Finance · Banking
DBS Cuts AI Spending While Deployment Accelerates
Singapore's largest bank is optimizing its artificial intelligence stack by matching models to tasks, achieving lower per-token costs even as employee usage climbs.
KEY TAKEAWAYS
- ·DBS has reduced per-token AI costs while employee usage expands by matching smaller language models to simpler tasks and testing multiple providers.
- ·The bank's overall technology spending remains at 10% of revenue despite wider AI adoption, indicating efficiency gains are offsetting new costs.
- ·DBS reported record Q2 net profit of S$3.08 billion and is evaluating which AI models to build internally rather than license externally.
Multi-Model Strategy Delivers Efficiency
DBS Group Holdings has cracked a puzzle that many financial institutions face: how to scale artificial intelligence use without ballooning costs. The Singapore lender is deploying a tiered approach that assigns routine queries to smaller language models while reserving heavier compute for complex requests, according to Chief Executive Officer Tan Su Shan.
The strategy has driven down the cost per token even as staff adoption accelerates across the organization. Rather than committing to a single provider, DBS evaluates models from multiple vendors and selects them based on specific use cases. The bank also relies on caching mechanisms to prevent redundant processing when employees submit identical queries.
Tan noted that staff are learning to refine their prompts and interactions with AI tools, which has further improved efficiency. The optimization reflects a maturation phase in enterprise AI adoption, where early experimentation gives way to disciplined deployment.
Technology Budget Holds Steady
Despite the surge in AI activity, DBS has kept overall technology spending at roughly 10% of revenue, a level that has remained stable over recent quarters. The discipline suggests that efficiency gains from AI are offsetting incremental costs, allowing the bank to absorb new capabilities within existing budgets.
DBS is also evaluating which models and capabilities should be built in-house rather than licensed from external providers. The internal development track gives the bank more control over proprietary workflows and data handling, particularly for sensitive financial operations that require regulatory compliance and security.
The open architecture approach allows DBS to avoid vendor lock-in and shift resources as the competitive landscape among AI providers evolves. Singapore's regulatory environment, which has encouraged sandbox testing and innovation in financial technology, provides a supportive backdrop for this experimentation.
Record Earnings Backdrop
The AI efficiency push comes as DBS reported record net profit of S$3.08 billion for the second quarter, prompting the bank to lift its full-year outlook. The strong financial performance gives management room to invest in technology infrastructure without immediate pressure on margins.
Asian banks have been among the most aggressive adopters of AI in the financial sector, driven by high digital penetration, competitive retail markets, and government support for technology initiatives. Singapore, Hong Kong, and South Korea have each launched frameworks to encourage responsible AI use in banking, from credit underwriting to fraud detection.
DBS has positioned itself as a technology leader within the regional banking sector, investing in cloud infrastructure, API platforms, and now generative AI. The bank's ability to contain AI costs while scaling usage will serve as a test case for other institutions navigating similar trade-offs.
What to Watch
The trajectory of per-token pricing across major AI providers will shape whether DBS can sustain its cost efficiency as usage grows. Competition among model vendors has driven prices down sharply over the past year, but that trend may flatten as the market consolidates.
Internal development of specialized models could become a differentiator for banks with sufficient scale and technical talent. DBS is well-positioned in Southeast Asia to recruit engineering staff and collaborate with research institutions in Singapore, but maintaining in-house AI capabilities requires sustained investment.
Regulatory clarity around data residency, model explainability, and liability for AI-generated outputs will also influence how aggressively banks can deploy these tools in customer-facing and risk-sensitive functions. DBS operates across multiple jurisdictions with varying regulatory maturity on AI, which adds complexity to any regional rollout.
RELATED STORIES
Singapore Banks Hit Record Highs on Wealth Management Surge
Priyanka V. Rao · 2 min
Singapore Approves 37 Crypto Firms From Nearly 300 License Applications
Priyanka V. Rao · 2 min
Singapore's Big Three Banks Post Solid First-Half Gains as Wealth Offsets Margin Pressure
Priyanka V. Rao · 3 min
Singapore Expands Bank Network for Financial Crime Intelligence Platform
Priyanka V. Rao · 3 min
Spot something wrong? Email editor@briefasia.com. We log every correction publicly.