Asia-Pacific Investment Report 2017 — The Mobile-First Divergence

2017 was the year the shape of Asia-Pacific's consumer internet became clear enough to analyse properly, and the finding was that it was not a delayed version of the West's.

Several of the region's largest markets skipped a generation of infrastructure. Where developed markets built the internet on desktop computers, fixed broadband and credit cards, large parts of Asia went directly to mobile phones, mobile data and digital wallets. The intermediate infrastructure was never built at scale.

That produces structurally different outcomes, and the differences are not cosmetic:

Mobile-first means different product architecture. A service designed for a small screen, intermittent connectivity and a single device used by a whole household is a different product from one adapted from desktop.

Wallet-first means different transaction economics. Where credit cards never achieved penetration, digital wallets became the default. Wallets have different cost structures, different data characteristics, and different relationships to merchants — and a wallet provider sits in a different position in the value chain than a card network does.

Lower income levels mean different business models. Average revenue per user is lower, which changes what unit economics can support. Businesses requiring high per-user revenue do not work; those monetising through transaction volume, advertising at scale, or commerce margin can.

The most visible consequence was the super-app — a single application spanning messaging, payments, commerce, transport and services. This is frequently described as a strategic choice. It is better understood as a consequence of the architecture: where the wallet is the payment layer and the phone is the only device, bundling services around the wallet is the natural structure.

The analytical error that persisted through this period was applying Western comparables. A regional platform is not "the X of Asia." It occupies a different position in a differently-structured value chain, and the comparison misleads in both directions — overstating value where the analogue is more profitable, understating it where the regional platform's position is structurally stronger.

Why leapfrogging produces different structures, not delayed ones

The mechanism deserves setting out because "leapfrogging" is often used loosely to mean "catching up faster."

Technology adoption is path-dependent. What gets built depends on what already exists. An economy with extensive fixed telephone infrastructure builds internet access on it. An economy without one builds on mobile networks, because that is cheaper than laying wire.

The path determines the destination, not just the speed. An economy that never built card networks does not eventually build them and then move to wallets. It builds wallets, and the card network stage never happens.

This produces genuinely different structures:

• No incumbent to displace. A wallet in a market with high card penetration must displace an entrenched, well-functioning system. A wallet in a market without cards is competing with cash, which is a much weaker incumbent.
• Different regulatory starting points. Card networks came with decades of accumulated regulation. Digital wallets in markets without them were regulated from scratch, often in ways that permitted structures cards do not.
• Different data. A wallet provider sees the full transaction, the merchant relationship and often the surrounding context. A card network sees a payment authorisation. That difference in data richness supports different adjacent businesses — credit, commerce, advertising.
• Different economics. Card networks carry an established fee structure. Wallet economics were built later, frequently at lower cost, sometimes subsidised to drive adoption.

A market that skips a technology generation does not arrive at the same place later. It arrives somewhere else. The company that dominates there is not the local version of the company that dominates here.

Why super-apps do not transplant

The super-app is the clearest illustration of architecture determining strategy, and it is worth explaining why the model has repeatedly failed to transfer.

In a wallet-first mobile-first market, bundling is natural. The user has one device, one primary application, and a wallet that is already the payment layer. Adding a service to that application requires no new download, no new account, no new payment setup. The marginal cost of adding a service to an existing relationship is very low.

In a card-first, multi-device market, bundling is unnatural. Payment is already handled by an incumbent network. Users have many applications and no single dominant one. Adding a service to an application confers little advantage, because the user can as easily download a dedicated one — and specialised applications are usually better than bundled ones at any single task.

The failure of super-app strategies in Western markets is therefore not a failure of execution. It is a strategy whose advantage depends on conditions that do not exist there.

The general principle matters well beyond this case: a business model's viability depends on the market's structure, not on the model's merits in the abstract. A model that works in one market may be structurally impossible in another, and investors evaluating a "proven model" from another geography should first ask what conditions the proof depended on.

The reverse error occurs equally often — assuming a Western model will work in Asia. Subscription businesses requiring high per-user revenue, services assuming card payment, and products designed for desktop-primary use have all struggled in markets where the underlying conditions differ.

Unit economics at different income levels

A specific and frequently underestimated constraint is what average income permits.

Revenue per user is bounded by income. A consumer with lower disposable income can spend less on any service, regardless of how good it is. That is a hard constraint, not a marketing problem.

Costs do not scale down proportionally. Some costs — engineering, infrastructure, content licensing — are broadly similar regardless of market. Others — delivery, customer service, local operations — scale with local wages and are genuinely lower.

The consequence is that business models requiring high revenue per user do not work at lower income levels, and models that do work must find revenue elsewhere:

• Advertising, monetising attention rather than payment, which requires very large scale to produce meaningful revenue.
• Transaction margin, taking a small share of a large volume of commerce.
• Financial services, where lending and payment economics can be attractive even at modest transaction sizes.
• Cross-subsidy, where one profitable service supports others that build engagement.

This is why the super-app structure and the income constraint reinforce each other. A business that cannot charge much for any single service can be viable by capturing a small amount across many — which requires the bundled structure to be economic in the first place.

