On Day One of the Bali Annual Telecommunication International Conference (BATIC) 2025, the closed-door Leadership Forum session cut through the stagecraft. Co-hosted by Telin and the Global Leaders’ Forum (GLF), the program brought together executives and strategists to wrestle with some hard questions: how to build digital infrastructure for AI, how to extract value from it, and how to power it all in a region still reliant on cheap coal. The themes were unmistakable: urgency, imbalance, and opportunity.
1. AI Has Shifted from Hype to Hard Infrastructure
In 2023, industry leaders were still talking about Web3 and the metaverse. By 2025, AI has not just arrived. It is dictating capital flows. Executives agreed that data centers and subsea fiber are now the most attractive investments in Asia-Pacific. The reason is clear: AI.
One operator described the transformation in physical terms: racks that once consumed 3 kilowatts now draw as much as 600. The scale of change requires new approaches to land, power, cooling, and facility design. Another added that planning data centers has become a guessing game: build too little and you miss the wave, build too much and you waste capacity. Either way, retrofitting at these loads is nearly impossible.
2. Enterprise AI Is Delivering Results, but Adoption Is Uneven
Boardrooms are instructing CIOs to “use AI”, but many organizations lack clear budgets or business cases. Current adoption is concentrated on cost optimization, often tied to workforce automation, while revenue-generating use cases remain elusive.
Yet pockets of success are undeniable. Some operators reported that internal AI-driven automation has cut processing times from hours to minutes. Large-scale networks are already resolving the majority of faults with AI before human intervention. The pattern is becoming clear: while consumer AI generates headlines, enterprise AI is quietly proving its value in operations and service delivery.
3. Sovereign AI Is Rising as a Parallel Track
Governments across the region are investing billions into sovereign AI projects and national “factories” for model training and strategic compute. These initiatives run parallel to enterprise adoption and reflect concerns over dependency on hyperscalers.
Enterprises themselves often prefer private clouds, especially when handling sensitive data, creating new opportunities for regional and local providers.
4. Telcos Are Missing the Edge Monetization Play
While subsea cables and hyperscale facilities dominate, participants pointed to a less visible but equally significant opportunity: the edge.
Operators already control central offices, points of presence, and access nodes. AI requires low-latency, high-quality, secure networks, ideally located closer to the user. Executives noted that these edge assets remain underutilized as a missed opportunity for monetization that could become critical as AI moves into real-time inference.
5. Energy Security Could Become the Bottleneck
If AI is the engine, energy is the fuel, and in Asia-Pacific, supply is fragile.
Data centers in Southeast Asia are expected to consume as much energy as entire small cities. This forces governments to make trade-offs: should 10 megawatts of power be allocated to data centers, or to schools and hospitals?
Some countries have already paused data center projects over energy security concerns. Policy swings add further risk, with governments slashing electricity tariffs to attract investment, then reversing course when subsidies prove unsustainable.
6. Policy Volatility Creates Investment Risk
Executives warned that sudden changes in tax and tariff regimes can destabilize long-term projects. One case was cited where electricity prices were cut to single digits per kilowatt-hour, only to jump again two years later, leaving investors stranded.
This volatility, compounded by surging demand for electrification across all sectors, raises questions about the durability of the AI build-out.
7. Standards, Trust, and Collaboration Still Lag
Technical capacity is racing ahead of governance. Fraud, spam, and opaque pricing in messaging have already eroded consumer trust. AI will amplify those vulnerabilities.
The consensus was that collective action is essential, from shared codes of conduct to standardized APIs and transparent supply chains. Without systemic trust, AI-driven services will struggle to scale sustainably.
8. Capex Is Pouring into AI, but ROI Remains Unclear
Investors are writing record-size checks for hyperscale facilities. But with three-year build cycles and six-month chip cycles, ROI calculations are increasingly tenuous.
Executives admitted they are making big bets without clear visibility on demand patterns. Bold choices are necessary. How many fibers to deploy, how much load to build for, but none come without risk.
9. The Next Wave Will Be Inference at Scale
Thus far, most spending has gone into training large models. The next inflection point will be inference, deploying models in production at scale. This will demand distributed infrastructure, ultra-low latency, and secure network integration.
That shift could elevate the importance of telecom edges and force new collaboration between carriers and cloud providers.
10. The Industry Must Think in Decades, Not Quarters
The session ended with a call for long-term vision. Global players emphasized that planning for AI requires horizons of two or three decades, not quarterly cycles. The takeaway was clear: demographics, economics, and adoption patterns all favor Asia-Pacific. But without affordable energy, trusted infrastructure, and systemic collaboration, the region risks missing the opportunity.
AI is not just a software story. It is an infrastructure, energy, and trust story. The region is bullish, but nervous. The winners will be those who:
- Build subsea, data centers, and edge capacity fast enough to match exponential demand
- Monetize edge assets as secure AI platforms
- Manage energy security through renewables, nuclear, and credible policy frameworks
- Rebuild trust with standards and transparency
- Shift from pilots to scaled inference that delivers ROI
The urgency is real. The AI future in Asia-Pacific will not be determined by hype cycles, but by steel, silicon, and power.