The AI token concept for ASEAN and Taiwan telecoms is more of a traffic and plan tool in the short term, and is not yet a primary earnings growth driver.
AI summary card
The AI token concept for ASEAN and Taiwan telecoms is more of a traffic and plan tool in the short term, and is not yet a primary earnings growth driver.
JPMorgan believes that AI token and model bundling services launched by telecom companies can easily be replicated, with more of the benefits flowing to consumers; the companies most likely to benefit are those with AI computing infrastructure, such as ST and ISAT.
- Current AI token products are mostly bundled model subscriptions or plan giveaways, lack exclusivity, and are unlikely to generate sustainable excess returns.
- If AI models are bundled only with higher-priced plans and encourage users to upgrade, industry ARPU may benefit, but this still depends on whether incremental revenue can cover token costs.
- Telecom companies building their own AI infrastructure, such as ST and ISAT, may gain incremental revenue from rising token consumption and computing demand.
- Telecom stocks broadly rose over the past month on sector rotation; FET outperformed among Taiwan telecoms, while ISAT and TLKM benefited from a mild spectrum auction and asset monetization.
Report interpretation
Overview
This report discusses the investment implications of recent AI token, AI model subscription, and plan bundling services launched by ASEAN and Taiwan telecom companies. The core view is that the current product format itself is unlikely to drive sustained earnings growth for telecom companies, as most products rely on third-party model distribution and successful offerings can easily be replicated by competitors. However, if telecom companies own or build AI computing infrastructure, rising token consumption could generate incremental revenue for that infrastructure.
Core views
First, AI token and model bundling services are currently more likely to represent consumer benefits than a sustainable source of telecom company profits. Second, AI features may improve industry revenue through higher ARPU only if they drive users to migrate to higher-priced plans, which requires validation of upgrade levels and token costs. Third, companies with AI infrastructure investments, such as ST and ISAT, are more likely to benefit from growth in AI usage. Fourth, recent share price gains were driven mainly by sector rotation, FET earnings growth, ISAT fiber asset monetization, and a mild spectrum auction, rather than by AI tokens themselves.
Analysis framework
The report uses a cross-sectional comparison of regional telecom companies, assessing the potential impact of AI tokens on telecom earnings and valuations based on AI product formats, plan pricing, the likelihood of user upgrades, token costs, AI infrastructure investment, and recent share price performance.
Methodology notes
Distinguishing AI product distribution from infrastructure benefit pathways
The report distinguishes between two earnings pathways: one involving bundled distribution of third-party AI models, which has limited earnings leverage due to a lack of exclusivity; and the other involving owned or combined third-party AI computing infrastructure, which may generate incremental revenue as token consumption increases.
Whether upgrades to higher-priced plans can cover token costs
AI features may have a positive impact on industry earnings only when they encourage users to select higher-priced plans and incremental revenue exceeds incremental token costs.
Asset mapping & comparison
Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).
- Singapore Telecom (STEL.SI)Preferred telecom stock; potential beneficiary of AI infrastructure
- Strengths
- Disclosed capex of up to S$600m and plans to deploy 11MW of AI compute by FY27; has the potential to generate infrastructure revenue from rising AI token consumption.
- Weaknesses
- AI token distribution itself may not constitute a sustainable earnings driver.
- Comparison
- Compared with telecom companies that only distribute third-party models, ST's owned AI infrastructure pathway offers greater potential differentiation.
- Risks
- Insufficient AI usage, lower-than-expected returns on computing investment, and token costs eroding plan revenue.
- Indosat Tbk PT (ISAT.JK)Preferred telecom stock; potential beneficiary of asset monetization and AI/5G reinvestment
- Strengths
- Received $650m from fiber asset monetization, valued at approximately 7x EV/EBITDA; proceeds can be used for AI, 5G growth projects, or shareholder distributions.
- Weaknesses
- The contribution from the AI business still requires validation; near-term share price benefits are more attributable to asset monetization and the mild spectrum auction.
- Comparison
- Compared with peers lacking capital release, ISAT has an advantage in redeploying capital.
- Risks
- Lower-than-expected capital utilization efficiency, insufficient returns on AI and 5G investment, and gains weakened by competitive replication.
