAI, Anthropic, Google, Microsoft—each charging for access and competing fiercely. China has taken a different path, releasing open-source models that anyone can use and modify. Europe, meanwhile, has almost nothing. There is Mistral, a French company that French officials insist on calling French rather than European, but it stands alone and small against the continental weight of its competitors.
The adoption of AI has been climbing steeply, particularly among individual users who have embraced the technology with enthusiasm. Businesses have been more cautious, approaching AI with skepticism about whether it actually solves their problems. Yet the hyperscalers continue to invest at a staggering pace, betting that the market will eventually justify the spending. The question that haunts the industry is whether it will. "What we don't know is whether all these companies will survive," Godinho de Matos said. It is one of the great unknowns.
OpenAI offers a case study in the mathematics of this uncertainty. In 2020, the company generated 3 billion dollars in revenue. By 2025, that had grown to roughly 13 billion. But the company had set a target of 30 billion dollars in annual revenue, and it missed that mark. Looking forward, the accumulated cash the company has burned through—the money spent on research, infrastructure, and operations without yet generating equivalent returns—creates a brutal arithmetic. For the numbers to make sense, for OpenAI to justify what it has spent and what it continues to spend, the company would need to be generating approximately 200 billion dollars in annual revenue by 2030. It is theoretically possible. But it would rank among the most extraordinary business successes in history. The company would have to grow from its current scale to something almost unimaginable in just five years.
This is the bind facing every hyperscaler in the field. The infrastructure costs are real and enormous. The revenue streams are growing but remain uncertain. The gap between what has been invested and what has been earned keeps widening. Some of these companies will not survive the pressure. Which ones, and when, remains the central question hanging over the entire industry.
Notable Quotes
The performance of language models has accelerated at an incredible speed, and right now we have a race— Miguel Godinho de Matos, Católica Lisbon School of Business & Economics
What we don't know is whether all these companies will survive. It is one of the great unknowns— Miguel Godinho de Matos
The Hearth Conversation Another angle on the story
Why has Europe fallen so far behind in this race?
It's partly about scale and capital. Building the infrastructure to train and run these models requires billions of dollars. American companies have access to venture capital and public markets willing to fund them. Chinese companies have state backing. European companies lack both.
But Europe has wealthy companies and investors. Why couldn't they compete?
They could have, but the moment passed. By the time European firms recognized what was happening, the Americans and Chinese had already built moats—they had the data, the talent, the computing power. Catching up now would mean spending more than the leaders, which is a losing proposition.
What about Mistral? Isn't that a European success?
It's a French company, and it's real, but it's also tiny compared to OpenAI or Google. The French government insists on calling it French, not European, which tells you something about the fragmentation. Europe doesn't have a unified strategy or a unified market. Each country is trying to protect its own interests.
So the question is really whether these hyperscalers can make money?
Exactly. They're spending 400 billion dollars a year on infrastructure. OpenAI needs to hit 200 billion in annual revenue by 2030 just to justify what it's already spent. That's not profit—that's just breaking even on the cash burn. If they can't get there, some of them will collapse.
And if they do collapse?
Then the entire industry recalibrates. Maybe the market doesn't need as many AI companies as we think. Maybe the winners consolidate and the rest disappear. Europe's absence from this race means it won't be a European company doing the consolidating.