Countries such as Brazil have become hot spots for data centers needed for artificial intelligence (AI), and the use of AI itself could help drive energy expansion by overcoming labor and other restrictions, the industry forecasts.
"AI is the great leveler right now," global energy software provider Quorum's chief product and technology officer Radhika Krishnan said at the CERAWeek by S&P Global energy conference in Houston, Texas. "Developing countries are leaning in to it."
Quorum's work in Latin America has included in Colombia, Brazil and Argentina.
In Brazil, state-controlled Petrobras has used AI to dramatically cut the time needed to process and analyze seismic data, chief engineering, technology and innovation office Renata Baruzzi said.
It has also launched the ChatPetrobras AI query tool that lets employees tap knowledge organization-wide, including seismic data. Employees can access some geological data 50pc faster than before AI, she said.
Petrobras first began integrating AI in non-core areas such as predictive maintenance, and then scaling, she added.
But AI will eventually move into core areas, with unmanned offshore rigs possible within 10 years, service provider SLB's president of digital Rakesh Jaggi forecast.
Legacy and newer companies will face different challenges, speakers said.
Larger companies can find 30-40pc improvements in some workflows with bolting on AI to current processes, but smaller companies should "ask what you could do if you start with AI from the ground up", Jaggi said.
Venezuelan opposition leader Maria Corina Machado this week said that while Venezuela's energy industry will require almost rebuilding, it would allow such a greenfield opportunity by "starting from scratch" in the age of AI if investment conditions improve.
Transition carryover
Markets have started to focus on whether AI is actually creating promised savings, but Baruzzi still sees AI-related investment as "good" costs rather than unproductive ones. The savings will eventually emerge, and help more than only traditional energy.
"AI savings will create funds we need for the energy transition," Baruzzi added. "This is not about choosing between transition and conventional energies ... but combining them."
Quorum's Krishnan noted that internal processes for both renewable and conventional energies can be similar in terms of suitability for AI. One successful use case they see in conventional energy is using AI in production allocation — determining how to divide up production or revenue from energy sales between governments and outside producers. The process has long been manually intensive, including evaluating data from everything from advanced sensors on physical field equipment to accounting records.
The same type of AI process could be adapted for deploying renewable energy, she said. This could feed data centers needed to support AI deployment.
Latin America's data center market is projected to double its value to as much as $10bn by 2029 from $5bn-6bn in 2023, with more than 37pc of those in Brazil, the UN Development Programme said.
Next steps
Countries now need to focus on developing stable regulations for AI, so that providers can help ensure data security, safety and privacy, Krishnan said. Both Colombia and Brazil have moved to the front in Latin America in these efforts, she noted.
Labor issues will need to be confronted in Latin America, where many countries face higher unemployment and sometimes historically inflated employee rolls at national oil companies.
Speakers saw fears of a labor crises as exaggerated, as "humans are still required for judgement and exception management", SLB's Jaggi said.
Populations have already peaked in roughly 25pc of the world's countries, according to UN projections, and fewer jobs may be needed in some places to meet growing energy demand.
But skill levels will need to be higher, and companies must get ahead of the curve in training, speakers noted. Petrobras has worked AI training into its internal Petrobras University education program.
Still, Baruzzi warned in the race to AI against "overestimating automation, and underestimating people and connections".
