Creative destruction in digital ecosystems revisited

For a long time, a prevailing view was that due to the dynamic nature of competition in digital markets intervention by antitrust authorities would do more harm than good. It was assumed that the permanent risk of disruptive innovations, Joseph Schumpeter’s “perennial gale of creative destruction,”[1] would sufficiently discipline any digital incumbent. Even a position of dominance was presumed to be inherently temporary as rivals were only “one click away,” switching costs low, and network effects reversible. Accordingly, when OpenAI/Microsoft launched ChatGPT in November 2022, expectations went viral that “Google is done,” only to witness Google counter with its own chatbot Bard within just three months. This article outlines why many assumptions relating to the disciplining force of potential disruptions are no longer sustainable in digital ecosystems with incontestable moats. 

1. Creative destruction in digital ecosystems

A. Schumpeter’s innovation theory

Proponents of a laissez-faire approach in digital markets have frequently turned to the theories of Joseph Schumpeter on “creative destruction.” Schumpeter believed that competition is driven less by the number of buyers and sellers or elements such as the price or output of a product but rather by innovation. What really matters would be “competition from the new commodity, the new technology, the new source of supply, the new type of organisation – competition […] which strikes not at the margins of the profits and the output of the existing firms but at their foundations and their very lives.” He argued that such innovation – perennial gales of “creative destruction” – is the centre of economic change, driving both economic development and business cycles. A company, even a successful one, would consistently feel such threat. This in turn could discipline its behaviour in a similar manner as a perfectly competitive market would.[2]

B. Role of innovation for platform competition

Schumpeter’s rationale worked well for digital (platform) markets because it coincides with some economic particularities of such markets.

The economics of multi-sided markets has taught us that when a platform has accumulated many users and therefore generates such strong positive network effects that, from a user’s perspective, any other platform appears inferior, a market may “tip” towards the platform. Once a market has tipped, newcomers offering the identical product are unlikely to attract a critical mass of users to generate sufficient network effects to trigger any growth. Therefore, where products and services are not interoperable (thereby “sharing” and neutralising the network effects created by the joint user bases), such tipping of a market may be irreversible. The competition “for the market” has been decided.[3] It may only be challenged by a new product with a significantly different quality or product design. Thus, for incumbent platforms, the biggest threat will come from drastic product differentiations, that is from a solution that – while offering (at least the low-end) functionalities of the incumbent – gains separate attention through different and new features that the incumbent does not provide and which add value. If such significant product differentiation originates from a neighbouring market rather than from within the incumbent’s market, and therefore catches the latter off-guard, economists refer to them as “disruptive innovations.”[4] This article will refer to both (i) drastic product differentiations from within a market, and (ii) disruptive innovations from outside as “significant innovations.”

The conceptual overlap between Schumpeter’s innovation theory and the functioning of multi-sided digital markets is striking. Digital markets that have tipped may only be contested by means of significant innovations that overcome the incumbent’s network effects. Hence, such innovations are tantamount to Schumpeter’s “creative destructions” that even dominant firms need to fear, and which may discipline them most.

C. Laissez-faire approach

Courts and economists drew different conclusions for digital markets from Schumpeter’s theories. Often, the dynamic arguments were used to dismiss antitrust interventions.[5] Because innovation-driven digital markets are highly dynamic, it was argued that even high market shares did not imply dominance[6] and that, in any case, due to short innovation cycles any dominance was either ephemeral or permanently subject to threats by disruptive forces. Since this would sufficiently discipline the incumbent firm, antitrust intervention was deemed unnecessary.[7] Accordingly, for two decades, authorities argued against interventions in digital markets on the grounds that “Type 1 errors” (the risk of any over-enforcement) could reduce the crucial incentive to innovate and invest in the dynamic digital markets, while “Type 2 errors” (the risk of under-enforcement) would matter far less as such errors would quickly be corrected by new market entries.[8]

2. Destruction of the creatives in digital ecosystems

Twenty years ago, such reasoning bore some empirical backing. In the early days of the internet, we did indeed witness an open and dynamic economy, in which new firms sprung up frequently, no internet company appeared to enjoy dominance for long, and fluctuation among them was high. Amongst others, we witnessed AltaVista/Yahoo! being pushed aside by Google as the dominant general search engine, and Meta’s Facebook toppling over the once strong social media platforms Friendster, Orkut and MySpace.

