Nvidia is building 100 “artificial intelligence factories”

A NVIDIA wants to reshape the concept of data centers with the creation of AI Factory: Structures designed to develop and operate large -scale artificial intelligence systems. Today, about 100 of these facilities They are under construction in various regions of the world, signaling a profound transformation in digital infrastructure.

These specialized centers will be aimed at training and execution of AI models, with the promise of dealing with multiple applications and architectures simultaneously. For that, they combine CPUS, GPUS, Networks and Advanced Softwarewith Nvidia’s own Cuda ecosystem at the center of the strategy.

According to Jensen Huang, CEO of the company, it is a long-term vision:

IA infrastructure will cover the planet, as well as the internet infrastructure. We are investing hundreds of billions of dollars in an AI infrastructure construction that will take five decades

Global partnerships and expansion

A NVIDIA It is mobilizing an international partnership network to implement this plan. In the United States, announced an investment of up to US $ 500 billion To build AI -dedicated supercomputers, with new centers planned in Texas and Arizona – projects that have companies like Foxconn e Wistron

In the Middle East, the company signed an agreement with the Humanfrom Saudi Arabia, to provide 18 mil GPUs Blackwellhelping to structure the local AI ecosystem. Also participates in the project Stargate UAEin Abu Dhabi, who intends to erect one of the largest AI centers on the planet.

In addition, the NVIDIA keep agreements with TSMC, Gigabyte, Asrock, Rack, Asus, Pegatron, Supermicro e Winstronamong others, to accelerate the implementation of AI infrastructure in both cloud and on on-premises.

Innovation and efficiency in the project center

AI factories are being equipped with state -of -the -art systems such as GB200 e GB300which are part of GPUs and CPUs on high performance platforms.

The company also invests heavily in energy efficiencyplanning to replace evaporative cooling systems with closed circuitswhich consume less energy and allow to operate with greater computational density.

The next generation of architectures of the company, named after Rubinintend to double the performance of current systems and reinforce the position of the NVIDIA No hardware market for IA.

Performance as a business model: the new logic of AI factories

To the basic antiga from NVIDIAthe gamer market, no longer occupies the center of attention, although it has sustained the company for more than three decades. In this segment, logic was clear: hardware needed to justify its performance with noticeable visual gains, such as new lighting effects or higher resolution. Now, with AI Factorythe game has changed.

“In a factory, performance is cost, and performance is recipe,” Summed Jensen Huang during his presentation on Computex. With each new generation of hardware, energy efficiency and processing capacity increaseallowing data centers to generate more results with lower energy consumption.

Huang’s reasoning is simple: If the performance by Watt is four times higherthe AI ​​factory operator can quadruple the revenue or drastically reduce costs. This is the essence of the new sales cycle that NVIDIA It wants to feed: a continuous flow of hardware and software updates, with each generation promising concrete gains for customer business.

The company also bets on long -term effect of your software ecosystemwith the platform CUDA serving as a competitive differential. The argument is that even old GPUS continue to improve with software updates, offering customers a sustained evolution in timebesides the incentive to migrate to the latest generations.

If we present a new generation, customer revenue can grow, and costs may fall. In this model, Customers come to depend on Nvidia’s technological cadencevery similar to what happened in the past with cryptocurrency mining, a market that strongly boosted GPUS sales in previous cycles.

But this race for performance has a price. Energy consumption and cooling systems become important bottlenecks. Therefore, the NVIDIA is already promoting the transition to Closed circuit cooling systemswhich promise superior energy efficiency and greater economic viability for large -scale operations.

Behind the view of AI factories, there is a recurrent monetization model: The more powerful the new generation of GPUS, the greater the competitive pressure for the IA companies to update their infrastructures – consolidating Dependence on Nvidia as a supplier of this ecosystem.

The new monetization cycle

As we mentioned above, the AI ​​factories business model resembles (partly) the logic of cryptocurrency mining, only applied to the corporate market. The idea is for companies to transform gross computational capacity in revenueusing the GPUs and systems of NVIDIA To train and operate AI models more efficiently.

As Jensen Huang explained during Computex:

In a factory, performance means cost, and performance means revenue. If we can offer four times more performance by Watt, customers can increase revenue and reduce the costs of their data centers

In addition to hardware, the company bets on the longevity of its software platform. With the constant evolution of the ecosystem CUDAa NVIDIA It seeks to ensure that their systems remain relevant for years, offering not only more power, but also optimizations that prolong the life of customer investments.

Also read:

AI’s future is still uncertain, but Nvidia bets high

Although the project of the AI ​​factories of NVIDIA Be grand, it is not risk free. The sector still faces challenges such as high operating costs, high energy consumption and the need for constant technological update. Furthermore, International competition grows rapidlywith Chinese companies offering lower cost solutions.

Still, the company of the beautiful leather jackets seems willing to lead this race. By turning their partners into customers of a complete ecosystem – that goes from silicon to software tools – the company bets that its AI factories will be a new revenue engine for the next decades.

If the future of AI will justify such a bet, it is still unknown. But the NVIDIA You have already put your chips on the table, and want to lead this game as far as you can.

Source: Tom’s Hardware