Current AI token products are unlikely to become the main driver of earnings growth for ASEAN and Taiwan telecom companies
AI summary card
Current AI token products are unlikely to become the main driver of earnings growth for ASEAN and Taiwan telecom companies
J.P. Morgan believes that existing telecom products bundling AI models and tokens are easy for competitors to replicate, making consumers more likely to benefit in the short term than the industry’s earnings to improve meaningfully.
- Most ASEAN and Taiwan telecom stocks outperformed over the past month due to sector rotation, with Taiwan telecom stocks standing out and FET rising on earnings growth.
- AI token and model bundle products are likely to be replicated by competitors after succeeding if there is no exclusive partnership, making it difficult to establish a sustainable earnings advantage.
- If AI models are tied only to premium plans, they may drive ARPU through customer upgrades, but the net impact depends on whether incremental revenue covers token costs.
- Telecom companies such as ST and ISAT that are building their own AI infrastructure may gain incremental revenue opportunities as token consumption rises.
- The report’s preferred telecom stocks include ST, ADVANC, TRUE, 4904, and ISAT.
Report interpretation
Overview
This report discusses the investment implications of ASEAN and Taiwan telecom operators launching AI tokens, AI model aggregation subscriptions, and telecom bundle services. The core conclusion is that current AI token products themselves are unlikely to significantly drive earnings growth for telecom companies, because these services are mainly based on third-party models, lack exclusivity, and successful products are easy for peers to copy. However, telecom companies that own or are building AI computing infrastructure may gain additional revenue from rising token usage.
Core views
First, current AI token products are more like value-added telecom plan benefits or model aggregation distribution, and are unlikely to create durable differentiated earnings. Second, the key condition for industry benefit is whether AI benefits can drive users to upgrade to higher-priced plans, and whether incremental revenue exceeds token and model usage costs. Third, compared with simply distributing third-party models, telecom companies building their own AI infrastructure are more likely to benefit from growth in AI usage. Fourth, recent share price performance has been driven more by sector rotation, earnings growth, spectrum auction results, and asset monetization than by AI token products themselves.
Analysis framework
The report uses industry thematic comparisons, company case reviews, and relative shareholder return analysis to compare AI token products of Chinese and global telecom companies, recent share price performance of ASEAN and Taiwan telecom companies, valuation metrics, ratings, and key financial data, while also assessing potential beneficiary pathways in conjunction with AI infrastructure investment.
Methodology notes
Compares three models: direct token sales, multi-model aggregation subscriptions, and telecom plan bundling.
Using cases such as Chinese telecom operators, AIS, TRUE, StarHub, Globe, and TLKM, the report shows that most current products are based on third-party model aggregation or promotional benefits, with limited core differentiation.
Compares recent total shareholder returns, relative performance versus local indices, PE, EV/EBITDA, dividend yield, and EPS growth across ASEAN, Taiwan telecom, and tower companies.
The report combines recent share price performance with earnings growth, spectrum auctions, asset monetization, and valuation levels to judge whether the AI theme has already become a stock price driver.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Singapore Telecom (STEL.SI)Preferred name and is building AI infrastructure
- Strengths
- Has an AI computing investment plan; the report mentions up to S$600m in capex to deploy 11MW of AI computing capacity by FY27.
- Weaknesses
- Revenue realization from AI infrastructure still depends on token consumption and growth in customer demand.
- Comparison
- Compared with operators distributing third-party models, ST is more likely to participate in growth in AI usage through its own infrastructure.
- Risks
- Uncertain capex returns and AI demand or pricing may fall short of expectations.
- Indosat Tbk PT (ISAT.JK)Preferred name, benefiting from asset monetization and potential reinvestment in AI/5G
- Strengths
- Fiber asset monetization brings in $650m in cash, and the approximately 7x EV/EBITDA valuation is considered value accretive by the report.
- Weaknesses
- Subsequent value depends on capital deployment efficiency and returns on growth projects.
- Comparison
- Unlike pure AI plan promotions, ISAT has room to reallocate capital after asset monetization.
- Risks
- Returns on AI and 5G investments are uncertain, and there is a trade-off between cash distribution and growth investment.