However, today, the digital economy looks different.

There are still vast areas witnessing a high level of innovation, such as in the field of artificial intelligence (AI). Yet, at least in the vicinity of some of the commercially most relevant online platform services, vigorous competition appears to have significantly cooled down, market concentration has risen to new highs and significant innovation (such as AI) is tied to very few. Today, some incumbent firms appear to focus more on innovative means to reduce competition and suppress innovation rather than on improving their own services. So where did reliance on Schumpeter’s innovation theory run short?

The main reason is that, contrary to the assumptions of some laissez-faire proponents, significant innovations do not just emerge out of nowhere and then magically topple over any incumbent firm if they are just good enough. For significant innovations, two requirements must be fulfilled. As next discussed, companies first need to have an incentive to innovate and, second, they must have the ability to effectively develop and, more importantly, bring innovation to the market. Both requirements are closely interlinked. Where it is unlikely that an undertaking would be able to monetise an innovation, there is no incentive to innovate. And where there is no incentive, even the best abilities do not suffice.

This is where Big Tech comes into play. As has been mentioned in the European Digital Markets Act, “[s]ome of those gatekeepers exercise control over whole platform ecosystems in the digital economy and are structurally extremely difficult to challenge or contest by existing or new market operators, irrespective of how innovative and efficient those market operators may be.”[9] Gatekeepers that control the operation of entire ecosystems around their core “cash cow” platform services lack the incentive to innovate against their own technology. And they have an incentive and the ability to suppress any significant innovation from third parties within their ecosystems which could, potentially, weaken their core platform business.

A. Incentives to innovate

Larger firms may have more resources to finance major research and development initiatives. They may also be better positioned subsequently to diffuse any resulting innovations and appropriate its value. “But monopolies, if protected from competition, are unlikely to be vigorous innovators.”[10] To be sure, digital incumbents are aware of the constant threat from significant innovations disrupting their business. However, a dominant firm with sunk investments in its technology has no incentive to invest in innovation that could not gain any additional business but may only cannibalise its existing business by replacing revenues it already secured, or even opening the door for third parties to displace such business in the long run.

Most recently, this could be observed at the example of the use of AI for search engines. As early as 2013, Google’s DeepMind had “amassed one of the biggest concentrations of deep learning experts in the world.”[11] It was Google engineers who in 2017 published a seminal paper on the “transformer-based” new deep-learning model[12] that is now used by Microsoft-funded OpenAI’s foundation model GPT-3 to power its famous ChatGPT, the fastest growing consumer application to date (with 100 million users in 2 months). With Large Language Models (LLMs) such as FLAN, LaMDA, BERT, PaLM and MuM, Google had been years ahead of OpenAI. Nevertheless, Google only decided to present its own AI-chatbot for its search engine, called Bard, in February 2023,[13] nearly the same day Microsoft had announced to embed ChatGPT into its search engine Bing and its browser Edge.[14] The fact that Google was able to react to ChatGPT in such a short period of time suggests that it had previously been holding this innovation back in fear that it could disrupt some of its core search-advertising business (as chatbot answers render sponsored links to advertisers’ websites redundant). Thus, far from sleeping in the AI race, Google waited until it was really necessary for it to bring innovation to market.

When the incumbent cannot be expected to drive innovation in relation to its business model, any dynamic competition depends on the incentives of third parties to invest in innovations that could overcome the incumbent’s existing technology. The most likely candidates would be remaining rivals within the relevant market (as in the case of Microsoft’s Bing integrating ChatGPT) or firms operating on neighbouring but related markets from which an attack could be launched (as in the case of Microsoft’s browser Edge integrating a chatbot). The prospect of getting a share of the incumbent’s monopoly profits will likely create a sufficient financial incentive for such an attack. However, any third party’s innovation incentive will also depend on the party’s likely ability to appropriate any value from a significant innovation, in particular by opening the door to the incumbent’s revenues. And here again, Big Tech has something to say.

B. Abilities to innovate

To appropriate value from any innovation, entrepreneurs need to master several phases. First, they need to actually invent something new. However, Schumpeter already had highlighted that for an innovation to lead to a “creative destruction” the initial phase of the invention is far less relevant than the subsequent diffusion. This is the period when the profitable potential of a new product or service is realised and it is widely rolled out.[15] This also applies to digital services.

