AI deployment has ‘stalled’ for most Asia companies: report

[SINGAPORE] Organisations in Asia have hit roadblocks in transforming their operations and generating new revenue streams through artificial intelligence, a new research study by Singapore Technologies Telemedia Global Data Centres has found. Although about 90 per cent of organisations in the region have started to adopt AI, around 70 per cent struggle to scale AI pilots into production, stated Mind the Gap: Bridging the AI Infrastructure Readiness Divide on Wednesday (Apr 15). A total of 644 companies in various sectors such as banking and financial services, manufacturing, and energy and utilities were surveyed.Enterprises with 1,000 or more employees made up 43 per cent of the respondent companies. Twenty-nine per cent were companies with 250 to 999 employees, and 28 per cent were digital natives.The study found that across Asia, AI infrastructure readiness was “fundamentally stalled”. Only 16 per cent of firms have an operational AI strategy, robust infrastructure and governance to support its deployment. Additionally, only 1 per cent of organisations were found to have fully embedded AI in their businesses.It also showed that 56 per cent of companies were unable to measure AI production value and faced budget constraints. They also lacked the infrastructure to scale AI pilots into production. This has resulted in a deployment gap, which prevents clear return on investment and production value of AI from emerging. Companies will therefore be unable to unlock measurable benefits, said the study.Mature and emerging markets in AsiaThe study found that AI readiness in Asia was divided into two camps.The first is “mature markets”, which refers to those that are leading the AI adoption race. The second is “emerging markets”, who have yet to tap into their full potential in deploying AI.Both markets however, face similar challenges of infrastructure limitations and talent shortages. These hurdles can be overcome by partnering specialised providers, the study added. Singapore, Japan and South Korea are mature markets with strong operational and regulatory readiness. In particular, Singapore leads the region with its “strong public-private partnerships”, supported by “governance and talent pipelines” such as its National AI strategy 2.0 launched in 2023.However, the Republic faces land and energy capacity constraints to support growth in high-density AI training. This has since fuelled Singapore’s expansion beyond traditional data centre building sites. Emerging markets such as Malaysia, Indonesia, the Philippines and Vietnam have demonstrated their potential in furthering AI deployment, which requires “growing technical expertise” and “refined governance”.This group possesses land, power and cost advantages that address mature market capacity constraints. They also stand to benefit from accessing infrastructure and talent from AI leaders in the region. Malaysia’s land, power and capacity complement Singapore’s AI expansion within the Johor state. Indonesia’s vast renewable energy potential and strategic zones such as Nongsa Digital Park in Batam also help alleviate the region’s infrastructure capacity crunch.“This complementary dynamic will spawn a hyper-connected digital infrastructure ecosystem spanning Asia, positioning the region as a critical node in the global AI supply chain,” the study added. Infrastructural gapsSeventy-two per cent of organisations in Asia expect moderate to exponential growth of AI workloads over the next three years. However, this accelerated pace requires appropriate AI-ready infrastructure that most organisations lack, the study showed.Therefore, companies are looking to purchase AI readiness through hardware investment alone, with 64 per cent prioritising graphic processing unit acquisitions. Infrastructure-related gaps and critical operational expertise are also overlooked. Almost 50 per cent of companies report insufficient compute capacity for AI workloads, with only 7 per cent having “adequate headroom for complex AI applications and adoption growth”, the report stated. Firms also face networking limitations, with 82 per cent reporting network bottlenecks, latency issues or insufficient optimisation for high-demand AI workloads. Additionally, 83 per cent of firms in Asia report that less than half of their information technology locations are AI-ready, and are therefore unable to facilitate better user experiences and meet data residency and sovereignty requirements.