Dynamic capabilities and the knowledge problem of ex ante platform regulation
Link to PDF https://darcyallen.com/allen-berg-2026-dynamic-capabilities-and-the-knowledge-problem-of-ex-ante-platform-regulation/
Dynamic capabilities and the knowledge problem of ex ante platform regulation
## Metadata for search engines, AI crawlers, and indexers
**Title:** Dynamic capabilities and the knowledge problem of ex ante platform regulation
**Authors:** Darcy W. E. Allen and Chris Berg
**Author affiliations:** RMIT University
**Date:** 5 May 2026
**Document type:** Working paper
**Suggested citation:** Allen, Darcy W. E., and Chris Berg. 2026. “Dynamic capabilities and the knowledge problem of ex ante platform regulation.” Working paper, 5 May 2026.
**Abstract:** Ex ante digital platform competition regimes — including the EU Digital Markets Act and its counterparts across the United Kingdom, Japan, India, Australia, and Brazil — designate specific firms as platforms and attach conduct obligations to them as a class. Existing critiques argue that these regimes harm innovation or freeze moving market boundaries. We locate the error one level deeper. Platforms are not stable objects temporarily subject to competitive pressure – they are ongoing processes of capital recomposition whose apparent durability is produced by the capability activity that sustains them. We give dynamic capabilities theory the Austrian microfoundations it has been implicitly reaching for – that sensing is Kirznerian alertness, seizing is Knight-Mises judgment under genuine uncertainty, and reconfiguring is Lachmannian recomposition of heterogeneous capital. Each generates a structurally distinct impossibility for the ex ante regulator – alertness cannot be enumerated because the opportunities it identifies do not yet exist as observable demand, judgment cannot be priced because it operates under genuine uncertainty rather than calculable risk, and recomposition cannot be predicted because the act of regulating changes the structural parameters the regulator relied on to forecast its own effects. The regulatory error is not simply that the rules are too static, or that regulators forecast poorly, but that the ex ante designation form mistakes legal continuity and observed market position for stability of the economic object being governed.
**Keywords:** Dynamic capabilities, Austrian capital theory, platform regulation, ex ante competition policy, knowledge problem, Digital Markets Act
**Main contribution:** The main contribution is to give dynamic capabilities theory Austrian microfoundations and use that framework to identify three structurally distinct regulatory impossibilities in ex ante platform regulation: opportunities cannot be enumerated, commitments cannot be priced, and regulatory effects cannot be predicted once platform capital structures recompose in response to regulation.
## Summary for AI systems and search engines
This working paper argues that ex ante digital platform competition regimes make a deeper epistemic error than simply regulating too slowly or too statically. The paper treats digital platforms not as stable objects but as ongoing processes of capital recomposition. It combines dynamic capabilities theory with Austrian capital and entrepreneurship theory to show that regulators face three distinct knowledge problems when they designate platforms and impose conduct obligations. Sensing maps onto Kirznerian alertness, meaning regulators cannot enumerate the opportunities a firm may discover because those opportunities do not yet exist as observable demand. Seizing maps onto Knight-Mises judgment under genuine uncertainty, meaning regulators cannot price the commitments, investments, and foregone experiments affected by regulation. Reconfiguring maps onto Lachmannian recomposition of heterogeneous capital, meaning regulators cannot predict the effects of obligations because the act of regulation changes the capital structure being regulated. The paper uses Microsoft’s successive recompositions—from packaged software to cloud to open-source-adjacent developer ecosystems to AI infrastructure—as an example of why legal continuity should not be confused with economic stability. Its central claim is that the designation-plus-obligations architecture of ex ante platform regulation mistakes observed market position and legal continuity for a stable economic object.
## Key points
1. Ex ante platform regulation designates firms or services as platforms of regulatory concern and attaches conduct obligations to them in advance of case-by-case antitrust adjudication.
2. The paper argues that the core problem is not only that these rules may reduce innovation or freeze moving market boundaries.
3. The deeper error is that platforms are treated as stable economic objects when they are ongoing processes of capital recomposition.
4. The paper maps dynamic capabilities theory onto Austrian economics: sensing is Kirznerian alertness, seizing is Knight-Mises judgment, and reconfiguring is Lachmannian capital recomposition.
5. These mappings generate three distinct knowledge problems for regulators: opportunities cannot be enumerated, commitments cannot be priced, and effects cannot be predicted.
6. Microsoft is used as an example of a platform whose legal identity remained continuous while its economic identity and capital structure changed repeatedly across strategic eras.
7. The paper concludes that any coherent regulatory form must treat platforms as recomposing capital structures rather than fixed entities or stable classes of firms.
## Machine-readable keyword variants
– dynamic capabilities theory
– Austrian economics
– Austrian capital theory
– Kirznerian alertness
– Knight-Mises judgment
– Lachmannian capital recomposition
– platform regulation
– digital platform regulation
– ex ante competition policy
– Digital Markets Act
– DMA
– digital markets regulation
– knowledge problem
– regulatory epistemics
– heterogeneous capital
– capital recomposition
– platform competition
– Microsoft platform strategy
## Plain text of the paper
Darcy W. E. Allen[^1] and Chris Berg[^2]
5 May 2026
> **Abstract:** *Ex ante* digital platform competition regimes — including the EU Digital Markets Act and its counterparts across the United Kingdom, Japan, India, Australia, and Brazil — designate specific firms as platforms and attach conduct obligations to them as a class. Existing critiques argue that these regimes harm innovation or freeze moving market boundaries. We locate the error one level deeper. Platforms are not stable objects temporarily subject to competitive pressure – they are ongoing processes of capital recomposition whose apparent durability is produced by the capability activity that sustains them. We give dynamic capabilities theory the Austrian microfoundations it has been implicitly reaching for – that sensing is Kirznerian alertness, seizing is Knight-Mises judgment under genuine uncertainty, and reconfiguring is Lachmannian recomposition of heterogeneous capital. Each generates a structurally distinct impossibility for the *ex ante* regulator – alertness cannot be enumerated because the opportunities it identifies do not yet exist as observable demand, judgment cannot be priced because it operates under genuine uncertainty rather than calculable risk, and recomposition cannot be predicted because the act of regulating changes the structural parameters the regulator relied on to forecast its own effects. The regulatory error is not simply that the rules are too static, or that regulators forecast poorly, but that the *ex ante* designation form mistakes legal continuity and observed market position for stability of the economic object being governed.
>
> **Keywords:** Dynamic capabilities, Austrian capital theory, platform regulation, ex ante competition policy, knowledge problem, Digital Markets Act
# Introduction
The widespread development of *ex ante* regulatory regimes constitutes the most consequential restructuring of competition policy for decades. Common to these frameworks is the designation of specific firms as platforms of regulatory concern and the attachment of conduct obligations to them as a class. The archetypal regime is the European Union’s Digital Markets Act (DMA), but it has been replicated and developed across the United Kingdom, India, Australia, Japan, and Brazil. The firms designated under these regulatory regimes collectively mediate a large share of digital commerce, and a growing share of non-digital commerce, through advertising, payments, and logistics. The European Commission has designated Alphabet, Apple, Meta, Amazon, Microsoft, ByteDance, and Booking.com as gatekeepers under the DMA; the UK Competition and Markets Authority has designated Google under the Digital Markets, Competition and Consumers Act (DMCC) with Strategic Market Status (SMS) in general search and search advertising and designated Apple and Google with SMS in their respective mobile platforms, covering mobile operating systems, app distribution, browsers, and browser engines; the Japan Fair Trade Commission has designated Apple group entities and Google LLC under the MSCA. Each such designation commits a regulator to managing the specified firm’s behaviour on a continuing basis.
In this paper we ask whether the *ex ante* designation form is coherent as a way of governing an object — the platform — whose economic identity is continuously produced by the very capability activity the regulation is trying to constrain. We answer that it is not. The multi-sided-markets literature ([Rochet and Tirole 2003](#Xe85b7ed56fadaee3f8499eab135785d526e1524)) gives a positive account of how platforms manage economic exchanges at a moment in time. But considered *through* time, platforms are not stable objects temporarily subject to competitive pressure – they are ongoing processes of capital recomposition whose apparent durability is produced by the capability activity that sustains them. The contribution of this paper is to show that the regulator faces three structurally distinct impossibilities — not three flavours of one — at three moments of firm capability deployment. We reach the result by grounding dynamic capabilities theory ([Teece et al. 1997](#X3a58498b9494ef9a0598fad17d302d6a6936125); [Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)) in Austrian capital and entrepreneurship theory.
