A global artificial intelligence battle looks to be building between China and the US that could well benefit Middle East AI startups through lower development costs.

OpenAI CEO Sam Altman is planning a visit to the UAE, reportedly to raise funds with the investment group MGX, established in 2024 by the government of Abu Dhabi to channel investments into AI-driven technologies.

But the arrival of China’s DeepSeek, claimed to have been developed for only $6 million, means that OpenAI’s ChatGPT, which cost $100 million, now has a serious competitor, many observers say.

DeepSeek’s lower price could make it easier for Middle East startups, which have generally relied on more expensive models from the American companies OpenAI and Anthropic, to invest in specialised AI applications.

“The introduction of DeepSeek provides substantial cost reductions in AI-solution development,” says Yousef Khalili, Middle East and Africa CEO at Quant, a New York-based company specialising in data, AI and advanced analytics.

Sectors such as government services, healthcare and finance could benefit the most, as they require tailored AI services, Khalili says.

But to capitalise on this, companies in the region must move quickly in developing AI that is optimised for Arabic and Gulf dialects, invest in graphics processing infrastructure and local data centres, and balance the relationship between Chinese and Western AI companies to avoid dependence on any single provider, Khalili says.

Despite ambitious AI investments, the Middle East has yet to produce a model that is as affordable and scalable as DeepSeek. The reasons are structural, Khalili says.

“The Middle East started its AI investment initiative later than both China and the US, which have been investing heavily for over a decade,” Khalili says. This late start means the region lacks an established AI research base compared with global leaders.

Another critical factor is computing limitations. Training a cutting edge large language model (LLM) requires vast computational power, primarily driven by graphics processing unit (GPU) clusters, most of which are controlled by US and Chinese firms.

Without comparable AI infrastructure, Middle Eastern firms have focused on applied AI rather than foundational models.

“Middle Eastern AI companies prioritised developing AI for smart cities, fintech and logistics rather than competing in the LLM space,” Khalili says. 

LLMs are very large, deep-learning models that are pre-trained on vast amounts of data. GPUs, meanwhile, which were originally created for graphics rendering, have become increasingly important for AI, enabling training and deployment of complex AI models that were once impossible to imagine.

All this said, there are signs of change in the Middle East. The UAE’s Falcon and Saudi Arabia’s Allam models are signs of growing ambitions for sovereign AI.

DeepSeek’s launch could accelerate these efforts, putting pressure on the Middle East to develop its own cost effective models.

One of the biggest controversies surrounding DeepSeek is its relatively low cost. Some argue that an AI model that is more affordable must be inferior in quality. However, this assumption ignores important factors that contribute to cost.

“People who think DeepSeek is low quality because of its affordability fail to understand how AI costs operate,” says Khalili.

DeepSeek probably employs specialised techniques to cut costs without sacrificing performance, he says. It also benefits from China’s lower operational expenses relative to the US, and seeks to address practical applications rather than outperform ChatGPT-4 in every metric.

DeepSeek’s greater affordability is less about compromise and more about optimisation. Gregor Amon, an entrepreneur in Dubai and founder of the AI Expert Academy, says: “DeepSeek trained its model for $6 million, but that was done on a $200 million hardware cluster.” 

Its greater efficiency stems from necessity: China’s access to the latest chips from the American company Nvidia is restricted because of US sanctions, forcing it to innovate with available resources.

Implications for Middle East AI

Donald Trump’s return to office, and his pro-crypto, pro-tech stance, could shift global AI dynamics, potentially benefiting US AI companies.

For AI companies in the Gulf and wider Middle East, this raises concerns around balancing the relationship between competing US and Chinese technology partners.

Khalili says the key lies in interoperability. Rather than bet on a single model, companies in the Middle East should integrate multiple LLMs, such as OpenAI, Falcon and DeepSeek, to optimise performance and costs, he says.

Amon suggests that AI’s future will be less about foundational models and more about the applications: “People will just not care if the underlying model is DeepSeek, OpenAI or Gemini. They will focus on the tools and platforms built on top.”

This perspective offers an opening for Middle East AI firms. Rather than compete directly in LLM development, regional companies can lead in AI-application innovation, building industry specific solutions on top of any of several different models.

 

Read more