One figure suffices to measure the seismic shift: over $500 billion. Nvidia and SK Group have unveiled an exceptional initiative aimed at building artificial intelligence infrastructures and securing the next generation of memory essential for the most powerful systems. Behind this announcement lies not merely a new contract between two tech giants but an attempt to redraw the global map of digital power.

Presented on July 24, 2026, the project combines Nvidia with South Korea’s second-largest conglomerate, which controls SK Telecom and SK Hynix, among others. It is based on letters of intent covering the construction of “AI factories,” long-term supply of high-bandwidth memory, and the joint development of new technologies. According to Nvidia and reports from Reuters, the first facility is expected to come online as early as 2027.

A 2-Gigawatt AI Factory: The Figure That Changes the Scale

The visible heart of the initiative will be an infrastructure capable of reaching 2 gigawatts, developed by SK Telecom in South Korea. Such power places the project in the category of major industrial complexes, far beyond the traditional data center. This capacity is intended to meet the growing demand for computing for training models, autonomous agents, sovereign AI, and physical AI used in robotics or industry.

The architecture will rely on the Nvidia DSX platform and future Vera Rubin systems. These are expected to be powered by SK Hynix’s HBM4 memory. This combination is strategic: accelerators cannot deliver their full performance without memory capable of moving massive volumes of data at very high speeds. In the current race, memory has become almost as decisive as the processor itself.

Why Nvidia Wants to Lock in the Memory of the Future

Nvidia dominates the AI accelerator market but relies on a complex ecosystem to manufacture its systems. SK Hynix is one of the most significant players in HBM memory, a rare and difficult-to-produce component essential for AI servers. The agreement aims to secure Nvidia’s supply while co-developing solutions tailored to upcoming platforms.

The partnership is set to cover several applications: training large language models, agents capable of executing tasks, industrial simulation, and autonomous machines. It also gives SK Hynix visibility into the future needs of its primary client ecosystem. For both groups, the calculation is clear: it is better to design processors and memory together than to attempt to adjust them afterward.

South Korea Aims to Transition from Supplier to AI Powerhouse

The political and industrial implications of the announcement are considerable. South Korea already holds a central position in semiconductors, telecommunications, batteries, and electronics. With this new infrastructure, Seoul aims to go further: no longer just supplying components for the digital economy but hosting the computing capabilities that will produce the models and services of tomorrow.

Chey Tae-won, president of SK Group, encapsulated this ambition by stating that a country’s competitiveness now depends on the amount of “intelligence” it can produce. Jensen Huang, founder and CEO of Nvidia, believes that Korea possesses the networks, data centers, chip expertise, and industrial scale necessary to become a global AI powerhouse.

The message is directed squarely at the United States, China, Japan, Europe, and Gulf states, all of which are also ramping up investments in advanced computing. The battle is no longer just between laboratories or applications. It now involves electricity, land, networks, cooling systems, chips, and data sovereignty.

$500 Billion: Real Investment or Potential Business Volume?

The announced amount must be interpreted with precision. Nvidia speaks of a global initiative and partnership exceeding $500 billion, formalized at this stage by letters of intent. This does not necessarily mean that a single check for this amount will be immediately invested in one site. The figure may cover, over several years, infrastructure, equipment, memory, services, and the business volume generated by the cooperation.

This nuance does not diminish the project’s scale, but it serves as a reminder that the timeline, detailed financing, and final contracts will be crucial. Nvidia itself mentions risks related to economic conditions, manufacturing, energy, competition, and industrial execution. The first concrete step announced remains the commissioning of an initial AI factory in 2027.

A Colossal Energy and Environmental Challenge

A 2-gigawatt infrastructure immediately raises questions about its power supply. AI data centers require continuous electricity, reinforced networks, and significant cooling capacities. Their footprint will depend on the energy mix, equipment efficiency, and how the facilities are deployed.

Nvidia emphasizes an architecture aimed at reducing the energy cost per computing unit. However, the efficiency of a machine does not guarantee a decrease in total consumption if demand explodes faster than technical gains. For South Korea, transforming this announcement into a sustainable advantage will therefore require simultaneously solving the equation of chips, energy, and the electrical grid.

What This Alliance Could Change for Europe and France

The announcement should also prompt reflection in Europe. France has strengths in energy, research, mathematics, cloud computing, and AI models, but the continent remains dependent on foreign suppliers for accelerators and a large part of advanced memory. Closer integration between Nvidia and SK Group could exacerbate this dependency if the most powerful capabilities are primarily reserved for major partners capable of signing massive commitments.

Conversely, the construction of new capacities could increase the global supply of computing and create opportunities for European companies. French groups involved in energy, infrastructure, cooling, telecommunications, and industrial software could find a place in this new value chain. The central question will be access: who will be able to rent this power, at what price, and under what data governance rules?

The Signal That the Global Industry Can No Longer Ignore

This alliance confirms a major shift. Artificial intelligence is no longer merely a competition of models. It is becoming a heavy industry, organized around assets valued in the hundreds of billions: power plants, networks, semiconductor factories, specialized memory, and computing campuses.

If Nvidia and SK Group realize their ambitions, South Korea could establish itself as one of the nerve centers of this global infrastructure. But the real test will come after the announcement: the ability to finance, build, power, and operate these facilities on time. The figure of $500 billion captures attention. The battle, however, will be fought over every megawatt, every chip, and every HBM4 module actually delivered.

Sources

• Nvidia Newsroom, official statement from July 24, 2026.
• Reuters, July 24, 2026.
• Reuters via Investing.com, accessible version of the report.