This package also includes: Data centres meet upbeat growth

Artificial intelligence (AI) is a potential enabler for the economic diversification programmes of the GCC’s hydrocarbons-exporting states. 

The UAE launched an open-source large-language model (LLM) last year. Falcon 40B, shortly followed by Falcon 180B, cemented the reputation of the Abu Dhabi government-funded Technology Innovation Institute as a major player in generative AI.

With 180 billion parameters and trained on 3.5 trillion tokens, Falcon 180B soared to the top of the Hugging Face Leaderboard, a benchmark for pre-trained LLMs. Falcon 180B outperformed competitors such as Meta’s Llama 2 in areas including reasoning, coding, proficiency and knowledge tests.

The launch of Falcon followed cumulative investments in research, talent acquisition and digital infrastructure. In recent years, Abu Dhabi has formed government-attached agencies and commercial entities backed by its sovereign wealth funds to focus on AI.

One such company is G42, which has partnered with the US’ OpenAI to develop sector-focused generative AI models, and with Microsoft to run applications on Azure and undertake AI skilling initiatives in the UAE and beyond.

Global AI hubs

The UAE aims to become a world-leading AI hub alongside the US and China, but the country will have to tread carefully when choosing partners to avoid geopolitical complications involving its most important security ally and its largest energy client.

Riyadh seems determined to give Abu Dhabi a run for its AI money. The GCC region’s two largest states have placed
separate multimillion-dollar orders for graphics processing units – powerful chips designed for training AI – from top US supplier Nvidia.

They have also formed AI-focused investment vehicles with a view to maximising investments and returns from AI ventures at home and abroad. Abu Dhabi formed MGX, which aims to build $100bn in assets under management within a few years, while Saudi Arabia’s Public Investment Fund formed a $100bn platform to transform the kingdom into a semiconductor and electronics hub, with AI playing a central role in the plan.

In May this year, the Saudi Data & Artificial Intelligence Authority and New York-based technology company IBM launched an open-source Arabic LLM called Allam on IBM’s Watsonx AI and data platform.

With AI promising to be a $1tn market by 2030, it offers attractive opportunities

Computer power

A potential issue facing the determined push for AI leadership is that AI requires enormous computational power and energy, in addition to vast capital and talent.

A recent article published by the World Economic Forum (WEF) suggests that the computational power required to sustain the rise of AI doubles approximately every 100 days.

Related read: Global AI market to top $1tn in 2030

“The energy required to run AI tasks is already accelerating with an annual growth rate between 26% and 36%. This means by 2028, AI could be using more power than the entire country of Iceland used in 2021,” the WEF article says.

The AI lifecycle impacts the environment in two stages. First is the training phase, when the models learn and develop by digesting vast amounts of data; and second is the inference phase, when they solve real-world problems.

At present, the environmental footprint is split, with training responsible for about 20% and inference taking up 80%.

“As AI models gain traction across diverse sectors, the need for inference and its environmental footprint will escalate,” the WEF warns.

A peer-reviewed analysis in the science journal Joule says that a continuation of the current trends in AI capacity and adoption will likely result in Nvidia shipping 1.5 million AI server units a year by 2027.

When running at full capacity, these servers are expected to consume at least 85.4 terawatt-hours of electricity annually, which is equivalent to 100GW of installed capacity in the next three years.

Data centres, which make up the main AI digital infrastructure, already account for about 1%-1.5% of global electricity use.

In a hypothetical scenario in which everyone shifts to AI for mundane tasks such as performing searches on Google, every data centre would effectively experience a 10-fold increase in energy consumption, according to Alex De Vries, a data scientist at the Central Bank of the Netherlands, which conducted the analysis published by Joule.

As a result, the hydrocarbons-exporting and energy-transitioning GCC states – particularly the UAE and Saudi Arabia – appear to be a natural fit for AI, due to the presence of abundant and cheap fossil-fuel or renewable-energy resources, and the need to diversify their revenue sources away from oil. With AI promising to be a $1tn market by 2030, it offers attractive opportunities.

According to a Dubai-based senior executive with a global infrastructure investor, each country and company will eventually need to consider what part they can play in the AI value chain. 

Since Nvidia seems to have captured the microprocessor space, the other areas of opportunity are in developing computing power, algorithms and implementation. “Both Saudi Arabia and the UAE have the theoretical capability to grow into the computing power and implementation spaces, which require computing capacity through data centres and medium-skilled manpower to deploy, migrate, train and maintain [AI],” the executive says.

Greening AI

Policy adjustments could be needed to support such advances, especially when it comes to minimising AI’s carbon footprint, even as it enables the curbing of those in other sectors – including the power sector.

In addition to the vast computing and wattage requirements of AI, the region’s arid weather and very hot summer temperatures mean that regional data centres have greater cooling requirements.

To address this, the Dubai state utility has started to build a solar-powered data centre, which is understood to be the first of its kind in the world.

Saudi Arabia, which aims to have 58.7GW of renewable energy installed capacity by 2030 – accounting for about 50% of its electricity production mix – could follow a similar model. 

Abu Dhabi’s quantum computer project, in partnership with researchers at Spain’s Qilimanjaro Quantum Tech, is under way.

Unlike a classic supercomputer that operates on binary states, a quantum computer uses quantum mechanics phenomena including superposition and entanglement to generate and manipulate subatomic particles such as electrons or photons, or qubits. 

This allows greater processing powers that can enable the performance of complex calculations that would take much longer to be solved, consuming less power than a supercomputer.

The growing electricity surplus in Abu Dhabi, as all four reactors at the Barakah nuclear power plant come onstream this year, could also be allocated to data centres and AI applications.

In addition, Abu Dhabi’s plan to start procuring phase two of its Barakah nuclear energy plant may not only boost energy exports, but could also create sufficient margins to accommodate future AI computing demand.

Related read: Nuclear power will help region achieve AI ambitions

“I don’t know if that means only nuclear power can solve the demand, but it certainly is a good option and carries some strategic advantage as well,” says Karen Young, senior research scholar at Columbia University’s Centre on Global Energy Policy.

While AI needs a significant amount of electricity for computations, there should be savings through productivity increases 

Efficiency gains

While it is difficult to accurately quantify and forecast AI’s overall carbon emissions, a holistic view of its overall environmental impact is required.

In theory, while AI itself needs a significant amount of electricity for computations, there should be savings through productivity increases. “Will people need to go to the office less often, and how about the improved performance of machines?” asks the Dubai-based infrastructure investor.

However, it is also important not to overstate AI’s potential benefits to the region’s economies. While AI could be a major driver of economic diversification, Young has yet to be convinced that it will significantly boost the GCC’s GDP growth.

Job creation is a vital element of economic diversification, she tells MEED, but AI is often used to replace roles in the service sector and lower-skilled opportunities, such as those in the retail banking sector. This could impact efforts under way in several GCC states to boost employment among citizens, such as the Saudi Nationalisation Programme and the UAE’s Emiratisation drive.

On the upside, however, AI can be very good at improving efficiencies in the oil and gas industry and the power sector, and at boosting productivity.

The need of the hour appears to be establishing a clear path towards efficient AI deployment, despite the fact that the results of the technology’s full-fledged implementation remain hard to ascertain. 

“The UAE is doing a lot to attract skilled people to provide more value-added services, but that is an organic process and needs a more vibrant ecosystem of education institutions – and companies establishing more than just sales offices – to be truly called a hub,” the infrastructure investor tells MEED. “Saudi Arabia is still a bit far from that.”

Data centres meet upbeat growth