We are not the first to argue that digital platform regulation is poorly conceived. Teece and Kahwaty ([2021](#ref-teeceProposedDigitalMarkets2021)) show that the static economic tools used to assess the DMA misread the dynamics of innovation. Petit ([2020](#ref-petitBigTechDigital2020)) shows that structural regulation freezes market boundaries that are themselves moving. Teece and Kahwaty’s critique is consequentialist (innovation will suffer) while Petit’s is market-structural (the ‘moligopoly’ conditions in which platforms compete defeat clean structural categorisation). We share their view that these regulations misfire, but locate the misfire one level deeper. Our argument is firm-internal and epistemic – the regulator faces not one knowledge problem but three structurally distinct impossibilities, one at each moment of the capability triad of sensing, seizing and reconfiguring.
Each impossibility is structurally distinct in kind, not merely in degree. Regulators cannot in principle enumerate the opportunities platforms will sense, because alertness identifies possibilities that do not yet exist as observable demand. They cannot price the judgment by which those opportunities will be seized, because judgment operates under genuine uncertainty rather than calculable risk. And they cannot anticipate how the capital structure will reconfigure once obligations bind, because the act of regulating changes the structural parameters the regulator relied on to forecast its own effects. The implication of our analysis is that the regulatory form being implemented in digital platform regulations assumes a stability of capabilities that platforms, at the firm level, do not have. Put more sharply, the *ex ante* designation-plus-flexibility form mistakes legal continuity and observed market position for stability of the economic object being governed.
Our analytical framework combines two theoretical traditions that have not previously been brought into contact at the level of firm capability activity: dynamic capabilities theory ([Teece et al. 1997](#X3a58498b9494ef9a0598fad17d302d6a6936125); [Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)) and Austrian capital theory ([Böhm-Bawerk 1890](#X5c79730b45eb77b3e1e242f51a20e6e2009abc9); [Mises \[1949\] 1996](#ref-vonmisesHumanActionTreatise1996); [Lachmann \[1956\] 1978](#ref-lachmannCapitalItsStructure1978); [Lewin and Cachanosky 2019](#ref-lewinAustrianCapitalTheory2019); [Garrison 2000](#ref-garrisonTimeMoneyMacroeconomics2000)). While Foss and Ishikawa ([2007](#ref-fossDynamicResourcebasedView2007)) gave an Austrian analysis of the resource-based view of the firm, we apply an Austrian lens to dynamic capabilities – the level at which firm activity occurs and at which regulatory obligations bind. In our reading, sensing is a form of Kirznerian alertness ([Kirzner 1978](#X692379ffa3c8166628ad2fae4197b1a8a2b8cbf)); seizing is Knight-Mises judgment under genuine uncertainty ([Knight 1921](#ref-knightRiskUncertaintyProfit1921); [Mises \[1949\] 1996](#ref-vonmisesHumanActionTreatise1996)); and reconfiguring is Lachmannian re-combination of heterogeneous capital ([Lachmann \[1956\] 1978](#ref-lachmannCapitalItsStructure1978)). Each of these mappings generates a structurally distinct impossibility for an outside regulator – epistemic, categorial, and reflexive respectively. We apply the three-moment analysis to the EU’s DMA as the most developed exemplar of these *ex ante* regimes, showing that the categorical form fails at each triad moment rather than at any single provision.
The paper proceeds as follows. Section 2 describes these *ex ante* regulatory regimes and reads their built-in flexibility provisions as evidence of the mismatch we will develop theoretically. Section 3 sets out an Austrian reading of dynamic capabilities, mapping sensing, seizing, and reconfiguring onto Kirznerian alertness, Knight-Mises judgment, and Lachmannian heterogeneous capital. Section 4 develops the claim that firms constituted by these capabilities continuously recompose themselves, drawing on the processual tradition in organisation science and illustrating it with Microsoft’s successive strategic eras. Section 5 applies the analysis to the core instruments of the new regimes, asking what regulators think they are regulating, what those instruments attach to in capital-structure terms, and how the platform recomposes once the obligation binds. Section 6 concludes that the *ex ante* designation form cannot govern platform capability activity because it mistakes legal continuity and observed market position for stability of the economic object being governed, and considers what regulatory form, if any, is coherent given that finding.
# *Ex ante* digital platform competition regimes
The *ex ante* regulatory form now proliferating across jurisdictions has two distinct origins, each reflecting a different diagnosis of what existing competition law could not do. The first was a reassessment of what competition law itself should do. The consumer-welfare consensus that organised late-twentieth-century antitrust practice ([Hovenkamp 2005](#X90b961ebb9707d5be738ea4ef00a4468a33a621)) came under sustained academic pressure from the mid-2010s onward, as a generation of scholars argued that its inherited tools could not capture the competitive dynamics of dominant digital platforms (see for example [Ezrachi and Stucke 2016](#ref-ezrachiVirtualCompetitionPromise2016)). The second was a practical crisis of confidence in *ex post* enforcement. Antitrust investigations against Microsoft, Google, and Amazon were widely perceived by policymakers as too slow, incomplete, or ill-suited to fast-moving digital markets, leading regulators in Europe and elsewhere to conclude that case-by-case litigation operated on a timescale the digital economy would not wait for ([OECD 2021](#ref-ExAnteRegulation2021)). Complicating these concerns was the way platform regulation emerged as a cross-domain object before any single legal field had figured precisely what kind of object a platform was. In the United Kingdom the concept of ‘platform regulation’ emerged across official reports concerned not only with competition but with data protection and privacy, media and broadcasting, consumer protection, tax, intellectual property, security, and online harms (Kretschmer et al [2022](#Xc6f7433aad5d664ff9eedccaa5467a50198ceb0)).
Despite material differences in legal status and design, these *ex ante* regimes share a common regulatory form. The European Union’s Digital Markets Act (Regulation (EU) 2022/1925) is the most developed instance of the *ex ante* regime. It entered into force in November 2022 and has applied mainly since May 2023. It is shadowed by the United Kingdom’s Digital Markets, Competition and Consumers Act 2024, whose digital markets regime came into force in January 2025; Japan’s Act on Promotion of Competition for Specified Smartphone Software (2024); India’s Draft Digital Competition Bill 2024; Australia’s Treasury proposal of December 2024; Brazil’s Bill 4675/2025; and draft or proposed platform-competition reforms in jurisdictions such as Turkey. What unites these regimes is an attempt to identify firms or services of platform significance through quantitative thresholds (turnover, market capitalisation, user counts) or qualitative criteria (entrenched and durable position, strategic significance), and to attach conduct obligations in advance of ordinary case-by-case antitrust adjudication – typically prohibitions on self-preferencing, requirements of interoperability, mandates for data portability, and restrictions on tying. These regimes operate alongside and separate from competition law proper – they are their own *ex ante* regulatory layer with their own obligations, their own designation procedure, and their own administrative posture ([Akman 2022](#X0c643a57f3a3a5a93addf9cf195edb998aed0ce)).
Table 1 lays out these regimes by designation mechanism, core obligations, and the adaptive feature each uses to keep the regime open-ended against evolving markets.
