AI startups held back by lack of GPU clusters in Sri Lanka

• Experts discuss Sri Lanka’s AI challenge and opportunities
• Data readiness, governance and compliance seen as first step for enterprises
• Banking, GovTech and agriculture flagged as key opportunity areas
• Education and professional bodies urged to integrate AI skills into training

Information Communication Technology Agency of Sri Lanka (ICTA) Board Member and Microimage Founder and CEO Harsha Purasinghe at a webinar last Saturday (27) on ‘How AI Is Reshaping Jobs, Skills and the Workplace’ said Artificial Intelligence (AI) is reshaping skills, workplaces, and business models, but Sri Lanka must urgently address gaps in computing capacity or Graphics Processing Unit (GPU) clusters, infrastructure, data readiness, and education to capture its potential.

Purasinghe noted that AI entrepreneurs face high entry barriers due to the lack of local computing capacity.

The webinar was organised by the Daily FT, Association of Chartered Certified Accountants (ACCA), International Chamber of Commerce Sri Lanka, Women in Management, the Sri Lanka Institute of Directors, MBA Alumni Association of the University of Colombo, Employers’ Federation of Ceylon, American Institute of Certified Public Accountants (AICPA), Chartered Institute of Management Accountants (CIMA), and BIMT Campus.

Sri Lanka does not yet have GPU clusters within its data centres, forcing startups to rely on public hyperscalers. A GPU, or Graphics Processing Unit, is a type of computer chip that can perform thousands of calculations simultaneously, making it essential for training AI models. While CPUs can process tasks sequentially, GPUs are required to run large volumes of data through models quickly and cost-effectively. 

“You can come up with fantastic ideas, but you need to train your models and experiment. For a startup, the biggest challenge is paying for compute,” Purasinghe said.

He pointed to the need for Government and industry collaboration to attract global hyperscalers and encourage local players to establish GPU clusters. Such infrastructure, he said, would also strengthen university research and help smaller startups build models more cost-effectively.

Purasinghe argued that Sri Lanka’s opportunity in AI lies not in scale but in cultivating high-quality talent. Unlike the outsourcing boom, which depended on large numbers of workers, AI requires small but skilled teams capable of building proprietary systems and intellectual property. 

“One of the key things we need to do as a country is up-skilling. Curricula have to change, and young people must be guided to make use of these tools. They will figure out how to build phenomenal systems,” he said.

On whether businesses and individuals can choose to opt out of AI, Purasinghe said access and affordability remain the decisive factors. Free tools offer entry points, but broader digital inclusion is critical. He cited the potential of AI to transform agriculture, where farmers could use tools to decide planting cycles, improve yields, and reduce costs.

The ICTA has piloted AI awareness among the public sector, starting with the Presidential Secretariat. The initiative has since expanded to cover more than 5,000 officials, introducing them to AI tools, prompting techniques, and workplace applications. “We found that training officials on AI tools had a massive impact on efficiency,” Purasinghe said.

Nimbus Cloud Lanka Managing Director Damith Hettihewa said the foundation for AI adoption lies in data consolidation and governance. “AI models are only as good as the data they are trained on. The first step is consolidating and organising data, establishing governance frameworks, and ensuring compliance and ethics.” 

Sri Lanka’s Personal Data Protection Act and the need to align with international regulations, particularly in Europe, already shape enterprise practices. 

Hettihewa argued that before worrying about computing power, companies can make progress by preparing data frameworks using existing CPU capacity or leveraging cloud credits, which lowers the initial entry barrier for startups.

NCINGA Chief Technology Officer Chathura Peiris said the opportunities for AI in Sri Lanka span multiple sectors, including banking, healthcare, education, agriculture, and Government services. 

He pointed to the rise of “AI-first” banking globally, where institutions use data-driven systems to deliver hyper-personalised services, cut operating costs, and accelerate customer engagement through digital-only platforms. 

“The foundation is data, which must be harnessed with scalable infrastructure and analytics layers. On top of that, hyper-personalised engagement through AI-powered customer platforms will transform how banks engage with clients,” Peiris said. 

He added that AI is also expected to play a central role in Sri Lanka’s GovTech agenda, especially in identity, payments, and data exchange.

ACCA Asia Pacific Policy and Insights Manager Dean Hezekiah said professional bodies also highlighted the role of education and credentialing in preparing the workforce. 

Accountancy qualifications like ACCA are being revised to integrate digital skills, with a new version of the program emphasising applied digital competencies. 

Universities and professional associations, Hezekiah said, should ensure at least baseline awareness of AI technologies among graduates.