Form factor, the shape, size, input methods, and real-time availability of the prevailing medium, has always been the silent yet omnipresent architect of digital value. It determines not only how information is displayed but how attention is structured, how habits are conditioned, and how economic moats are built or eroded.1 Every major transition in consumer technology, such as the transition from the desktop to the laptop to the smartphone, has shown that the interface itself is never neutral: it quietly selects which business models thrive and which ones wither.
For the past decade and a half, the prevailing form factor has been unusually stable: a slim rectangle carried in the pocket, governed by familiar touch gestures, notifications, and always-on connectivity. This prolonged stability has fostered a subtle but pervasive assumption across the industry. Companies, investors, and strategists now routinely project the current medium outward as if it were permanent, allocating capital, designing organizations, and evaluating competitive position on the implicit premise that the smartphone will remain the center of the digital universe indefinitely. The result is widespread investment in assets and capabilities that are, in truth, borrowed from today’s device rather than owned by the business itself—an assumption that works beautifully while the medium holds, and becomes costly the moment it begins to shift.
This framework is a practical method for separating durable competitive advantage from advantages that are rented from the current technology medium.2 The relevant work for the analyst is not to forecast the next device, but to separate which parts of a current business would survive translation to any different medium from those predicated on the current prevailing medium. The first category is durable. The second is borrowed. In practice, the distinction is often a gradient rather than a sharp binary; some assets sit right on the seam and are valuable in both layers at once. Both layers typically coexist within a single business, and the distinction tends to go unexamined while the medium is stable.
“Durable” Means Valuable Independent of the Medium
An asset is durable to the extent that it would retain its value even if the interface changed completely. The relevant test is not whether the company could rebuild for a new medium, but whether the underlying asset retains its value during and after that rebuild.
Durable assets may differ in form, but collectively they are products of accumulated time under specific conditions, which capital cannot often compress. A licensed catalog requires years of negotiation under terms a new entrant cannot easily replicate. A global shipping and fulfillment network requires years of physical warehouse construction, route optimization at massive real-world scale, regulatory approvals across jurisdictions, and supplier relationships forged through consistent high-volume operations. An indexed corpus of human conversation requires years of those conversations actually occurring. These assets are durable not because they are technically complex, per se, but because the production function involves time as an irreducible input.
Durable assets also address a need that exists independent of how the user expresses it. They rest on infrastructure that is expensive and slow to replicate: physical assets, regulatory licenses, supplier relationships, accumulated content, or accumulated user data that meaningfully improves the service. These then produce switching costs unrelated to interface familiarity, including contracts, integrations into the user’s adjacent systems, and history that would be costly to reconstruct elsewhere.
A useful sharpening question: if the company had to deliver its core value through a different form of interaction, what fraction of the current organization, infrastructure, and competitive position would still apply? Answer this question, and you define the durable core; everything else is interface.
“Borrowed” Means the Medium Is Doing the Work
A capability is borrowed when its value depends on the specific properties of the current technology medium. Borrowed value is not illusory: it frequently accounts for most current revenue and engagement, but it is held on the medium’s terms, terms that are not within the company’s control.
On any given medium, user behavior is largely conditioned rather than freely chosen. Repeated cue-actionreward cycles3 (the lock screen lights up, the thumb opens the app, the feed delivers variable reward) produce habits that operate below deliberate preference. The cues, the actions, and the reward delivery are all properties of the device. Change the device and the conditioning loop breaks; the user is presented, briefly, with the conscious choice of what to do, and conscious choice is a far weaker driver of behavior than habit. Borrowed advantages depend on the strength of the conditioning loop. They erode at exactly the moment the loop is interrupted.
Some indicators of a borrowed advantage:
- The interaction is the product. When users return for the feel of scrolling, swiping, glancing, or tapping in a particular rhythm, the product is the rhythm as much as the content. Alter the rhythm and the product becomes something else, since the cue and the action that constituted the experience no longer exist in the same form.
- Monetization is dimensionally specific. Revenue tied to a particular screen aspect ratio, session length, notification cadence, or in-feed ad slot is revenue tied to the persistence of those dimensions as the structure of attention. The pricing of an ad slot is the pricing of a unit of conditioned behavior; the slot retains value only as long as the behavior pattern that produces views of it persists.
- Retention depends on hooks specific to the device. Notifications, lock-screen presence, home-screen real estate, and the ambient availability of a phone in a pocket function as cues in the conditioning loop. They are not invariants. They are properties of one form factor’s operating system conventions, and a medium that redefines those terms redefines the cues, and therefore the retention, with them.