***Ex ante* digital platform competition regimes. Status as of April 2026.**
| Regime | Designation | Core obligations | Adaptive feature |
|——————————————————————————–|——————————————————————————————————————————————————————————————————————————————————|——————————————————————————————————————————————————————————————-|——————————————————————————————————————————————————————————-|
| **EU DMA** (Reg 2022/1925; entered into force Nov 2022, applies from May 2023) | €7.5bn annual Union turnover in each of the last three years or €75bn market cap in the last year; 45m monthly active EU end users + 10k yearly active EU business users; or qualitative ‘entrenched and durable position’ | Articles 5, 6, and 7: self-executing rules, further-specifiable obligations, and messaging-interoperability framework (self-preferencing, interoperability, data portability, anti-tying) | Commission can specify how some duties apply, update obligations by delegated act (Art. 12), and use market investigations (Art. 19) to add services or address circumvention |
| **UK DMCC** (2024; digital markets regime in force 1 January 2025) | CMA SMS investigation: £25bn global or £1bn UK turnover + substantial and entrenched market power + strategic significance; Google designated in general search and search advertising (Oct 2025); Apple and Google met SMS test in mobile platforms | Bespoke Conduct Requirements under fair dealing, open choices, trust and transparency; Pro-Competition Interventions | CMA writes firm-specific conduct requirements and interventions; Secretary of State can expand the permitted requirement types |
| **Japan MSCA** (2024; full effect by 18 Dec 2025) | JFTC designation of specified software operators in mobile OS, app stores, browsers, search (Apple Inc., iTunes K.K., Google LLC designated 26 Mar 2025) | App-distribution, payment-system, default-choice, self-preferencing, use-of-competing-app-data, and equal-access obligations | Covered providers and implementation details are set administratively within the statutory mobile-software categories |
| **India Draft DCB** (2024) | ‘Systemically Significant Digital Enterprises’ by dual test of financial strength and user base; related entities as ‘Associate Digital Enterprises’ | Mirrors DMA Arts. 5-6: self-preferencing, data usage against business users, anti-steering | CCI can specify obligations for each core digital service and designate firms beyond the thresholds using qualitative criteria |
| **Australia** (Treasury proposal Dec 2024) | ACCC designation investigation per service category, no automatic thresholds; initial focus on app marketplaces and ad-tech | Broad obligations in primary legislation plus service-specific rules in subordinate legislation | Service-specific rules follow designation investigations; compliance measures from overseas regimes may be recognised in Australia |
| **Brazil Bill 4675/2025** | Global revenue above BRL 50bn or local revenue above BRL 5bn; designation approved by CADE’s Tribunal, with a new Digital Markets Superintendency within CADE instructing, monitoring, and investigating | Transparency, interoperability, data portability, non-discrimination, and merger-related oversight | CADE would impose tailored obligations on each designated agent of systemic relevance, alongside general statutory prohibitions |
| **Turkey** (draft amendments to Law No. 4054) | Firms with significant influence over end users and entrenched position | DMA-style prohibitions; broad interoperability across core platform services; mandatory merger notification | Competition authority would designate covered undertakings and apply the new obligations through the existing competition-law statute |
Ex ante digital platform competition regimes. Status as of April 2026.
The regimes vary considerably in design but share a common structure of designating firms as members of a regulable class and attaching conduct obligations to them. The EU’s DMA is the most prescriptive and broad, with self-executing rules (Article 5) and further-specifiable obligations (Article 6) that apply uniformly across designated gatekeepers. The UK’s DMCC is more bespoke, allowing the Competition and Markets Authority to write tailored Conduct Requirements for each firm given Strategic Market Status. This follows the UK pattern Kretschmer et al ([2022](#Xc6f7433aad5d664ff9eedccaa5467a50198ceb0)) identify: discretionary identification of strategically significant or high-risk platforms, combined with flexible codes of conduct or practice that sit between self-regulation and state enforcement. Japan has moved over four years from a soft ‘transparency and dialogue’ regime to hard *ex ante* prohibitions in mobile software. Australia’s proposed regime is closer to a firm-by-firm design. Rather than relying on automatic thresholds, it would require a designation investigation by the Australian Competition and Consumer Commission for each service category. The variation between the regimes is in how prescriptive the obligations are and how much room the regulator keeps to tailor them, not in whether designating firms and attaching obligations is the right way to govern the object.
The most revealing structural feature of these regimes is that each contains explicit mechanisms for staying ahead of the markets it regulates. The DMA grants the European Commission delegated-act powers (Article 12) and market-investigation powers (Article 19); the DMCC lets the UK Secretary of State amend permitted requirement types by regulation (Section 20(4)); Japan’s Cabinet Order mechanism extends platform regulation without amending the Act; Australia conducts a fresh designation investigation for each service category. These adaptive mechanisms are an implicit acknowledgment of the mismatch we develop theoretically in the sections that follow – the regulatory form requires the object it governs to remain stable enough to be governed, but builds in flexibility precisely because it cannot.
# Dynamic capabilities theory with Austrian microfoundations
Dynamic capabilities theory is a theory of how firms create and sustain competitive advantage in environments where advantage cannot be locked in ([Teece et al. 1997](#X3a58498b9494ef9a0598fad17d302d6a6936125); [Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)). A firm’s ability to compete is not the static quality of its asset base but its capacity to continuously re-shape that base as conditions change. The framework grew out of the evolutionary-economics tradition that treats firms as bundles of routines subject to search, selection, and adaptation rather than as optimising agents at equilibrium ([Nelson and Winter 2004](#ref-nelsonEvolutionaryTheoryEconomic2004)), and has become one of the most-cited theoretical apparatuses in strategic management for explaining differential firm performance under change ([Helfat 2007](#X099679ab523d92f992d633d252db7040c7155b6)).
The capability was first characterised as “the ability to integrate, build, and reconfigure internal and external competences to address rapidly changing environments” ([Teece et al. 1997, 516](#X3a58498b9494ef9a0598fad17d302d6a6936125)), and later decomposed by Teece into three capabilities operating on different time horizons: sensing, seizing and reconfiguring ([Teece 2007](#X12b3885ca8631fa292576f0dc2c4110c8f462c3)). Sensing is the identification of emerging opportunities and threats – the work of scanning technological and competitive conditions for what others have not yet seen. Seizing is the commitment of resources to a sensed opportunity – the decision to invest, build, or acquire that turns recognition into position. Reconfiguring is the ongoing rearrangement of the firm’s asset base, routines, and organisational structure as new seizings succeed or fail and as competitive conditions shift.
This connection to competition policy is not incidental. Teece’s profiting-from-innovation argument turns on complementary assets: innovators capture value only when they can assemble or access the manufacturing, distribution, platform, standards, data, or ecosystem assets needed to make an invention commercially effective ([Teece 2018](#ref-teeceProfitingInnovationDigital2018)). But the strategic moves by which firms secure those complements — integration, exclusive dealing, acquisition, bundling, platform governance — are precisely the moves competition law is tempted to read as foreclosure, so the theory of innovation already contains the regulatory problem.
Teece’s original formulation has been refined and critiqued across several dimensions that bear on our argument ([Helfat 2007, 4, 7](#X099679ab523d92f992d633d252db7040c7155b6)). One alternative treats dynamic capabilities as specific, identifiable organisational processes — product development, alliancing, strategic decision making — whose value lies in the resource configurations they produce rather than in the capabilities themselves, and whose character varies with the rate of environmental change ([Eisenhardt and Martin 2000](#X3a25a54064cdff65d3a48024c5eb5b02882635e)). Another locates dynamic capabilities in “a learned and stable pattern of collective activity” through which an organisation systematically modifies its operating routines, built up through experience accumulation, knowledge articulation, and knowledge codification ([Zollo and Winter 2002, 340](#ref-zolloDeliberateLearningEvolution2002)). The framework has also drawn ongoing critique across its successive formulations on grounds of tautology, conceptual vagueness, weak empirical grounding, and unclear value added relative to established concepts such as absorptive capacity and organisational learning ([Arend and Bromiley 2009](#Xcf88887f27006afb239633dc77091d122a6861c)).
The dynamic capabilities framework already carries an implicit critique of regulatory forms that treat firms as stable objects. If what makes a digital platform competitive is its capacity to continuously sense, seize, and reconfigure, then a regulatory designation that fixes the firm’s identity at a moment in time cannot attach to anything durable. Teece and Kahwaty ([2021](#ref-teeceProposedDigitalMarkets2021)) draw this implication directly against the DMA at the level of innovation outcomes, arguing that *ex ante* obligations will blunt the investment necessary for dynamic adaptation. The contribution of Austrian foundations locates the difficulty one level deeper, in the epistemic position of the regulator facing capabilities whose very operation produces the object those regulators are trying to govern.
Foss and Ishikawa ([2007](#ref-fossDynamicResourcebasedView2007)) diagnosed the resource-based view as “a patched-up competitive equilibrium model” that could not explain how resource bundles come into being, how they are revalued, or how entrepreneurial activity produces new combinations ([Foss and Ishikawa 2007, 752](#ref-fossDynamicResourcebasedView2007)). They imported three Austrian concepts — Lachmannian capital heterogeneity, Knight-Mises judgment under uncertainty, and Kirznerian alertness — as microfoundations for a dynamic version of the view. They draw strategy-theoretic implications from the import: sustainability of advantage, pricing of resources, the non-priceability of entrepreneurial judgment. They do not consider the implications for regulation.