- Advanced Info Services (ADVANC.BK)Preferred name, launched AIS Alisa AI service
- Strengths
- Alisa AI bundles 10 AI models at Bt220/month, offering a strong consumer value proposition.
- Weaknesses
- Without exclusive partnerships, the product is easy for competitors to replicate once successful.
- Comparison
- Similar to multi-model aggregation offerings from TRUE, StarHub, and Globe, the core challenge is converting user benefits into ARPU and profit.
- Risks
- Rising token costs, competitive replication, and lower plan pricing may weaken earnings contribution.
- Far EasTone Telecom (4904.TW)Preferred name, with recent share price performance driven by earnings growth
- Strengths
- 1H26 mobile revenue grew 6%, ICT revenue grew 33%, and net profit grew 17%, significantly above full-year guidance of 3% growth.
- Weaknesses
- The report does not mainly attribute its recent performance to AI token products.
- Comparison
- Among Taiwan telecom stocks, FET’s performance is more clearly driven by earnings growth.
- Risks
- If earnings growth slows, the stock price catalyst may weaken.
- True Corporation PCL (TRUE.BK)Preferred name, offering a multi-model aggregated AI product
- Strengths
- The report lists TRUE’s enterprise multimodal platform as aggregating more than 50 AI models, covering image, web search, document processing, reasoning, research, and video generation functions.
- Weaknesses
- Multi-model aggregation does not equate to an exclusive advantage and is easy for peers to replicate.
- Comparison
- TRUE is closer to a multimodal aggregation platform model rather than a giveaway tied to a single-model subscription.
- Risks
- Enterprise client adoption, usage costs, and competitive pricing may affect profitability.
Key data
- FET 1H26 mobile revenue growth6%The report says FET's mobile revenue growth supported its share price performance.
- FET 1H26 ICT revenue growth33%ICT revenue growth is an important factor behind FET's earnings improvement.
- FET 1H26 net profit growth17%Above the full-year guidance of 3% growth.
- Cash proceeds from ISAT fiber asset monetization$650mThe report believes the transaction is accretive to shareholder value.
- Valuation multiple of ISAT fiber asset monetization约7x EV/EBITDAFunds may be used for growth projects such as AI and 5G networks, as well as cash distribution to shareholders.
- AIS Alisa AI priceBt220/月AIS says the service offers a single subscription to 10 AI models for eligible customers.
- AIS claimed standalone subscription cumulative price约Bt3.7kUsed to compare the discount value of the Alisa AI bundled subscription.
- StarHub gigaFLEX+ promotional priceS$13.90/月The plan includes unlimited AI text usage and 300 image generations per month, showing that AI benefits may be used to promote low-priced plans.
- ST AI computing capex plan最高S$600mST plans to deploy 11MW of AI computing capacity by FY27.
Impact & implications
For investors, ASEAN and Taiwan telecom operators’ AI token products should not be simply analogized to the stock catalysts seen in Chinese telecom operators. If current products are merely third-party model distribution or promotional benefits, earnings elasticity is limited; what truly deserves attention is whether AI services can drive upgrades to higher-priced plans, lift ARPU, and whether self-built AI infrastructure can generate verifiable incremental revenue. In the short term, stock selection should still place greater weight on earnings growth, asset monetization, spectrum costs, cash returns, and valuation.
Risks
- AI token products lack exclusivity, and successful models are easy for competitors to replicate.
- If AI benefits are used to promote low-priced plans, they may enhance customer value but fail to improve industry profits.
- Customer upgrade intensity may be insufficient, or incremental revenue may be lower than token and model invocation costs.
- Building AI infrastructure requires relatively high capex, and the payback period and utilization rate are uncertain.
- Recent gains in telecom stocks may mainly reflect sector rotation, earnings growth, spectrum auctions, and asset monetization, rather than the AI theme itself.
What to watch
- Whether AI services are tied to higher-priced plans and actually drive ARPU upward.
- The spread between token usage costs, model procurement costs, and incremental revenue for telecom companies.
- Progress in utilization, customer acquisition, and revenue recognition for AI infrastructure at companies such as ST and ISAT.
- Allocation of proceeds from ISAT’s fiber asset monetization among AI, 5G networks, and shareholder returns.
- Whether subsequent earnings growth at FET, ADVANC, TRUE, ST, and ISAT can support their OW ratings.