- Discovery flows assume the current topology. App stores, share sheets (the native operating-system menus that appear when a user taps “Share”), deep links (special URLs that open directly to a specific screen or piece of content inside an installed app rather than a homepage or web page), and the general choreography by which users find and move between experiences are conventions of the current stack. Businesses that scaled on those conventions scaled on decisions made by platform owners; the conventions are revisable, and any revision unsettles the discovery patterns built on top of them.
Durability is Rarely Binary
The goal is to identify the boundary between what the company architects and owns and how the user currently experiences it. The further apart those two are, the more durable the business. The closer they are—the more the experience and the asset intersect—the more borrowed and less durable. Some assets sit directly on that seam, part durable data or relationship and part borrowed surface expression. Most assets sit on a gradient between the two: the relevant question is the ratio and the location of the seams.
A business with a deep underlying asset and a thin experience layer is positioned for transitions, since the experience layer can be rebuilt while the asset persists. A business whose underlying asset and experience layer are coextensive—where the way users interact with the product is the product—is exposed, since rebuilding the experience layer entails rebuilding the product itself, and during the rebuild, challengers native to the new medium operate in an open field.
Common Advantage Misconceptions
Although the durable-versus-borrowed framework is conceptually straightforward, it is surprisingly easy to misapply in practice. There are four recurring errors that frequently render the analysis ineffective or misleading. Each of these pitfalls stems from a specific category mistake about what actually drives user behavior and creates sustainable competitive advantage.
- Conflating scale with durability. A company with hundreds of millions of users on the current medium appears unassailable, and on the current medium it typically is. The logical error is to treat the user count as a measure of the company’s hold on those users when, in significant part, it measures the medium’s hold on them. Users exist because the conditioning loop is present. Disrupt the loop and a portion of the user base is, in effect, redistributed to whichever business establishes the next loop most effectively. The history of media transitions is largely a history of dominant incumbents that conflated scale-under-conditioning with scale-as-loyalty.
- Conflating brand with moat. Brand recognition built within a specific interaction context does not automatically transfer to a different one, because brand recognition operates as a learned cue rather than as an abstract preference. The user has been trained that a particular icon, in a particular location, leads to a particular reward. Strip away the icon, the location, and the routine that connects them, and what remains is the abstract concept of the brand, which is far weaker than the cue-bound version. Users who reflexively open a particular app on a phone do not necessarily reflexively reach for that brand through a different interface because the reflex was constructed from elements the new interface does not preserve.
- Treating current conventions as user preferences. “Users prefer this format” and “users prefer the format the dominant device renders effortless” are not equivalent statements, though they appear identical from within the current era. Preferences in a constrained environment reveal as much about the constraint as about the preference. Users select from what is offered along the gradient of least friction; the friction gradient is set by the medium. Preference data describes the equilibrium of usersplus-medium, not users alone, and it has limited predictive power for behavior under a different friction gradient.
- Conflating early bets with a durable position. Early bets on new mediums fail regularly—sometimes because the medium does not arrive, because it arrives differently than projected, or because the early version is technically capable but socially premature. The objective is not to chase the next thing. It is to identify, with discipline, which parts of the current business would remain standing if the next medium arrived and to invest in those parts disproportionately while the current medium is funding everything.
Investment in durable assets compounds across transitions, but investment in borrowed assets is dependent on the persistence of the current medium. Both positions are legitimate, but they should not be conflated since they carry different risk profiles and time horizons. Conflating them can easily lead to errors of misjudgment.
Important Disclosures
Past performance is no indication or guarantee of future performance and no representation or guarantee is being made as to the future investment performance of Tsai Capital’s separately managed accounts or any entity.
The information provided herein includes certain opinions, portfolio and/or market characteristics, and other information provided by Tsai Capital, as of May 2026. Any opinions or other information included herein may change without notice and should not be relied upon in selecting any investment or investmentrelated product or service. Nothing contained herein should be construed as investment advice.
All investments carry risk, including the risk of loss of investment principal. Additionally, short-term market volatility may present increased risks for investors who have shorter investment horizons due to impending or current liquidity needs.
1 This perspective draws from Marshall McLuhan’s published theories, most notably Understanding Media, in which the characteristics of the medium itself shape human perception, behavior, and society more profoundly than the content it carries.
2 These themes are examined at length in Tsai Capital’s 2025 Investor Letter, specifically in the section entitled “The Medium is the Moat: McLuhan’s Take on Tech’s Hidden Power,” which applies these ideas to investment analysis.
3 The concept of repeated cue-action-reward cycles draws from B. F. Skinner’s foundational theories of operant conditioning in behavioral psychology, in which environmental cues trigger behaviors that are then strengthened or weakened by subsequent reinforcements or rewards.