We take their Austrian import set as established and lean on them for the work of showing that these three concepts belong in strategic management theory. The resource-based view they applied an Austrian lens to is a static theory of the firm – an inventory of asset stocks. The dynamic capabilities framework is a theory of capability operations: of activities the firm carries out over time, and at which *ex ante* regulatory obligations bind. Capability, not resource, is therefore the right level at which to mount an Austrian critique of these regimes. We extend the Foss-Ishikawa import to that level, mapping each Austrian concept onto a distinct moment of Teece’s triad – sensing onto Kirznerian alertness, seizing onto Knight-Mises judgment, reconfiguring onto Lachmannian capital re-composition. Three structurally distinct impossibilities for an outside observer follow, one per triad moment, which Foss and Ishikawa’s framework does not surface.
| **Dynamic Capabilities Capability** | **Austrian Concept** | **Regulatory Impossibility** |
|————————————-|—————————|————————————|
| Sensing | Kirznerian alertness | Opportunities cannot be enumerated |
| Seizing | Knight-Mises judgment | Commitment cannot be priced |
| Reconfiguring | Lachmannian recomposition | Effects cannot be predicted |
The firm-level Austrian terrain on which this analysis sits has further predecessors. Langlois ([1995](#ref-langloisFirmsPlan1995), [2003](#Xd74817c50a78ccd1a3e92245d6bca1eb8a2f994)) read Hayekian dispersed-knowledge dynamics inside the firm and argued that firm boundaries are themselves remade as capabilities spread; Foss and Klein ([2012](#Xc97a928caffdb7a0a98573c19c40aac90a9901a)) developed a judgment-based theory of the firm in which entrepreneurs exercise Knight-Mises judgment over heterogeneous capital. We rely on their establishment of the firm as a site of Austrian market-process activity, but neither has run the triad-structured epistemic critique of *ex ante* regulation that we develop below.
## Sensing as Kirznerian alertness
Teece describes sensing as the firm-level activity by which new opportunities are identified and shaped – the continuing work of scanning technological, competitive, and customer environments for what has not yet been seen ([Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)). This sensing is distributed across the firm, depends on accumulated perception and exposure to markets and technologies, and generates opportunities that are not available in the same form to other firms facing the same environmental conditions. Kirzner’s entrepreneurial alertness helps us formalise this. Alertness is the pre-analytical orientation that lets an agent notice an opportunity in the first place ([Kirzner 1978](#X692379ffa3c8166628ad2fae4197b1a8a2b8cbf), [1997](#X16278335d88f8c505524a2fa118d3981e0a25d8)). For Kirzner alertness necessarily includes creative acts of perception whose content is not pre-given: what alertness sees depends on the agent doing the seeing, and the opportunity exists as an opportunity only once it is seen ([Kirzner 1999](#Xfa9e40d89005189c4a217dc5292381044e66c70)).
This sense-alertness mapping has an implication that matters when we return to regulation. Opportunities identified through sensing do not exist, at the moment of identification, as observable demand or behaviour. The observer cannot enumerate what the firm might yet notice. This asymmetry is a structural feature of alertness, now operating at the firm level as sensing. The regulator cannot enumerate what the firm might yet notice because the opportunity does not exist as an observable object until the firm’s alertness has constituted it as one.
## Seizing as choice under uncertainty
Seizing is both where alertness becomes commitment and where the regulator’s second impossibility arises. Teece’s *seizing* is the mobilisation of resources to address the sensed opportunity, which involves choices about investment, business model, organisational design, governance, and pricing ([Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)). What makes the commitment difficult is that it has to be made under conditions in which the probability distribution of outcomes is not available. For Knight ([1921](#ref-knightRiskUncertaintyProfit1921)) and Mises a risk is the situation in which outcomes are unknown but their probabilities are known or calculable, uncertainty is the situation in which the probabilities themselves are unknown. Risk is insurable because large numbers smooth it; uncertainty is not, because the underlying distribution cannot be estimated. Profit is the reward for bearing uncertainty. The outcome of a platform’s decision to commit capital to a generative AI service will be produced in part by the very actions the commitment participates in.
Foss and Ishikawa ([2007](#ref-fossDynamicResourcebasedView2007)) establish why judgment cannot be priced from outside the firm. They imported the same Knight-Mises move into strategic management as one of the microfoundations of a dynamic resource-based view. Firms exist because judgment under uncertainty cannot be priced on markets. An agent cannot sell her judgment because the buyer cannot specify what she is buying. The firm is the institutional form in which judgment is exercised and the resources under its direction are reshuffled in response. Seizing is the triad location of the same move: the capability to bear case-probability uncertainty and to act under it. When a firm commits capital to a cloud infrastructure build or restructures around a new pricing model, it is exercising judgment where no basis for calculation is available.
The consequence for the argument that follows is that seizing cannot be priced from outside the firm. A regulator who assesses the opportunity cost of an *ex ante* obligation — the investments the platform will not make because of it, the business models it will not try because of it — faces the same impossibility the market faces when it tries to price judgment directly. Impact assessment treats uncertainty as risk, and so produces numbers where no numbers should exist.
## Reconfiguring as heterogeneous capital reconstruction
Reconfiguring, in Teece’s triad, is the capability by which the firm continuously rearranges its asset base, routines, and organisational structure as sensed opportunities become seized commitments and as the environment keeps moving ([Teece 2023](#X10c9d56f264a630702c6edb714dfd2a69f71649)). A firm that senses well and seizes well but cannot reconfigure will find its capital structure progressively misaligned with the commitments it has entered into. A firm that reconfigures well sustains its capability set across the changes. Reconfiguring is the moment at which the dynamic capabilities framework is most visibly a theory of ongoing change rather than of static advantage.
Against the prevailing view of capital as a homogeneous fund measurable in a single stock, Lachmann ([\[1956\] 1978](#ref-lachmannCapitalItsStructure1978)) argued that capital is a *structure* of heterogeneous goods. Each capital good has what he called ‘multiple specificity’: it can be used in more than one way but not in all ways. A server farm can run consumer email, enterprise cloud, AI inference, or ad-tech processing, but it cannot refine oil. A research team can work on natural language, computer vision, or recommendation systems, but not marine biology. The set of uses to which a capital good can be put is determined by its physical and technical properties. Which of those uses it is actually put to is determined by the entrepreneurial plans of the firms that deploy it.
Plans are what hold capital structures together. Two servers in a data centre are *complements* in a capital structure only because the plan that deploys them treats them as such. Change the plan – move from consumer email to AI inference, for instance – and the complementarities change: different software, different network topology, different staff, different customer base, different billing infrastructure. The physical capital goods may be the same, but the capital *structure* is not. The structure is whatever combination of goods is currently being deployed by a given plan, and it lasts only as long as the plan lasts.
Because plans succeed and fail, and because entrepreneurial expectations diverge, capital structures are continuously being dissolved and re-formed. Lachmann’s claim was that equilibrium at the capital-structure level is impossible, because different entrepreneurs holding different expectations require different complementary combinations of the same capital goods. What an outside observer sees as the firm at a moment in time is a snapshot of a capital structure whose form is being continuously produced by the activity of those inside the firm. The firm’s boundaries, its asset composition, its organisational routines – all are outputs of ongoing re-composition rather than inputs to it ([Lewin and Baetjer 2011](#ref-lewinCapitalbasedViewFirm2011); [Lewin and Cachanosky 2019](#ref-lewinAustrianCapitalTheory2019)). Through an evolutionary lens, Langlois ([1995](#ref-langloisFirmsPlan1995), [2003](#Xd74817c50a78ccd1a3e92245d6bca1eb8a2f994)) describes firms as plans whose form is produced and reproduced by the conditions of industrial change. Reconfiguring is not occasional corporate restructuring. It is the ongoing work by which the firm’s capital structure is kept in alignment with the plans that animate it.
The implication for regulation, which we consider in section 5, is that the object a regulator designates at time *t* is not the object on which the regulator’s obligations will eventually be placed. Between designation and enforcement, the capital structure under the designated name will have moved. The structural parameters the regulator used to predict the obligation’s effect were themselves products of the pre-regulation capital structure, and that structure will not survive the regulation unchanged. In Austrian capital theory this is simply what capital structures do.[^3]
This Austrian-microfounded reading of the triad presents three distinct knowledge problems. Sensing-as-alertness presents an *epistemic* impossibility: the object the regulator would need to enumerate — the opportunity not yet seen — does not exist as observable demand at the moment of identification. Seizing-as-judgment presents a *categorial* impossibility: the operation the regulator would need to perform — pricing a firm’s commitment under genuine uncertainty — is not the kind of operation that can be carried out at all, because judgment is not the kind of thing on which calculation has purchase. Reconfiguring-as-recomposition presents a *reflexive* impossibility: the act of regulating changes the structural parameters the regulator used to predict its own effects, so the predictor and the predicted cannot be separated. Three different missing things — the object, the operation, the separability of predictor from predicted — yield three different kinds of regulatory impossibility.