When strong network effects are at play, for any significant innovation to displace an incumbent, it first needs to gain a critical mass of users to generate any positive network effects. The common strategy to gain such a foothold on the market is to aim for customers at the “low-end” of the market. Incumbent firms tend to focus on constantly improving their products to pull the market to the (more lucrative) “high-end.” This creates opportunities for other firms to attract users through low-end products that meet the basic requirements of users (the “value network”) while offering added value through their respective innovation. If this initial phase of gaining a foothold is successful, the innovator can successively redefine the factors that matter to users (i.e., the value network) and progress to also cater to the interests of mainstream users of the incumbent, thereby ultimately replacing it.[16]

The problem we are facing today, however, is that with a view to effectively prevent innovation that could challenge their business, some incumbents can and actively do interfere in all stages of the innovation process within an ecosystem, from the invention to the diffusion.

3. Gatekeepers’ suppression of innovation in closed ecosystems

The more entrenched an incumbent’s market position for its core “cash cow” service, and the higher the threat that an innovation could displace such service, the lower the incumbent’s incentive to invest in such innovation, but the stronger its incentive and ability to prevent any third party from doing so.[17]

Applying this to digital markets, twenty years of antitrust under-enforcement left us a very dim prospect for innovation. Over these years, a small group of companies has built up and connected a web of products and services around their core “cash cow” businesses that effectively shields such business from disruptive innovation or disintermediation.[18] At a global level, this is most striking as to Alphabet (Google) and Apple through their respective control of the iOS and Android-based mobile ecosystems.[19] However, to a lesser extent Amazon (for its marketplace and its cloud service AWS), Microsoft (for Office, Azure and Xbox) and Meta (for Facebook/Instagram/WhatsApp) also have built ecosystems around their core services. By establishing such ecosystems, those companies aim at being included in all the most profitable value chains built on or around their platforms (advertising, subscriptions/payments, data). They also gain influence on the innovation process to protect their core platform monopoly from entry and disruption. The ancillary services act as a protective “moat” or ”wall” around the incumbent’s “castle,” i.e., its core revenue-generating services, to identify any potential disruptive innovation and prevent that it may gain a foothold.

The control over such “walled-off” or “closed” ecosystems allows digital gatekeepers to suppress innovation. In particular, it enables them to effectively do all of the following:

