How can Europe compete with AI semiconductor giant Nvidia if the chips are broken?

Nvidia, which was worth $3.01 trillion (€2.76 trillion) on Thursday, just overtook Apple as the world’s second most valuable company.

The company’s share price rose by more than 60 points, or five percent, to more than $1,224 (€1,125) per share on Wednesday. Nvidia is only the third company in the world, after Microsoft and Apple, to be valued at a trillion dollars.

Nvidia has been designing semiconductors or microchips since 1993; first for video game consoles and now for the generative artificial intelligence (GenAI) companies.

The most important question now is: can European companies compete with the rapid rise of Nvidia?

Who’s who in the booming AI microchip industry

Microchips are the engines that power the world’s largest technology companies and their AI divisions.

There are five different types of companies that make up the semiconductor market, according to Cem Dilmegani, principal analyst at consultancy AIMultiple.

At the core of the industry are those who make the machines needed for the mass production of microchips, including the Dutch company Advanced Semiconductor Materials Lithography (ASML), which is worth 361.67 billion euros.

Then, foundry companies like Taiwan’s Semiconductor Manufacturing Company (TSMC) use these machines to manufacture the chips that other design companies, like Nvidia, will refine for AI models. The largest global players at this level of the semiconductor industry are in China, South Korea and Taiwan.

Next come design companies like Nvidia, which, according to the company’s website, are designing “the most advanced chips, systems and software for the AI ​​factories of the future.”

Those developed chips are then bought by the big tech companies – from France’s Mistral AI to major tech giants Amazon, Google, Meta and Microsoft – to program the large language models (LLMs) that power AI.

Nvidia particularly specializes in graphics processing unit (GPU) chips that display higher quality images than the central processing units (CPUs) that their US competitors Intel and Advanced Micro Devices (AMD) are known for.

These GPUs can perform calculations in ways that CPU chips cannot, meaning they can better handle the kind of work AI companies need.

For companies that can get their hands on Nvidia’s chips, Dilmegani says this also means they will need fewer staff to train their languages.

“It’s going to take a lot more work and you’re going to have to put more engineers on it,” he said. “You may also need more calculation time or the [language] model may work slowly”.

‘We are creating an entire ecosystem’

Nvidia Vice President EMEA Serge Palaric joined the company in 2004 when they were still designing GPU chips for video consoles like the Xbox.

Then in 2006, Nvidia launched Compute Unified Device Architecture (CUDA), their own programming language for their GPU chips. It is now deep learning software that companies can buy in addition to the developed chips.

Palaric attributes his company’s rise in the semiconductor market to both the hardware it designed and the software that simplifies the work of these big tech companies.

“We don’t sell chips, we sell software,” Palaric said. “We are creating a complete ecosystem to address and support these companies that want to make the transition [generative] AI”.

“ChatGPT was built for over 10,000 GPUs, so we knew what was coming. We’re always looking at what’s next, how we do that, [and] how do we help our partners with this.”

In 2012, Nvidia and the University of Toronto discovered that GPUs make deep learning – the type of learning that powers today’s AI – more accurate. From that moment on, the race began within Nvidia to make way for the eventual AI boom.

So when OpenAI’s ChatGPT launched in 2022, Palaric said Nvidia wasn’t surprised at all and was ready to rise to the occasion.

“ChatGPT was built for over 10,000 GPUs, so we knew what was coming,” said Palaric. “We always look at what the next step is, how we do it, [and] how do we help our partners with this.”

AMD and Intel are both coming on June 2 announced new AI processors to regain some market share from Nvidia, which controls about 80 percent of the chip design market.

‘Nvidia is in a class of its own’

Based on market share alone, ASML is the largest player for Europe in the AI ​​semiconductor industry, according to Dilmegani.

But because Nvidia develops GPU chips and ASML supplies the machines to those who produce them, it’s not fair to equate ASML and Nvidia’s dominance in their respective parts of the industry, said Michelle Brophy, director of research at AlphaSense, a AI expert. -powered market platform.

“None of the Europeans are close [Nvidia’s market cap] not at all,” Brophy told Euronews Next. “They will find their niche areas, but Nvidia is in a league of its own.”

In theory, Dilmegani said ASML could scale vertically to start designing its own microchip technology. However, it would be “strange” and “very complex” because it would then compete with other chip makers that are part of its customer base.

Where ASML is competitive, according to Brophy, is in helping TSMC improve the production of its chips. The Taiwanese manufacturer has publicly said it wants to start making 2-nanometer chips in 2025, offering chips with 10 to 15 percent faster processing times than the advanced 3-nanometer chips currently on the market.

Brophy said it was the “only way.” [TSMC] get there” is to work with ASML.

Elsewhere, two European companies are already designing semiconductor chips: Germany’s Infineon and French-Italian STMicroelectronics, Brophy said.

Both companies work specifically on CPU microchips for the automotive industry and have not publicly indicated that they will compete with Nvidia.

Brophy said it wouldn’t be wise for them to do that.

“They would have a lot of work ahead of them,” Brophy said. “If their next step were toward AI, it would probably be in the area of ​​processing.”

Euronews Next contacted both Infineon and STMicroelectronics, but did not receive a response from either company at the time of publication.

‘We don’t take many risks’

According to Dilmegani and Brophy, if European companies or start-ups want to become competitive in the sector, they must develop a niche in another part of the semiconductor market, such as ASML.

Alternatively, they should develop the necessary relationships so that leading chip design companies such as Nvidia or AMD consider moving their operations to the continent.

Brophy pointed to a recent deal with Intel in Germany as an example of areas where Europe could invest more.

In 2023, Intel and Germany signed a €30 billion deal to establish a microchip factory in Magdeburg, but the deal is still awaiting approval from the European Union, according to Politics.

“Start-ups need to think about where technology is going and make bold bets and create machines that are different from other processors. But you need an ecosystem in which these companies can emerge… and in the EU we don’t accept that [many] risks in that dimension”.

An EU spokesperson told Euronews Next that the bloc cannot comment on the progress of this deal.

The goal for companies like Intel to invest in European chip factories is to move away from dependence on TSMC, the largest chip maker, due to political tensions between the West and China, Brophy said.

Nvidia’s Palaric said the company will continue to rely on TSMC for their chips.

What Nvidia is doing in Europe is investing in AI startups such as France’s Mistral AI, Britain’s Synthesia and Germany’s Northern Data Group’s AI Accelerator.

Their goal, according to Palaric, is to provide companies with the technology they need to develop their own languages.

Startups looking to design chips like Nvidia will likely need at least a decade to grow their presence enough to be competitive, according to Dilmegani.

“Startups need to think about where technology is going and make bold bets and create machines that are different from other processors,” Dilmegani said.

‘But you need an ecosystem in which these companies can emerge… and in the EU we don’t accept that [many] risks in that dimension”.

Leave a Comment