The investment implication is direct. A company in a lower-income market pursuing a model that requires high revenue per user is attempting something the market may not support. Conversely, a company with modest revenue per user but very large scale and multiple monetisation routes may be far more valuable than a per-user comparison with a Western analogue suggests.

Per-user metrics are the wrong comparison across income levels. Total addressable value, and the number of monetisation routes available, are better.

The supply chain benefits from synchronised growth

The 2017 global report describes synchronised global growth. Its clearest regional expression in Asia-Pacific was in the technology supply chain.

Korea, Taiwan, Japan and parts of Southeast Asia occupy critical positions in semiconductor manufacturing, components, displays and electronics assembly. These are cyclical businesses with a specific dynamic:

• Demand is derived from end-market demand for devices, data centres and industrial equipment, which amplifies cycles — a modest change in end demand produces a larger change in component orders.
• Capacity is lumpy. Semiconductor fabrication requires enormous capital investment with long lead times, so supply adjusts slowly and in large increments.
• The combination produces sharp cycles. When demand exceeds capacity, pricing power is substantial. When capacity arrives, it typically arrives in excess.

2017 was a favourable point in that cycle. Synchronised global growth raised demand across every end market at once, while capacity added in the preceding period had been absorbed.

The lasting observation is about how to analyse these businesses. The relevant variables are the capacity cycle and the position in the supply chain, not the growth rate of the end market. A component supplier in a growing end market can have poor economics if capacity is abundant; one in a flat end market can have excellent economics if it is scarce. Scarcity, not growth, determines pricing power — a point that became central again in 2024 when AI demand met constrained advanced-node capacity.

Why the wrong comparison is the default one

The habit of describing a regional platform as "the X of Asia" is worth examining, because it is not merely lazy shorthand — it produces systematic valuation errors in both directions.

Why the comparison is reached for. A Western analyst assessing an unfamiliar company needs a reference point. The nearest available one is a company they already understand, and analogy is how unfamiliar things are made tractable.

What the analogy imports without examination:

• The revenue model. If the Western analogue monetises through subscription and the regional company monetises through commerce margin, they have different economics with different margin structures and different sensitivities.
• The competitive position. The Western analogue may hold a dominant position built over decades against a specific competitive set. The regional company operates in a different market structure — frequently with a public-utility payment layer, different regulation, and different incumbents.
• The margin structure. Determined by income level, cost base and monetisation route. A company serving lower-income customers at large scale has a different margin profile from one serving higher-income customers at smaller scale, even at identical revenue.
• The addressable market. The analogue's market and the regional company's market are different sizes with different growth rates and different ceilings.

The error runs in both directions, which is why it is not a simple bias:

• Overstatement where the Western analogue is more profitable, has a stronger competitive position, or operates in a larger addressable market than the regional company does.
• Understatement where the regional company's position is structurally stronger — a super-app bundling services in a market where bundling is natural has an economic structure the Western analogue cannot replicate, and comparing it to a single-purpose Western application understates what it is.

The correct approach is to build the analysis from the market's own structure: what does the payment layer look like, what income supports what spending, what does the competitive set consist of, what monetisation routes are available. That is more work than an analogy and it produces an answer rather than a reference.

"The X of Asia" is a claim that two companies occupy the same position in the same structure. In a market that skipped a technology generation, the structures are not the same — which means the position is not either, and the comparison is doing no analytical work at all.

What an allocator could act on

Ask what the payment layer is before assessing a consumer business. Whether it is card-based, wallet-based or a public utility determines the economics, the available adjacent businesses, and where value accrues. A fintech thesis built for one architecture does not transfer to another.

Size the market by spending capacity, not population. A service requiring meaningful annual spending is addressable only to the segment with that capacity. Population-based market sizing overstates by one or two orders of magnitude in lower-income markets, and the error appears years later as a growth plateau read as an execution failure.

Compare total monetisation routes, not revenue per user. Per-user metrics are not comparable across income levels. A business with modest revenue per user, very large scale and several monetisation routes can be worth substantially more than a per-user comparison implies.

Assess supply chain businesses on the capacity cycle. Scarcity determines pricing power, not end-market growth. A component supplier in a growing end market can have poor economics if capacity is abundant; one in a flat market can have excellent economics if it is scarce. The relevant series is capacity additions relative to demand, and SEMI publishes equipment billings free and monthly as a leading indicator.

Check whether a proven model's proof depended on conditions that exist here. A business model's viability is a property of the market structure, not of the model. The super-app works where the wallet is the payment layer and the phone is the only device; it fails where neither holds. Before importing a model, name the conditions its success depended on and check whether they are present.

What 2017 established for Asia-Pacific

• Leapfrogging produces structurally different markets, not delayed versions of existing ones.
• The super-app is architecture-determined, and does not transplant to markets with different payment and device structures.
• Income level constrains business model viability, making per-user comparisons across income levels misleading.
• Western comparables produce systematic misvaluation in both directions.
• Supply chain analysis depends on the capacity cycle, not on end-market growth.