  • Deprive third parties of the assets required to innovate: Digital gatekeepers typically control several inputs (data, IP & privacy rights, computing infrastructure, and funding) that may be crucial for participants in their ecosystem to innovate. OpenAI is a good example. They started with the ambition to remain an independent open-source AI developer and were heavily funded, including by Tesla founder Elon Musk. Despite of that, in 2019 OpenAI had to turn to Microsoft to gain access to its compute power to train the GPT-3 foundation model. Owning Azure, Microsoft is the second largest cloud computing provider in the world (after Amazon and before Google). But even Microsoft struggled with OpenAI’s task. Reportedly, “[t]raining the models required far more computing power than Microsoft’s systems had ever handled before.”[20] Microsoft had to team up with Nvidia, the leading “supercomputer” chipmaker, to build the fifth largest supercomputer on earth for OpenAI.[21] Thus, while OpenAI (and firms deploying its AI models) were marketed in public as innovative “start-ups”, it actually took all efforts of one of the largest tech firms and the largest supercomputers in the world to bring OpenAI’s groundbreaking GPT to life. Similarly, when in reaction of ChatGPT, Google invested $ 300 million in AI company Anthropic, it came in the form of a discount for using Google’s cloud computing services.[22] Without access to such infrastructure, the AI innovations might never have seen daylight. Yet, currently compute infrastructure in highly concentrated in the hands of Amazon, Microsoft and Google, whose combined market share in cloud computing is between 75% and 90%, with very high barriers to entry.[23] Hence, even the most innovative entrepreneur would have few alternatives to cooperating with the digital gatekeepers in order to train and run AI models at scale. The same is true when it comes to access to data or functionalities of internet-connected devices (controlled by incumbents) that is required to provide innovative products and services.
  • Set the conditions for any innovation to reach end users: Digital services need to be present on the main platforms such as search engines or app stores to reach end users. These platform services unilaterally set the rules and commercial conditions by which businesses in downstream (intermediated) markets may access end users through the respective (upstream) platform.[24] This allows an upstream intermediary to set conditions that disfavour any downstream innovation that could threaten its upstream or any of its other core businesses.
  • Monitor any innovation on related markets for pre-emptive defensive measures: Gatekeepers have been very successful in spreading and deploying sophisticated surveillance tools to constantly monitor the development and performance of businesses within and even outside of their digital ecosystems. Gatekeepers are therefore amongst the first to spot which new products or services are getting traction and pose a competitive threat. This enables them to defend their dominance in a highly targeted manner, either by acquiring any promising innovation or by anti-competitively preventing its success.[25]
  • Prevent disruptive market entry by acquiring and “killing” innovations: Google, Amazon, Apple, Meta and Microsoft acquired more than 400 companies from 2009 to 2020. [26] Many of those were “killer acquisitions,”i.e., acquisitions with the sole purpose of discontinuing the target’s innovation projects because they potentially could disrupt the acquirer’s technology. In the field of AI, Apple alone bought 60 companies between 2016 and 2020.[27]
  • Hamper the diffusion of any innovation within their ecosystem: By controlling the digital infrastructure within walled-off ecosystems, gatekeepers can largely influence which products or services end users detect, see, engage with, and ultimately use – and which do not. By hiding new products or services on core platforms (e.g., devices, operating systems, app stores, results pages), gatekeepers can directly influence user behaviour and thereby determine the diffusion of innovations. Moreover, where a gatekeeper also controls the intermediation of advertising and/or the systems for the fulfilment of payments (subscriptions) within its ecosystem, as in the case of Google, Meta and Apple, it may also hamper any marketing and subscription activities of a newcomer.
  • Quickly imitate, integrate and thereby outcompete any innovation within the ecosystem: Gatekeepers that control a digital ecosystem are likely to find it much easier than others to identify significant third-party innovators at an early stage, imitate them effectively and, most importantly, swiftly present copy-cats in a prominent manner throughout their entire ecosystem as fulfilling at least the same functionalities as the original innovation. Thereby, a gatekeeper may deprive its mainstream users’ incentive to switch. To speed up the process it may also acquire and integrate a direct competitor with an emerging innovation in a still nascent market.[28] Again, Google’s quick announcement to embed a chatbot (Bard) in its search engine akin to Bing’s integration of ChatGPT serves as prime example.

4. Creative destructions across digital ecosystems—the lucrative ecosystem oligopoly

The main, if not the only, ”blind spot” that a gatekeeper may have are disruptions originating from outside of the digital ecosystem it controls. This can be referred to as an inter-rather than intra-ecosystem disruption. Microsoft’s recent head-turning integration of ChatGPT into its search engine Bing and its browser Edge could be seen as such a head-on attack. So, are Google, Amazon, Apple, Meta and Microsoft sufficiently disciplined and incentivized to innovate because their ecosystems compete on a meta level and there is a constant threat that they disrupt each other’s core platform services, as some have argued? Most likely not. In any event, we cannot rely on innovation by just a handful global firms operating ecosystems.

Microsoft Bing’s recent move against Google Search was the first genuine attack on a core platform service of another ecosystem since Google’s famously failed attempt to attack Facebook with the launch of the social network “Google Plus” in 2011. As outlined, it took billions of dollars in Google’s own AI research, and a close cooperation between OpenAI, Microsoft and Nvidia to train GPT. Considering the joint effort, ChatGPT hardly serves as an argument that in digital markets competition was just “one prompt away”.[29] In fact, looked at closely, Microsoft’s cooperation with OpenAI appears to be much more about boosting Azur, Microsoft’s largest growing “cash cow” cloud compute service (which made up 34% of its total revenue in 2022), than about search advertising (which stood for 6%). Microsoft’s calculation appears to be that the open battle with Google over AI is perfect marketing to push more companies and investors into the AI sector. Since AI requires large computing power and Microsoft is the second largest cloud provider in the world, any such shift should ultimately benefit Microsoft. Operating the third largest cloud service after Amazon and Microsoft, Google may not even be too annoyed about this development. Google may well offset losses in search advertising caused by AI with gains in cloud services thanks to AI. When it comes to pushing cloud-based AI services into every business, Microsoft’s and Google’s interests are actually fully aligned.