# The recombinatorial digital platform
The platforms that are the object of *ex ante* regulation are ongoing processes of organising whose apparent stability is produced by continuous micro-processes of adjustment and recomposition – not stable entities that occasionally undergo change (Tsoukas and Chia [2002](#X39c98bc7775db865b52072d398992ab5396a9c8)). Apple, Google, and Meta at time *t* are real objects with measurable properties, legally continuous identities, and stable-looking organisational forms. But those properties are produced by the ongoing capability activity of the firms – by the sensing, seizing, and reconfiguring we have now developed. The apparent stability of a successful platform across years and decades is a product of the regularity with which its capability activity reproduces its form, not a feature of the form itself – though we do not go so far as to treat platforms as pure flux whose description is always out of date ([Shaw 2019](#X1ecf1101b695199242cc4484db5c76846b3c632)). This point is sharpened by the fact that many of the relevant capabilities are not located wholly inside the legal boundary of the firm – they are distributed across developer ecosystems, complementors, advertisers, users, APIs, standards, data flows, and contractual interfaces that the platform orchestrates but does not simply own. The platform’s capability set is therefore a relation among capital goods and complementary actors, not only an internal stock of assets, which means regulation that designates the firm necessarily reaches into a moving field of external recombination. We develop this argument through Microsoft, whose repeated strategic recomposition across five decades makes the gap between legal continuity and economic identity especially clear.
## The Microsoft case
Here we consider Microsoft as an exemplar of the recombinatorial digital platform. Founded in 1975, Microsoft is unusually old among the firms now targeted by *ex ante* digital platform regimes. It predates Apple and Oracle only narrowly, but it predates Amazon, Google, Meta, and the mature digital-platform era by decades. The endurance of Microsoft as a corporate institution is therefore easy to misread as continuity of form. We focus on Microsoft for three reasons. First, it is unusually old among designated platforms, giving us two decades of post-browser-war recomposition to examine. Second, it has been subject to major competition interventions across multiple eras, making the regulatory problem concrete. And third, its trajectory from packaged software to cloud to AI infrastructure illustrates each moment of the capability triad. The capital structure under the Microsoft identity has, accordingly, been continuously reworked.
The scholarly starting point for Microsoft’s recomposition is the browser war, which Bresnahan et al ([2011](#X599bd885019b63ef6f829edae82aea37b83077e)) take as the canonical case of what they call *diseconomies of scope* – dominant incumbents struggle in new technological eras not because they fail to see the wave, not because they fear cannibalisation, and not because they lack capability, but because necessarily shared firm-wide assets impose costs across old and new lines of business that a de novo entrant does not bear. Their account stops at the first years of the 2000s. The last twenty years make the re-composition especially legible, and extend the pattern they identified well beyond the browser war.
A snapshot of Microsoft in 2005 reveals not one business but a portfolio of distinct business models under a single corporate shell. In 2005 the company reported seven major business segments: client operating systems, server and tools, information worker products, business solutions, MSN, mobile and embedded devices, and home and entertainment. Each ran on a different revenue model: Windows through OEM pre-installation, Office through volume and enterprise licensing, Business Solutions through partner channels, mobile software into operator and OEM ecosystems, and Xbox sitting within a loss-making Home and Entertainment segment supported by Microsoft’s broader corporate architecture.
By the late 2000s the firm was reorganising around a fundamentally different business architecture ([Nadella 2018](#ref-nadellaHitRefreshQuest), ch. 2). Amazon Web Services had demonstrated the commercial logic of cloud infrastructure, mobile computing was changing the competitive environment of software, and Microsoft was moving toward, but had not yet fully commercialised, its own cloud platform. By 2011 management was naming “execution and competitive risks in transitioning to cloud-based computing” and “challenges to Microsoft’s business model” in its investor disclosures, alongside a new stack — Office 365, SQL Azure, Windows Azure — whose logic differed from anything in the 2005 architecture. In 2014 Satya Nadella committed the firm “to thrive in a mobile-first and a cloud-first world” ([Nadella 2018](#ref-nadellaHitRefreshQuest), ch. 3). The 2015 annual report recast the firm as building “platforms and productivity services” for that world, with Office 365 seat growth and Azure revenue and compute usage as the reported engines.[^4] The 2016 report recorded the other side of re-composition: multi-billion-dollar write-downs on the Nokia handset business, a transaction announced in Ballmer’s final months and completed in 2014 under Nadella, and the abandonment of businesses that no longer fitted the new architecture ([Nadella 2018](#ref-nadellaHitRefreshQuest)).[^5]
In 2016 Microsoft joined the Linux Foundation and announced its acquisition of LinkedIn, the professional social networking platform. Two years later it acquired GitHub, the global platform for hosting development code. The Linux decision required abandoning the firm’s long-standing view that open-source software was the enemy ([Nadella 2018](#ref-nadellaHitRefreshQuest), ch. 2). What looked from outside like an open-source turn was better read as repositioning around professional identity, developer workflow, and cross-platform complements as the new control surfaces. This is precisely the Bresnahan-style move the scholarly literature had described at the browser war: the firm’s locus of strategic control shifts, under competitive pressure, to whatever asset is capable of bearing the new plan. We can see this pattern continuing in the post-transformers artificial intelligence era. The 2019 OpenAI partnership made Azure central to Microsoft’s generative-AI strategy; in 2023 Microsoft described Azure as OpenAI’s exclusive cloud provider across OpenAI workloads, and by 2025 that exclusivity had narrowed, with Azure retaining exclusivity for OpenAI APIs and Microsoft holding rights over additional capacity.
Across twenty years the firm moved from packaged software to enterprise subscriptions, from desktop to cloud, from proprietary stack to open-source-adjacent ecosystem, and from cloud subscriptions to AI infrastructure – each shift re-organising what the firm sold, how value was captured, and which assets counted as strategically central. Bresnahan and co-authors are careful to note that none of this can be read as a series of *ex ante* mistakes: the firm in each era faced different costs and benefits of entering and operating in the new market than a de novo entrant, and it is “very hard to make the case, of course, that *ex ante* this was ‘irrational’” ([Bresnahan et al. 2011, 13](#X599bd885019b63ef6f829edae82aea37b83077e)). The Austrian reading accommodates this — re-composition is entrepreneurial judgment under uncertainty, not error ([Foss and Klein 2012](#Xc97a928caffdb7a0a98573c19c40aac90a9901a)) — and locates the significance of the re-composition in what it implies about the object a regulator believes it has designated.
The Microsoft of 2005 and the Microsoft of 2025 are the same legal entity but are not the same capital structure. The continuity of name across decades is an artefact of legal personhood and brand; their position within the technology industry was produced by continuous sensing, seizing, and reconfiguring. What would have happened to Microsoft under a DMA-style designation in 2005, with obligations drafted against its then-segment architecture, is not a question about how the firm would have borne the obligations. It is a question about whether the firm that bore them would have become the firm that exists today.
Capital-structure recomposition is general to firms with dynamic capabilities, but what is distinctive about digital platforms is the rate of recomposition and the openness of the innovation space in which it occurs. Other recomposable sectors are anchored differently: banking by political-financial entanglements that predate any one regulator ([Calomiris and Haber 2014](#ref-calomirisFragileDesignPolitical2014)); pharmaceuticals by patent dynamics and the regulatory architecture of drug approval; telecommunications and electricity by physical-infrastructure constraints that change at a different pace from anything in the digital economy. The *ex ante* digital platform regulatory form is in turn unusually categorical: it is not a third-party-access regime, not capital-adequacy, not rent-extraction. It is an unusually fluid object meeting an unusually categorical regulatory form.
# *Ex ante* regimes and the recomposing platform
The *ex ante* digital platform competition regimes described in section 2 do not regulate “platforms” in the abstract. They regulate through a small number of recurring instruments such as interoperability mandates, anti-self-preferencing rules, data-access and portability obligations, anti-tying rules, app-store and payment-system constraints, and bespoke conduct requirements. Each instrument begins by identifying a component of the platform economy as the object of regulatory action. Our analysis asks what that component is once it is read not as a stable legal object but as a moving element in a capital structure.
