Race is on for AI chips as nations seek tech supremacy

Costfoto/NurPhoto/Reuters

• AI surge causing shortages at chipmaker Nvidia
• Saudi Arabia and UAE are buying up chips
• Rivalry over tech causing US-China tensions

The Gulf states are expected to invest heavily in powerful semiconductor chips in the years ahead to drive artificial intelligence forward.

However, US restrictions on exports of the world’s biggest player in the chips market are a potential stumbling block. 

AI technology is critical to developing global influence and expanding high-growth industries, and the high-performance chips – known as graphics processing units (GPUs) – process vast amounts of data much faster than traditional computer components.

• Investors turn to data centres in Gulf and Mena tech push
• UAE to play ‘central role’ in AI but needs greater regulation
• Nvidia to build Israel’s most powerful AI supercomputer

The computer power they provide makes them vital for building AI products such as ChatGPT, the large language model (LLM) that has become a household name.

US-based chipmaker Nvidia – the world’s most valuable manufacturer, worth nearly $1 trillion – dominates the market for GPUs, holding a 95 percent market share.

The AI surge has caused a huge shortage of Nvidia’s semiconductors, resulting in long waiting lists.

The company’s US rival AMD and other chip manufacturers have made significant strides in GPU technology, but Nvidia’s GPUs are preferred by many developers and organisations. 

Burgeoning demand

Both the UAE and Saudi Arabia had ordered chips from Nvidia, the UK’s Financial Times reported in August.

The kingdom has bought at least 3,000 of Nvidia’s H100 chips, costing around $40,000 each. The UAE has secured access to thousands of Nvidia chips, according to the newspaper. 

Earlier this year Abu Dhabi made Falcon 40B available for research and commercial use – Falcon 40B is a foundational LLM with 40 billion parameters (a parameter is the value used to configure a learning algorithm).

UAE technology consultant Carrington Malin expects two of the world’s fastest growing users of AI – the UAE and Saudi Arabia – to continue buying “thousands” of AI chips each year.

“With competition for AI chips getting fierce, it pays buyers to have a long-term forecast of required compute capacity,” Malin says.

“The biggest supercomputer order announced in the past year was from King Abdullah University of Science and Technology in Saudi Arabia,” he says.

“Its Shaheen III system alone will ultimately use 2,800 Nvidia Grace Hopper Superchips.”

Deepak Padmanabhan, partner at Dubai strategic advisory firm TenTen100, says the UAE is likely “augmenting its capacity” of GPUs.

Politicised industry

In August the UK became the latest country to join the race to acquire the powerful chips, with prime minister Rishi Sunak reportedly planning to spend up to £100 million ($121 million) on chips potentially from Nvidia, AMD and Intel, to help develop a national AI Research Resource.  

However, many countries do not have the vision or resources to commit to orders far in advance, according to Malin. 

What is more, the industry is starting to become heavily politicised. “The availability, pricing and geopolitics of access to AI chips may well start to look a lot like oil,” he says. 

“One only has to look at the US-China decoupling over the past five years.”

Competition over tech has caused tensions between the US and China, with Washington seeking to restrict China’s access to semiconductors.

Sales of Nvidia chips are already curtailed in China and Russia, but in August the Biden administration blocked sales of some of Nvidia’s most valuable products to the Middle East.

Nvidia announced the restrictions in a regulatory filing to the US Securities and Exchange Commission.

Nvidia/Handout via Reuters

The UK’s Daily Telegraph reported soon afterwards that the restrictions were thought to be an expansion of US efforts to undermine China’s AI ambitions, by preventing onward sale of the chip to China and limiting Middle East links to Chinese companies.

Earlier in August US tech giant Intel had called off a $5.4 billion acquisition of an Israeli chip manufacturer after Chinese regulators failed to sign off a big export deal.

The FT reported the previous week that Chinese tech giants had ordered $5 billion worth of Nvidia’s GPUs ahead of the Biden restriction on China’s access to such US exports. 

A super-competitor?

Meanwhile the Condor Galaxy 1 (CG1) supercomputer could give chipmakers such as Nvidia a run for their money.

Built by Silicon Valley-based Cerebras Systems with backing from UAE AI company G42, the CG1 can support up to 600 billion parameters and extend up to 100 trillion.

This makes it one of the largest AI supercomputers in production.

Andrew Jackson, CEO of Inception, a G42 subsidiary that developed a 13 billion-parameter Arabic and English LLM, tells AGBI the company is “doubling down on Cerebras” after using CG1.

“We are massive fans of Nvidia and we have been their largest customer in the region,” Jackson says.

“But we’re also very open to trying new things. We have run on other hardware and when we got access to CG1 we saw a dramatic speed-up in terms of our training time.”

The next frontier

Prashant Gulati, an angel investor in Dubai, likens the chipmaker race to “buying fighter jets”.

“Countries are picking up weaponry to fight the next frontier, which is AI,” Gulati says.

However, the intensity of the competition and the restrictions setback is unlikely to dampen the Gulf’s ambitions. 

“In the next generation of geopolitics, the size of the country may not be a measure for power. A country like the UAE with 10 million people could punch weight as big as a country far bigger in size.”