Middle East heavyweights eye tech firms for gains as DeepSeek model boosts AI investment prospects

Vijay Valecha, Arabian Business, January 30, 2025

Middle Eastern investors, including institutional investors and large family offices, are now betting on companies like Salesforce, ServiceNow, Adobe, AMD, and Intel to make windfall gains.

They expect to profit from the rise of low-cost, powerful large language models by DeepSeek, which will shift complex AI operations from data centres to mobile devices and desktop experts told Arabian Business.

Apple is another big stock Middle East investors are watching out as iPhone has demonstrated to successfully run models with a size of 3 billion parameters on local device as against the smaller, distilled R1 models of DeepSeek, which requires around 1.5 billion to 70 billion parameters.

“Investors [in the Middle East] are now looking at a scenario that should the DeepSeek model be able to operate with such a lower cost dynamic, top AI service names could provide a major opportunity for them,” Vijay Valecha , Chief Investment Officer at Dubai-based Century Financial, told Arabian Business.

“In this regard, some of the top names on the investors’ radar include Salesforce, Adobe, MongoDB, Okta and ServiceNow,” he revealed.

Valecha said for investors, one other opportunity could also arise from having exposure to the top mobile chip processors and edge device providers.

AMD, Qualcomm, Intel and even Apple are the top picks in this sector, he said.

“The entire narrative that could drive these stocks is the fact that the cost-efficient DeepSeek will now allow large language models to run locally on computers/mobiles etc,” the Century Financial investment chief said.

Low-cost AI shift

Sector and market experts said AI as a service will now move from high-cost and complex data centres to edge devices on our hands, bringing a revolutionary shift in the mainstream adoption of this fast-emerging new tech tool.

By drastically reducing the cost of deploying advanced AI capabilities, the DeepSeek-like models will accelerate the adoption of AI in real-world applications – from smart cities, and retail to healthcare and beyond, they said.

Andreas Hassellöf, CEO of Ombori, a global company operating at the intersection of technology and innovation, said this efficiency is particularly crucial in the current landscape, where recent updates to US export controls on certain AI hardware have created tiers of access for different countries.

“For instance, the UAE, classified as a Tier 2 country, faces limitations in accessing critical hardware resources. However, with technologies like DeepSeek, fewer GPUs (Graphics Processing Unit) are required to achieve the same – or even better – results, making the need for advanced hardware less pressing,” Hassellöf told Arabian Business.

The Ombori top executive said the advent of low-cost, super-powered models by companies like DeepSeek can open up to a much broader range of countries, encouraging collaboration across borders and creating a more globally interconnected AI ecosystem.

“The fact that DeepSeek is open source will further amplify its impact, enabling developers worldwide to build on this foundation and drive innovation at an unprecedented pace,” he said.

DeepSeek’s market impact

Valecha said there indeed seems to be a lot of investors in the Middle East shifting their focus to other US service and software names.

This is in anticipation of the lower cost dynamic newly unveiled LLM models by DeepSeek – and expectations of more such models coming up going forward – could be a boost to companies in the AI services sector, providing a major opportunity for the investors, he said.

The Century Financial investment chief also said among the top names investors should watch out for is Apple, as contrary to massive data centre spending by the rest of big tech, the iPhone maker has instead been focusing on how powerful its model phones can allow to run the AI model directly.

“In this context, DeepSeek is instead a blessing in disguise for Apple. The latest released R1 version model required around 1.5 billion to 70 billion parameters, as opposed to the original 671 billion.

“iPhone, on the other hand, has demonstrated to successfully run models with size of 3 billion parameters on local devices,” he said.

The Ombori chief executive said while the Wall Street reaction [to the news breakout on DeepSeek] has been exaggerated, history tells us that reductions in cost and increased efficiency often lead to greater demand over time – what economists call Jevons paradox.

“This is a pivotal moment for AI that will unlock new opportunities, foster global collaboration, and ultimately improve the everyday lives of people around the world,” Hassellöf said

Source

Arabian Business