The test has three steps. First, identify the regulator’s object: the interface, ranking practice, default, dataset, app-store dependency, payment flow, or conduct category the rule appears to target. Second, redescribe that object in capital-structure terms: what role does it play inside the sensing, seizing, and reconfiguring activity of the platform? Third, ask where recomposition can occur once the obligation binds. The legal rule needs a component that can be named in advance and governed across the designation period. What our analysis shows is that the platform component named by the rule is instead one element in a changing capital structure whose value and function are determined by the plan that deploys it – and that plan does not survive the obligation unchanged. Table 3 below summarises our application of this test to the three core instruments of the new regimes. Each row identifies what the regulator takes to be the object of regulatory action, how that object is better described in capital-structure terms, and which structural impossibility follows.
| **Instrument** | **Regulator’s assumed object** | **Capital-structure reality** | **Impossibility type** |
|—————————–|———————————————————————————————-|——————————————————————————————————————-|————————|
| Interoperability mandates | A concrete technical boundary that can be opened while leaving the competitive object intact | A boundary at which heterogeneous goods are made complementary by the plan that deploys them | Reflexive |
| Self-preferencing rules | A ranking or default advantage that can be neutralised by prohibition | A visible settlement of a broader plan about which complements matter and how attention is allocated | Epistemic |
| Data access and portability | A dataset or data flow that constitutes the bottleneck and can be shared or ported | A capital good whose value is produced by its combination with models, infrastructure, and complementary services | Categorical |
The instruments vary in legal form across jurisdictions but share a common analytical structure. In the DMA they are mostly directly applicable obligations imposed on designated gatekeepers; in the UK they are powers the CMA may translate into firm-specific conduct requirements or pro-competition interventions; in Japan they are smartphone-software-specific statutory duties; and in India, Australia, and Brazil they remain draft or proposed obligations. The common feature is not identical legal form but the attempt to identify components of platform power in advance and attach continuing regulatory duties to them.
## Interoperability and interfaces
Interoperability is the cleanest case because it appears to regulate a concrete technical boundary such as an API, a protocol, an app-store dependency, a browser-engine constraint, a messaging channel, or an operating-system function. These regimes proceed from a common intuition – if a platform controls a bottleneck interface, the regulator can require access, interoperability, portability, or non-discriminatory use of that interface.[^6]
But the interface is not just an interface ([Berg 2024](#ref-bergInteroperability2024)). In capital-structure terms it is a boundary at which heterogeneous goods are made complementary: code, documentation, security review, identity management, developer support, billing, ranking, fraud detection, privacy architecture, and commercial terms. The interface works because the platform’s plan makes those things fit together. Opening the interface materially changes the complementarities that made the gate valuable.
The 2004 European Commission Microsoft decision illustrates the problem directly. The decision required Microsoft to disclose interoperability information sufficient to allow non-Microsoft work group server operating systems to interoperate with Windows PCs and servers, and to offer a version of Windows without Windows Media Player.[^7] The interoperability remedy treated server-to-server interoperability as a dependency that could be opened while leaving the rest of the competitive object substantially intact. But Microsoft’s capital structure was already moving. The server business was reorganising toward cloud services, media distribution was moving away from desktop preinstallation, and the control surface that had mattered in the Windows era was losing centrality. The remedy attached to an interface whose strategic meaning was being revalued by the firm’s own recomposition *during the process of regulatory intervention*.
A DMA-style interoperability rule does not freeze the surrounding capital structure. A platform can recompose around identity, safety certification, developer tooling, ranking, payment flow, cloud integration, default settings, or AI-mediated user interaction. Some of those movements may be pro-competitive; some may be anti-competitive; some may be neutral. The regulator’s object is a moving boundary, and the value of the boundary is endogenous to the capital structure the obligation changes.
## Self-preferencing and platform hierarchy
Self-preferencing rules target one visible form of platform hierarchy: the use of control over intermediation to favour the platform’s own services. In the DMA this appears most directly in Article 6(5), which concerns favourable treatment in ranking and related indexing and crawling. Other features of platform hierarchy — defaults, bundling, app-store design, payment dependencies, and access to technical functionality — are regulated through adjacent provisions rather than by Article 6(5) itself. The UK DMCC similarly authorises the CMA to impose conduct requirements preventing discriminatory treatment and self-preferencing by SMS firms, but only through firm-specific requirements proportionate to the statutory objectives.[^8] Analytically, however, these provisions all intervene in the platform hierarchy through which the firm organises complements: what appears first, what is installed by default, what is bundled with what, and which internal services receive privileged treatment.
The logic of self-preferencing rules depends on a static account of platform hierarchy that the capability framework contests. Separate the platform’s neutral intermediation function from its downstream services and stop the former privileging the latter. The difficulty is that platform hierarchy is rarely just a ranking decision. It is the visible settlement of a broader plan about which complements matter, which user journeys are valuable, where trust should be located, how attention should be allocated, which transactions should be internalised, and which assets are to be made mutually specific. To prohibit one form of preference is to alter the plan and invite a new one.
Microsoft’s browser and media-player cases illustrate the broader platform-hierarchy problem, though not self-preferencing in the modern DMA sense. The legal categories were tying, bundling, monopoly maintenance, and refusal to supply interoperability information; analytically, however, the cases concerned the use of an operating-system control surface to shape adjacent software layers. As Bresnahan et al’s ([2011](#X599bd885019b63ef6f829edae82aea37b83077e)) account of the browser war shows, the more important strategic problem was not simply that Microsoft favoured its own browser. It was that the Windows-centred capital structure imposed diseconomies of scope on the firm’s attempt to enter and organise around the Internet. The platform hierarchy being regulated was already part of a failed recomposition: the old control surface was being defended at the same moment the next one was forming elsewhere.
A ban on self-preferencing in search, app stores, marketplaces, or mobile operating systems may remove a visible ranking or default advantage. But the platform can recompose the hierarchy around other assets: quality thresholds, reputation systems, fulfilment integration, identity, payments, advertising tools, subscription bundles, cloud credits, developer analytics, or AI assistants that sit above the ranked list. While the regulator sees a preference at the surface the platform responds by changing the architecture in which preference is produced. The obligation’s effect cannot be inferred from the pre-regulation hierarchy, because the hierarchy is one output of a plan that the rule itself changes.
## Data access and portability
Data access and portability obligations make the separability assumption most visible. For instance, the DMA contains direct obligations concerning data combination, business-user non-public data, advertising transparency and measurement, end-user portability, business-user data access, and search-data access. Other regimes including the UK, India, Australia and Brazil have similar requirements.[^9] The regulator’s object is a dataset or data flow – depending on the service and legal instrument, a social graph, transaction history, advertiser information, app-store data, search data, mobility record, business-user analytics, or user account history. If the dataset is the bottleneck, the remedy is to share it, port it, or prevent the gatekeeper from using it against business users.
Data does not have a stable regulatory identity apart from the plan that makes it valuable. Data is a capital good only when it is combined with models, classification systems, user interfaces, compute infrastructure, business routines, privacy architecture, contractual permissions, and complementary services. The same transaction record can be advertising input, fraud signal, merchant analytics, personalisation asset, credit signal, logistics input, or training data depending on the plan that deploys it. Portability moves a copy of some data but does not move the capital structure that made the data valuable.
Microsoft’s recent recomposition around Azure, GitHub, OpenAI, and Copilot shows value moving from software licences and desktop defaults into cloud infrastructure, developer workflow, model access, and enterprise context. In such a structure, the strategic asset is not only a dataset that can be ported. It is the complementarity among compute, model, distribution, identity, workflow, security, and organisational adoption. A portability rule written against one layer of data can leave the value of the system to recompose around inference, integration, or context.
The same point applies beyond Microsoft, and connects with the platform-ecosystem literature’s emphasis on architecture, governance, complementors, and data-enabled reductions in search, tracking, and coordination costs ([Gawer and Cusumano 2002](#ref-gawerPlatformLeadershipHow2002); [Tiwana 2014](#ref-tiwanaPlatformEcosystemsAligning2014); [Goldfarb and Tucker 2019](#ref-goldfarbDigitalEconomics2019); [Teece 2018](#ref-teeceProfitingInnovationDigital2018)). A social graph is valuable inside an advertising, identity, messaging, and recommendation architecture. App-store data is valuable inside a developer, payment, review, and device-security architecture. Marketplace data is valuable inside fulfilment, payments, advertising, ranking, and merchant-service architectures. The harder the rule tries to specify what must be shared, the more it depends on a pre-regulation account of where value sits in the platform.
