Foxconn and Nvidia choose Mexico to build their superchip plant

Mexico will manufacture chips for Nvidia

“We are building the largest GB200 production plant on the planet,” company vice president Benjamin Ting said at the event.

The GB200 chip, also known as Blackwell, is Nvidia’s most advanced Artificial Intelligence server. The company has described it as “the engine” of a “new industrial revolution,” enabling organizations around the world to create and run generative AI at 25 times less cost and energy consumption than its predecessor.

According to Jensen Huang, CEO of Nvidia, this component will help unlock advances in data processing, engineering simulation, electronic design automation, computer-aided drug design, quantum computing and generative AI. Organizations planning to adopt Blackwell include Amazon, Google, Meta, Microsoft, and OpenAI.

With this new chip, Nvidia seeks to amplify its leadership in the AI ​​industry, where it has had a predominant performance thanks to the H100 AI GPU, which turned it into a multi-million dollar company. According to Huang’s statements during his annual event last March, Blackwell exceeds the performance of an H100 by seven times.

According to statements by Young Liu reported by the Financial Times, the capacity of this new plant will be “very enormous”, in addition to surpassing other factories of its nature around the world.

So far the location of where this plant will be found is not known. Various reports affirm that it would be in Guadalajara. Last year it invested 453 million Mexican pesos to buy a 421,600 square meter land in Jalisco.

It is also important to highlight that Foxconn already has a presence in the state of Chihuahua, where it has invested more than 500 million dollars and already has a headquarters in Tijuana, Baja California, in addition to ensuring that it employs more than 5,000 people in the country.

On the other hand, Nvidia’s decision to work with Foxconn is due to the fact that, unlike its competitors, such as Intel, it does not manufacture its own chips, but rather hires a third party to carry out that process.

As part of this dynamic, the firm collaborated with Foxconn last year, with the aim of creating “AI factories”, that is, centers specialized in the creation of cutting-edge components to promote its most advanced technologies.

Blackwell is also part of a platform of six technologies, which together enable the training of real-time large language models (LLM) for models that scale 10 billion parameters and which will be available by the end of the year .

It should be remembered that a couple of months ago, Nvidia announced the design flaw of this chip, which caused a delay in its launch, which also coincided with Black Monday in August.

“The demand for Blackwell is exhausted,” Marcio Aguiar, director of Nvidia’s Enterprise division for Latin America, told Expansión, who highlighted that within the company they expect shipments to begin in the fourth quarter of the year.