- Advanced Info Services (ADVANC.BK)Preferred telecom stock; representative AI model subscription product
- Strengths
- AIS Alisa bundles 10 AI models into a Bt220/month subscription, well below the aggregate cost of individual subscriptions.
- Weaknesses
- Without an exclusive partnership, the product may be replicated by competitors after achieving success, with most of the profit benefit passed on to consumers.
- Comparison
- It is representative as an AI token subscription case, but its earnings leverage may not be stronger than that of companies with computing infrastructure.
- Risks
- Insufficient user conversion to paid subscriptions, price competition, and token costs exceeding incremental revenue.
- Far EasTone Telecom (4904.TW)Preferred telecom stock; recent earnings growth-driven share price
- Strengths
- 1H26 mobile revenue grew 6%, ICT revenue grew 33%, and net profit grew 17%, significantly above full-year guidance.
- Weaknesses
- The report does not primarily attribute its recent share price gains to AI tokens.
- Comparison
- Compared with sector rotation affecting most telecom stocks, FET's rise has stronger support from earnings growth.
- Risks
- A slowdown in earnings growth, unsustainable ICT growth, and valuation expectations exceeding business performance.
- True Corporation PCL (TRUE.BK)Preferred telecom stock
- Strengths
- Included in JPMorgan's list of preferred telecom stocks and rated OW.
- Weaknesses
- The report excerpt provides relatively few specific arguments regarding its AI or earnings drivers.
- Comparison
- It is a Thai telecom coverage name like ADVANC, but the report primarily uses AIS Alisa as its AI product example.
- Risks
- Industry competition, insufficient ARPU upgrades, and difficulty converting AI products into earnings.
Key data
- FET 1H26 mobile revenue growth6%The report states that Far EasTone's 1H26 mobile revenue grew 6% year over year.
- FET 1H26 ICT revenue growth33%The report states that Far EasTone's 1H26 ICT revenue grew 33% year over year.
- FET 1H26 net profit growth17%The report states that this growth rate was significantly above the full-year 3% growth guidance.
- AIS Alisa priceBt220/monthAIS Alisa is offered to eligible customers and bundles 10 AI models.
- Cumulative price of individual model subscriptionsapproximately Bt3.7k/monthThe report uses this to illustrate the price appeal of the bundled service to consumers.
- StarHub gigaFLEX+ youth starting priceS$13.9/monthThe plan includes unlimited AI access, showing that AI features are not necessarily tied to high-priced plans.
- ISAT fiber asset monetization cash proceeds$650mThe report considers the transaction value-accretive to shareholders, with fiber asset monetization valued at approximately 7x EV/EBITDA.
- ST AI computing capexup to S$600mST disclosed plans to deploy 11MW of AI compute by FY27.
- ST AI compute plan11MW by FY27This supports its infrastructure pathway, which may benefit from rising AI token consumption.
Impact & implications
For investors, ASEAN and Taiwan telecom AI token products should not simply be compared with the AI plan trend among Chinese telecom operators. At this stage, greater attention should be paid to whether AI genuinely drives upgrades to higher-priced plans, whether token costs are controllable, and which companies own monetizable AI infrastructure. The report's preferences are more concentrated in telecom companies supported by fundamentals, asset monetization, network investment, and computing investment.
Risks
- AI token and model bundling services lack exclusivity, and successful products may be quickly replicated by competitors.
- Consumers may capture most of the economic benefits, leaving telecom companies unable to retain sufficient profits.
- If AI features are tied to low-priced plans, they may fail to increase ARPU.
- Even if they drive user upgrades, incremental revenue may still be offset by token costs.
- AI infrastructure capex may face uncertainty regarding utilization and returns on investment.
- If recent share price gains were primarily driven by fund rotation, subsequent performance may lack sustained fundamental support.
What to watch
- Whether AI features shift from free giveaways or low-priced plans to being bundled with higher-priced plans.
- The gap between the user upgrade rate, ARPU changes, and token usage costs.
- Progress on ST's 11MW AI compute deployment and commercialization revenue.
- How ISAT's fiber asset monetization proceeds are actually allocated among AI, 5G, and shareholder returns.
- Whether FET can sustain its mobile, ICT, and net profit growth in the second half of the year.
- Whether AI products among regional telecom companies develop exclusive partnerships or non-replicable differentiation.