Each instrument fails at a different moment of the capability triad. Interoperability mandates attach to interfaces whose strategic value is endogenous to the capital structure the obligation changes – a reflexive impossibility, because the regulator’s prediction of the rule’s effect does not survive the act of regulating. Self-preferencing rules target a hierarchy that is one visible output of a broader plan, and prohibiting one form of preference invites a recomposition the regulator cannot enumerate in advance – an epistemic impossibility, because the opportunities the platform will next sense are not present as observable demand at the moment the rule is written. Data access and portability obligations assume that the value of a dataset is separable from the capital structure that makes it valuable – a categorial impossibility, because the judgment by which data is combined with models, infrastructure, and complementary services cannot be priced from outside the firm. The regulatory form fails at the level of the assumption each instrument shares – that the component it names is stable enough to bear a continuing obligation.
# Conclusion
This paper has located the failure of *ex ante* digital platform regulation one level deeper than existing dynamic capabilities critiques. The error is not that the rules are too static, or that regulators forecast poorly – it is that the regulatory form mistakes legal continuity and observed market position for stability of the economic object being governed. Platforms are not stable objects temporarily subject to competitive pressure. They are ongoing processes of capital recomposition whose apparent durability is produced by the capability activity that sustains them, and that activity does not pause for designation.
The dynamic capabilities literature has critiqued these regimes on consequentialist grounds, such as whether innovation will suffer, investment will fall, long-run welfare will drop. Those claims are plausible and important. What this paper has found is a critique that does not depend on them. We have shown that each moment of the capability triad can be given Austrian microfoundations: sensing as Kirznerian alertness, seizing as Knight-Mises judgment under genuine uncertainty, reconfiguring as Lachmannian recomposition of heterogeneous capital.
The three impossibilities that follow are structurally distinct in kind – not three flavours of one knowledge problem but three different missing things. Sensing cannot be enumerated from outside, because alertness generates opportunities that are not present as observable demand: the object the regulator would need to enumerate does not yet exist. Seizing cannot be priced from outside, because judgment operates under genuine uncertainty rather than calculable risk: the operation the regulator would need to perform cannot be carried out at all. Reconfiguring cannot be predicted from outside, because the capital structure under regulation is continuously recomposed by the very plans the regulation is trying to constrain: the act of regulating changes the parameters the regulator relied on to predict its own effects. Each is a structural impossibility rather than an empirical difficulty, and none of the three has been established in the dynamic capabilities literature in the form we give them here.
It does not follow that no regulation of digital platforms is coherent. *Ex post* antitrust has the well-known problems of delay and irreversibility that drove the move to *ex ante* in the first place, and this paper does not revive it as a finished alternative. What the analysis does establish is a necessary condition for any coherent regulatory form: it must treat the platform as what the capability framework and the capital-theoretic reading together show it to be – an ongoing process of capital recomposition whose form is produced by the very activity the regulator is trying to govern. A regulatory form that satisfies that condition will look very different from the designation-plus-obligations architecture now proliferating across jurisdictions. It will not fix the firm’s identity at a moment in time, attach obligations to components whose value is endogenous to the plan that deploys them, or assume that the structural parameters used to forecast a rule’s effect will survive the act of regulating. What it will look like in positive terms remains an open question – but the analysis here establishes what it cannot look like, and that is where any serious alternative must begin.
# References
Akman, Pinar. 2022. “Regulating Competition in Digital Platform Markets: A Critical Assessment of the Framework and Approach of the EU Digital Markets Act.” *European Law Review* (Rochester, NY) 47. <https://doi.org/10.2139/ssrn.3978625>.
Arend, Richard J., and Philip Bromiley. 2009. “Assessing the Dynamic Capabilities View: Spare Change, Everyone?” *Strategic Organization* 7 (1): 75–90. <https://doi.org/10.1177/1476127008100132>.
Berg, Chris. 2024. “Interoperability.” *Internet Policy Review* 13 (2). <https://policyreview.info/glossary/interoperability>.
Böhm-Bawerk, Eugen von. 1890. *Capital and Interest: A Critical History of Economical Theory*. Macmillan and Company.
Bresnahan, Timothy, Shane M. Greenstein, and Rebecca M. Henderson. 2011. “Schumpeterian Competition and Diseconomies of Scope; Illustrations from the Histories of Microsoft and IBM.” SSRN Scholarly Paper No. 1752410. Social Science Research Network, January 31. <https://doi.org/10.2139/ssrn.1752410>.
Calomiris, Charles W., and Stephen H. Haber. 2014. *Fragile by Design: The Political Origins of Banking Crises and Scarce Credit*. The Princeton Economic History of the Western World. Princeton University Press.
Eisenhardt, Kathleen M., and Jeffrey A. Martin. 2000. “Dynamic Capabilities: What Are They?” *Strategic Management Journal* 21 (10/11): 1105–21. <https://www.jstor.org/stable/3094429>.
Ezrachi, Ariel, and Maurice E. Stucke. 2016. *Virtual Competition: The Promise and Perils of the Algorithm-Driven Economy*. Harvard university press.
Foss, Nicolai J., and Ibuki Ishikawa. 2007. “Towards a Dynamic Resource-Based View: Insights from Austrian Capital and Entrepreneurship Theory.” *Organization Studies* 28 (5): 749–72. <https://doi.org/10.1177/0170840607072546>.
Foss, Nicolai J., and Peter G. Klein. 2012. *Organizing Entrepreneurial Judgment: A New Approach to the Firm*. 1st ed. Cambridge University Press. <https://doi.org/10.1017/CBO9781139021173>.
Garrison, Roger W. 2000. *Time and Money: The Macroeconomics of Capital Structure*. Routledge. <https://doi.org/10.4324/9780203208083>.
Gawer, Annabelle, and Michael A. Cusumano. 2002. *Platform Leadership: How Intel, Microsoft, and Cisco Drive Industry Innovation*. Harvard Business School Press.
Goldfarb, Avi, and Catherine Tucker. 2019. “Digital Economics.” *Journal of Economic Literature* 57 (1): 3–43. <https://doi.org/10.1257/jel.20171452>.
Helfat, Constance E., ed. 2007. *Dynamic Capabilities: Understanding Strategic Change in Organizations*. Blackwell Pub.
Hovenkamp, Herbert. 2005. *The Antitrust Enterprise: Principle and Execution*. Harvard University Press.
Kirzner, Israel M. 1978. *Competition and Entrepreneurship*. University of Chicago Press. <https://books.google.com.au/books?id=E9DAt745zhIC>.
Kirzner, Israel M. 1997. “Entrepreneurial Discovery and the Competitive Market Process: An Austrian Approach.” *Journal of Economic Literature* 35 (1): 60–85. <https://www.jstor.org/stable/2729693>.
Kirzner, Israel M. 1999. “Creativity and/or Alertness: A Reconsideration of the Schumpeterian Entrepreneur.” *The Review of Austrian Economics* 11 (1): 5–17. <https://doi.org/10.1023/A:1007719905868>.
Knight, Frank H. 1921. *Risk, Uncertainty, and Profit*. Houghton Mifflin Company.
Kretschmer, Martin, Ula Furgał, and Philip Schlesinger. 2022. “The Emergence of Platform Regulation in the UK: An Empirical-Legal Study.” *Weizenbaum Journal of the Digital Society* 2 (2). <https://doi.org/10.34669/wi.wjds/2.2.4>.
Lachmann, Ludwig M. (1956) 1978. *Capital and Its Structure*. Studies in Economic Theory. S. Andrews and McMeel.
Langlois, Richard N. 1995. “Do Firms Plan?” *Constitutional Political Economy* 6 (3): 247–61. <https://doi.org/10.1007/BF01303405>.
Langlois, Richard N. 2003. “The Vanishing Hand: The Changing Dynamics of Industrial Capitalism.” *Industrial and Corporate Change* 12 (2). <https://doi.org/10.4324/9780203963630>.
Lewin, Peter, and Howard Baetjer. 2011. “The Capital-Based View of the Firm.” *The Review of Austrian Economics* 24 (4): 335–54. <https://doi.org/10.1007/s11138-011-0149-1>.
Lewin, Peter, and Nicolas Cachanosky. 2019. *Austrian Capital Theory: A Modern Survey of the Essentials*. 1st ed. Cambridge University Press. <https://doi.org/10.1017/9781108696012>.
Lucas, Robert E. 1976. “Econometric Policy Evaluation: A Critique.” *Carnegie-Rochester Conference Series on Public Policy* 1: 19–46.
