Mark Zuckerberg wants to build gigantic data centers that could train the first artificial superintelligence on the planet. The first, called “Prometheus”, will be ready in 2026. The dimensions? As a neighborhood in Manhattan. Consumption? As a whole city.
Meta
“We build several titanic clusters. Only one of them covers a significant part of the surface of the Manhattan”, Zuckerberg wrote on Threads.
The AI has no limits. Only resources.
Massive investments in dedicated infrastructures signal an acceleration of the development of artificial intelligence on an unprecedented scale, warns Alexandru Dan, CEO TVL Tech and teacher at the Academy of Economic Studies (ASE) in Bucharest. Basically, he believes, we are approaching the moment when you will have almost unlimited resources for training and operation, which will lead to increasingly sophisticated models, capable of solving complex problems, collaborating with people at new levels and fundamentally change whole industries.
“This evolution could bring both huge benefits in medicine, education, science, as well as hard to anticipate risks if it is not managed responsibly. On the labor market, the advance will bring both large-scale automation, as well as the profound transformation of many existing roles. Repetitive or data analysis will disap Critical thinking, empathy, creativity and adaptation will become even more valuable. explains the specialist.
Global race for digital supremacy
In his opinion, the concentration of these huge resources of AI at the level of private companies in the US or from several other regions creates a global differentiation. “The countries and companies that control such digital infrastructures will have significant economic, political and military advantages, which can lead to power imbalances and a new global race for technological supremacy. Open access, international governance and collaboration become essential to avoid digital fragmentation of the world.”, Says the TVL Tech CEO.
When such a technology is developed by a private company, the actual control over AI belongs, first of all, to the shareholders and management of that company. Without a clear international framework for regulatory and without transparency mechanisms, there is a risk that decisions with a global impact will be made according to commercial or group interests. Therefore, it is urgent to involve the involvement of governments, international institutions and civil society to define clear rules on safety, ethics and responsibility in using superintelligence, warns Alexandru Dan.
In fact, he says, the huge energy consumption of these centers has a considerable ecological impact. “If the energy used does not come from renewable sources, carbon emissions and pressure on natural resources can cancel many of the benefits you promise. The industry has the responsibility to invest in green technologies and energy efficiency, and regulations must encourage or impose the use of sustainable sources for such infrastructures,”,, explains the teacher.
“The future will belong to those who know how to collaborate with AI”
At the same time, Alexandru Dan argues that as you get more and more efficient, human creativity will retain its essential value, but will have to evolve. “You can automate certain forms of creation, but it cannot fully replace the originality, intuition, cultural sensitivity and authenticity that only man can bring. The future will belong to those who know how to collaborate with AI, to use digital tools to amplify their ideas and to create meaning beyond what can generate an algorithm.”he adds.
For his part, social media expert Cristian China Birta says: “The only real obstacle to the path, at this moment, is the processing power. So. This prevents it from becoming something truly superintelligent or truly global. And this power comes from two directions: on the one hand, you have the processors, which, no matter how fast, have a limit. Possibilities.
AI is barely in childhood
According to him, these “cities” may not even be necessary, if the promises related to the quantum computer are confirmed. “Google seems already in an advanced phase. Then, instead of having large servers for square kilometers, we will have much more compact infrastructures, but with an incomparably greater processing power. What will happen from that moment onwards … it is hard to anticipate. Jobs are threatened and how many behavioral models are changing.
This is why we have to be open, to keep the radar on all the time, to be careful about everything in this area and to learn how we integrate the AI into our lives and business. And, just as important, as we protect ourselves. Because the truth is that no one knows, not even those who have created these systems, where this madness will be reached. And this, in itself, is a great danger ”concludes Cristian China Birta.
The mega-project meta, in short
According to Reuters, Meta has announced that she will invest hundreds of billions in several Database centers, as part of an aggressive superintelligence development strategy. The first center, Prometheus, will become operational in 2026. Another, Hyperion, will reach 5 gigawates in the coming years.
Meta has recently reorganized her efforts under the superintelligence division Labs, led by Alexandr Wang and Nat Friedman, after an investment of $ 14.3 billion in AI scales.
The total capital expenses estimated for 2025: between 64 and $ 72 billion. The stake: the supremacy in front of Openai and Google.