Other than Bing’s attack on Google Search, over the last few years there have been some measures to limit the scope of another ecosystem or its underlying business model, but there appear to have been no attempts to fully disrupt another Big Tech’s ”cash cow” service. Listed below are several possible reasons.

  • Gatekeepers are well aware of the power of the protective moats around the core platform services in other ecosystems and the resulting means to prevent a disruptive market entry. They know that in order to neutralise all incumbent advantages they would often have to take on the other entire ecosystem, not just a particular service. The barriers for such attack are very high.
  • Any gatekeeper starting to invest in innovations to disrupt another gatekeeper’s ecosystem would have to expect the target to launch a counter-attack. Given the comparable resources and technical means of the respective opponents, the likely financial damage suffered is likely to be significant. This likely ‘lose-lose’ “retaliation effect” serves as a strong deterrent against such attacks.
  • With the notable exception of cloud compute services, the core business models of Google (search advertising), Amazon (online marketplace), Apple (sale of mobile devices), Meta (display advertising), and Microsoft (operating systems) are largely complementary. At least over the last decade, these companies have grown neatly side-by-side in terms of revenues and market capitalization.
  • At a macro-level, the companies contributed to their joint overall growth, thereby helping each other. That is because the largest growth potential still results from the overall increase in global internet consumption. Since 2015, nearly 3 billion people worldwide came online for the very first time.[30] According to Google: “In the next four years, we expect another 1.2 billion new internet users.”[31] The user experience typically starts with a mobile phone.[32] Big Tech’s joint goal is to increase such online usage. The same now applies to the use of AI by any business and the cloud services powering them.
  • Their joint enemy is offline and non-consumption. Occupying the first touchpoints that shape the user experience and providing combined added value, together they all increased digitalisation and shifted consumer and business attention away from offline media and commerce to internet consumption – where these gatekeepers are unavoidable trading partners. “Anything that increases Internet use ultimately enriches Google”, Google’s chief economist Hal Varian once said.[33] Such common goals in enhancing overall consumption and digitalization of any goods available unites more than it divides.
  • It makes more economic sense for the largest tech firms to focus their (static) innovation efforts on their own products, and to co-operate with the other gatekeepers to enhance their mutual total revenues rather than to disrupt each other’s services. Over the last few years, more and more such co-operation has emerged. Considering the “Revenue Sharing Agreements” (RSA) between Google and Apple, for instance, a Google manager described the relationship as follows: “Our vision is that we work as if we are one company”.[34]

5. Schumpeter revisited in walled-off ecosystems

In order to prevent competition and disruption, a gatekeeper does not have to own all competitive resources itself. It is sufficient to technically or commercially control them. The gatekeeper’s operation of largely walled-off ecosystems with technically integrated business users allows it to exert such effective control over the use of innovation inputs. By setting the rules of the game and the technical parameters to play for every participant in a digital ecosystem, gatekeepers have many means to suppress even radical innovation.[35]

Schumpeter would turn in his grave if he knew how some of his arguments on “creative destruction” are being used today to justify anti-competitive measures that destroy the Creatives. Schumpeter was not against competition intervention. In fact, for him the main criterion for whether or not a market was competitive concerned its contestability.[36] The relevant question was not “how many firms are in this industry” but “what are the barriers to entry that are preventing firms from coming up with substitutes?”[37] As outlined above, the control of an ecosystem can create insurmountable barriers to enter markets and to grow, even for the most innovative firms. Gatekeepers can use such control to spot and hinder even disruptive innovation, by depriving rivals of crucial resources or the means to diffuse any innovation within the closed ecosystem. Such “barriers to innovation” can insulate companies in today’s high-tech economy for competition for decades.[38]

6. Consequences for competition policy

Competition policy may no longer assume that dynamic competition sufficiently disciplines even dominant companies, and that there is a higher risk from over-enforcement than from under-enforcement.

Where walled-off ecosystems are suppressing dynamic competition for core platform services, competition authorities need to pro-actively intervene to make such markets contestable again. The protective walls need to be broken down so that superior innovation can freely develop and be rapidly adopted and diffused within any ecosystem. Products and services need to find their way to end users on the basis of their quality, not on the basis of the incumbent’s goodwill. Specific obligations to keep intermediation fair and markets contestable (for innovation) such as in the European Digital Markets Act is the right approach.