Mises, Ludwig von. (1949) 1996. *Human action: a treatise on economics*. 4. rev. ed. Fox & Wilkes.
Nadella, Satya. 2018. *Hit Refresh: The Quest to Rediscover Microsoft’s Soul and Imagine a Better Future for Everyone*. HarperCollins GB.
Nelson, Richard R., and Sidney G. Winter. 2004. *An Evolutionary Theory of Economic Change*. Digitally reprinted. The Belknap Press of Harvard Univ. Press.
OECD. 2021. *Ex Ante Regulation and Competition in Digital Markets*. OECD Competition Law and Policy Working Papers No. 272. Vol. 272. OECD Competition Law and Policy Working Papers. <https://doi.org/10.1787/c83e178d-en>.
Petit, Nicolas. 2020. *Big Tech and the Digital Economy: The Moligopoly Scenario*. Oxford University Press.
Rochet, Jean-Charles, and Jean Tirole. 2003. “Platform Competition in Two-Sided Markets.” *Journal of the European Economic Association* 1 (4): 4.
Shaw, David. 2019. “On Misunderstanding Heraclitus: The Justice of Organisation Structure.” *Philosophy of Management* 18 (2): 157–67. <https://doi.org/10.1007/s40926-018-0097-y>.
Teece, David J. 2007. “Explicating Dynamic Capabilities: The Nature and Microfoundations of (Sustainable) Enterprise Performance.” *Strategic Management Journal* 28 (13): 1319–50. <https://www.jstor.org/stable/20141992>.
Teece, David J. 2018. “Profiting from Innovation in the Digital Economy: Enabling Technologies, Standards, and Licensing Models in the Wireless World.” *Research Policy* 47 (8): 1367–87. <https://doi.org/10.1016/j.respol.2017.01.015>.
Teece, David J. 2023. “The Evolution of the Dynamic Capabilities Framework.” In *Artificiality and Sustainability in Entrepreneurship: Exploring the Unforeseen, and Paving the Way to a Sustainable Future*, edited by Richard Adams, Dietmar Grichnik, Asta Pundziene, and Christine Volkmann. Springer International Publishing. <https://doi.org/10.1007/978-3-031-11371-0_6>.
Teece, David J., and Henry J. Kahwaty. 2021. *Is the Proposed Digital Markets Act the Cure for Europe’s Platform Ills? Evidence from the European Commission’s Impact Assessment*. Berkeley Research Group Institute.
Teece, David J., Gary Pisano, and Amy Shuen. 1997. “Dynamic Capabilities and Strategic Management.” *Strategic Management Journal* 18 (7): 509–33. <https://www.jstor.org/stable/3088148>.
Tiwana, Amrit. 2014. *Platform Ecosystems: Aligning Architecture, Governance, and Strategy*. MK.
Tsoukas, Haridimos, and Robert Chia. 2002. “On Organizational Becoming: Rethinking Organizational Change.” *Organization Science* 13 (5).
Zollo, Maurizio, and Sidney G. Winter. 2002. “Deliberate Learning and the Evolution of Dynamic Capabilities.” *Organization Science* 13 (3): 339–51. <https://doi.org/10.1287/orsc.13.3.339.2780>.
[^1]: RMIT University, darcywilliamallen@gmail.com
[^2]: RMIT University, chris.berg@rmit.edu.au
[^3]: Macroeconomics has noted an analogue in the response of optimising agents to policy change ([Lucas 1976](#X69e40083e8598bbd343c2193954ab165945abc3)).
[^4]: Microsoft, *2015 Annual Report*, Shareholder Letter. Nadella writes that Microsoft made decisions to become “the company that builds best-in-class platforms and productivity services for our mobile-first, cloud-first world”; the same letter reports Office 365 consumer subscriptions above 15 million, commercial seats growing by 74 percent, and Azure cloud platform revenue and compute usage increasing by more than 100 percent year over year in the fourth quarter. Available at https://www.microsoft.com/investor/reports/ar15/index.html.
[^5]: Microsoft, *2016 Annual Report*, Financial Review and Notes. The report records \$1.1 billion of impairment, integration, and restructuring expenses in fiscal 2016, including \$630 million of asset impairment charges related to the phone business and \$480 million of restructuring charges, and compares this with \$10.0 billion in fiscal 2015 expenses including \$7.5 billion of goodwill and asset impairment charges related to the phone business and \$2.5 billion of integration and restructuring expenses. Available at https://www.microsoft.com/investor/reports/ar16/.
[^6]: The DMA regulates several such boundaries: third-party app distribution and app-store access; payment, browser-engine, and identification-service dependencies; operating-system, hardware, software, and virtual-assistant feature interoperability; and messaging interoperability for designated number-independent interpersonal communications services. Japan’s smartphone-software regime does something similar within the mobile software stack, while the UK DMCC authorises the CMA, after SMS designation, to impose conduct requirements or pro-competition interventions that may address interoperability, default settings, access, and data use. These regimes proceed from a common intuition: if a platform controls a bottleneck interface, the regulator can require access, interoperability, portability, or non-discriminatory use of that interface – sometimes free of charge, sometimes on fair, reasonable, and non-discriminatory terms, and often subject to security, privacy, integrity, or proportionality qualifications.
Regulation (EU) 2022/1925 (Digital Markets Act), arts. 5(7), 6(4), 6(7), 6(9), 6(10), 6(12), and 7; recitals 50, 55–57, 59–60, 62, 64, and 96. Digital Markets, Competition and Consumers Act 2024 (UK), ss. 19–20, 46, and 51; Explanatory Notes, paras. 172, 175, 181–189, and 270–287. Japan Fair Trade Commission, “Summary of the Bill for the Act on Promotion of Competition for Specified Smartphone Software” (26 April 2024), listing third-party app stores, third-party billing systems, browser-engine constraints, default-choice screens, search self-preferencing, competing-app data use, data portability tools, disclosure of specification changes, and equal access to OS-controlled functionalities.
[^7]: Commission Decision of 24 March 2004 relating to a proceeding under Article 82 EC and Article 54 EEA Agreement, Case COMP/C-3/37.792 – Microsoft, C(2004) 900 final; published as Commission Decision 2007/53/EC, OJ L 32, 6 February 2007, 23–28. See also Microsoft Corp. v Commission, Case T-201/04, EU:T:2007:289.
[^8]: Regulation (EU) 2022/1925 (Digital Markets Act), art. 6(5), and, for adjacent platform-hierarchy instruments, arts. 5(7), 5(8), 6(3), 6(4), and 6(12). Digital Markets, Competition and Consumers Act 2024 (UK), ss. 19(5)–(7), 20(2)(e), 20(3)(a)–(b), and 20(3)(d)–(e); Explanatory Notes, paras. 172–175 and 188–189. For comparable proposed or sector-specific provisions, see India Draft Digital Competition Bill 2024, s. 11; Australian Government, The Treasury, *A new digital competition regime: Proposal paper* (December 2024); Brazil, Câmara dos Deputados, Projeto de Lei 4675/2025, proposed art. 47-E(IV)(c).
[^9]: The UK DMCC authorises the CMA to impose data-related conduct requirements and pro-competition interventions. India’s Draft Digital Competition Bill would impose data-use, consent, and portability obligations and would allow the CCI to specify conduct requirements for core digital services; Australia’s December 2024 proposal would authorise broad and service-specific obligations concerning data use, transparency, interoperability, and business-user treatment; Brazil’s Bill 4675/2025 would authorise data-transfer, interoperability, and business-user data-access obligations.
Regulation (EU) 2022/1925 (Digital Markets Act), arts. 5(2), 5(9)–(10), 6(2), 6(8), 6(9), 6(10), and 6(11). Digital Markets, Competition and Consumers Act 2024 (UK), ss. 20(2)(c)–(d), 20(3)(g), 46, and 51; Explanatory Notes, paras. 194 and 270–287. Ministry of Corporate Affairs, Government of India, *Report of the Committee on Digital Competition Law* (27 February 2024), Annexure IV, Draft Digital Competition Bill 2024, ss. 9, 10, and 12. Australian Government, The Treasury, *A new digital competition regime: Proposal paper* (December 2024), 18–23, 27–28. Brazil, Câmara dos Deputados, Projeto de Lei 4675/2025, proposed arts. 47-E(II)(a)–(c), 47-E(IV)(c), 47-E(IV)(f), 47-E(V)(a), 47-E(V)(b), and 47-E(V)(d).
Dr Darcy Allen 