*Prof. Dr. Thomas Hoppner is a Partner in the Berlin office. The article is a condensed and updated version of the peer-reviewed article „From Creative Destruction to Destruction of the Creatives: Innovation in Walled-Off Ecosystems” Journal of Law, Market & Innovation (JLMI) 2/2022, p. 10-38.


[1] Joseph Schumpeter, Capitalism Socialism & Democracy (first published 1943, Taylor & Francis e-Library 2003) 84, 87.
[2] Schumpeter (n. 1) 84, 85.
[3] See generally Crémer/Montjoye/ Schweitzer, ‘Competition policy for the digital era, Final report’  <>
[4] Alexandre de Streel and Pierre Larouche, Disruptive Innovation and Competition Policy Enforcement (OECD Background Note, DAF/COMP/GF(2015)7, 2015) para. 4. <>
[5] Michael L. Katz and Howard A. Shelanski, ”Schumpeterian” Competition and Antitrust Policy in High-Tech Markets’ (2005) 14 Competition 47. 
[6] United States v. Microsoft Corp., 253 F.3d 34, 49 (D.C. Cir. 2001).
[7] See Case T-79/12 Cisco Systems and Messagenet v Commission [2013] para. 69.
[8] See Cristina Caffarra, Gregory Crawford and Tommaso Valletti, ‘‘How tech rolls’: Potential competition and ‘reverse’ killer acquisitions’ (, 11 May 2020) <>.
[9] Recital (3) of Regulation (EU) 2022 of the European Parliament and of the Council on contestable and fair markets in the digital sector (Digital Markets Act).
[10] Gilbert and Melamed, ‘Innovation Under Section 2 of the Shearman Act’ Antitrust Law Journal, 2021 <>
[11] <>
[12] Vasmani et all., Attention is All You Need, 31st Conference on Neural Information Processing Systems (NIPS 2017), <>
[13] <>
[14]   <>
[15] Joseph Schumpeter, Konjunkturzyklen. Eine theoretische, historische und statistische Analyse des kapitalistischen Prozesses (first published 1939, Vandenhoeck 1961).
[16] de Streel/Larouche(n. 4) para. 4.
[17] Compare Gilbert and Melamed (n. 10) 620.
[18] Disintermediation means the process of reducing the use of intermediaries between producers and consumers, in particular by cutting out one or more middlemen from a transaction. See <>
[19]  <>
[20]  <>
[21] <>
[22] <>
[23] <>
[24] See Recital (13) Digital Markets Act (n. 9).
[25] Lina M. Khan, ‘Remarks of Chair Line M. Khan’ (Speech at CRA Conference, Brussels, 31 March 2022) <>
[26] Oliver Latham, Isabel Tecu and Nitika Bagaria, ‘Beyond Killer Acquisitions: Are there more common potential competition issues in tech deals and how can these be assessed?’ (2020) CPI Antitrust Chronicle May 2020 <>
[27] <>
[28] Khan (n. 25) 3.
[29] See Schrepel, Competition is One Prompt Away, Network Law Review, 17 February 2023, <>
[30] Digital 2022 Global Overview Report <>
[31] Google Developers, ‘Building better products for new internet users’ (Google Developers Blog, 5 May 2022) <>
[32] Digital 2022 Global Overview Report (n. 29).
[33] Google chief economist Val Harian in interview with Steven Levy, ‘Secret of Googlenomics: Data-Fueled Recipe Brews Profitability’ (Wired, 22 April 2009) <>
[34] United States v Google (Case No. 1:20-cv-03010 (District Court, District of Columbia). DOJ amended complaint (15 January 2021), para. 120.
[35] See Robert Andrews, ‘Google Won't Buy Ailing Newspapers, Could 'Merge Without Merging'’ (CBS News, 8 February 2010) <>.
[36] Art Carden, ‘The Essential Joseph Schumpeter: An Easy and Accessible Introduction to an Important and Complex Thinker’ (AIER, 14 July 2020) <>
[37] Art Carden (n. 35).
[38] Gilbert and Melamed (n. 10) para. 679.

Other Publications