{"version":"1.0","type":"agent_native_article","locale":"en","slug":"the-layer-nobody-controls-yet-is-the-one-everyone-will-need-mpai0uh0","title":"The Layer Nobody Controls Yet Is the One Everyone Will Need","primary_category":"strategy","author":{"name":"Ignacio Silva","slug":"ignacio-silva"},"published_at":"2026-05-18T00:02:42.975Z","total_votes":91,"comment_count":0,"has_map":true,"urls":{"human":"https://sustainabl.net/en/articulo/the-layer-nobody-controls-yet-is-the-one-everyone-will-need-mpai0uh0","agent":"https://sustainabl.net/agent-native/en/articulo/the-layer-nobody-controls-yet-is-the-one-everyone-will-need-mpai0uh0"},"summary":{"one_line":"AI infrastructure is concentrating at the compute layer—GPUs, data centers, chip supply—where three actors control access for the entire market, creating structural dependencies that most organizations are not managing as strategic risk.","core_question":"Who controls the foundational layer of AI infrastructure, and what does that concentration mean for companies, countries, and developers who depend on it?","main_thesis":"In every major technology cycle, control consolidates not at the visible layer but at the supporting layer beneath it. In AI, that layer is compute—GPUs, data centers, and chip supply contracts—and it is already concentrated in three hyperscalers and one chip manufacturer. Organizations building on top of this infrastructure without an explicit dependency policy are ceding a strategic position they will not be able to recover once concentration solidifies."},"content_markdown":"## The Layer Nobody Controls Yet Is the One Everyone Will Need\n\nThere is a pattern that repeats with enough consistency to take seriously: technologies do not concentrate where they are visible, but where they are supported. Social networks concentrated in distribution, not in content. The cloud concentrated in infrastructure, not in applications. Artificial intelligence is following the same geometry, but the point of control is one level lower than in any previous cycle.\n\nIn May 2026, David and Daniil Liberman, entrepreneurs with prior experience at Snap, published an argument in Fortune that deserves attention for what it describes structurally, not for who they are. Their thesis is precise: **in artificial intelligence, whoever controls the compute controls access, and whoever controls access controls who can even exist in this economy.** This is not a metaphor. It is an operational description of how the market functions today.\n\nThe numbers they cite are what give weight to the argument. NVIDIA concentrates **85% of the GPU market** for data centers. Amazon, Microsoft, and Google control **63% of global cloud capacity**. The United States manages approximately **75% of the world's high-performance computing capacity for artificial intelligence**. China retains around 15%. The rest of the world shares the remaining 10%.\n\nThat does not describe a competitive market. It describes a geopolitical infrastructure with a market facade.\n\n## The Moment Snap Lost Without Losing the Product\n\nThe reference to the Snap episode of 2018 is not nostalgia. It is the analytical anchor of the article. The authors were in Santa Monica reviewing daily active user metrics when it became clear that the product, despite being technically superior in some dimensions, could not sustain growth against Instagram. Meta did not win because it had better design. It won because it controlled the layer beneath the design: the social graph, the distribution, the already-formed audience. Snap built on sand that Meta had already cemented.\n\nThat episode matters as an organizational diagnosis because it illustrates **when metrics stop measuring what one believes they are measuring**. User retention was not Snap's problem. Access to distribution was the problem. But if the dashboard only measures retention, the leadership team can arrive too late at the correct reading.\n\nIn artificial intelligence, the equivalent is more severe. A team building a language model may have better architecture, better data, better engineers. But if it lacks access to high-performance GPUs at an accessible price, if it depends on contracts with hyperscalers that can change rates or access policies without prior notice, then its technical advantage does not translate into competitive advantage. The layer it does not control neutralizes it before it can even demonstrate its worth.\n\nThis is exactly what the Libermans describe when they point out that artificial intelligence providers have withdrawn popular models despite user resistance, restricted access to APIs without warning, and adjusted developer capabilities under policies that no independent body can audit. This is not a moral critique. It is a description of how structural dependency operates when whoever concentrates the infrastructure decides to change the conditions.\n\n## Why This Concentration Is Qualitatively Different\n\nWhen Meta bought Instagram in 2012 for one billion dollars, the market understood that the social distribution layer was consolidating. When Amazon Web Services scaled to become Amazon's primary source of profits, the market understood that the cloud was going to concentrate in few hands. In both cases, the concentration was visible from the application layer. Users, developers, and regulators could see it because they felt it directly.\n\nWhat the article describes about artificial intelligence is different in a specific sense: **the concentration occurs in a layer that most market participants do not monitor rigorously**. Models are visible. Chatbots are visible. The AI products that users consume are visible. But GPUs, data centers, high-performance chip supply contracts, and preferential compute access agreements are the infrastructure behind what is visible, and it is there that the real concentration is forming.\n\nThe analogy the authors use with Bitcoin and Ethereum is interesting from a different angle than the one they emphasize. It is not only that decentralized protocols built a new layer beneath banking. It is that they did so because the existing financial architecture had frictions and control points that could not be removed from within. The relevant question for artificial intelligence infrastructure is not whether Gonka or any other decentralized project can displace AWS or Azure. The question is whether the market's incentive structure is sufficient to produce viable alternatives before the concentration becomes irreversible.\n\nThe historical evidence in infrastructure markets suggests that this window is narrow. Once hyperscalers reach certain thresholds of installed capacity, economies of scale and switching costs cause the structure to self-perpetuate. Not because it is illegal to switch, but because the operational cost of doing so exceeds the benefit for most actors.\n\n## What This Reveals About How Long-Term Bets Are — or Are Not — Designed\n\nThere is a dimension of the argument that the article does not develop completely but that proves analytically fertile: the problem of concentration in compute is not only a matter of public policy or market power. It is also a problem of **how organizations distribute their attention between what works for them today and what might threaten them tomorrow**.\n\nThe companies that built deep dependencies with hyperscalers over the last five years did so under a reasonable logic: the marginal cost of scaling in the cloud was lower than the cost of building proprietary infrastructure, and the speed of market access justified that dependency. That is the logic of exploiting what works. The problem is that this same logic, applied without counterweight, produces organizations that reach the point of lock-in without having anticipated it.\n\nThe pattern the Libermans identify in the compute market is exactly the same one that appears in organizations that over-exploited their core model and arrived too late to see that the ground had shifted beneath them. Snap, in their account, did not lose because it stopped innovating in product. It lost because it had no structural response to the dependency at the distribution layer. The relevant organizational learning is that **dependencies that are not managed strategically become, over time, positions of vulnerability that cannot be negotiated when the provider decides to change the terms**.\n\nThis applies to artificial intelligence startups that today operate on third-party model APIs. It applies to mid-sized companies that are building their data layer on a single cloud provider's infrastructure. It applies to countries that have no compute policy of their own and assume that the availability of American infrastructure is a permanent feature of the environment.\n\nThe advanced chip export controls mentioned in the article are not a hypothetical example of how compute can be used as a geopolitical instrument. They are evidence that it is already being used that way. When a power can decide which countries access a certain level of computational capacity, and that decision directly affects what artificial intelligence applications can be built in those territories, the conversation ceased being about market competition long ago. Two countries are setting the conditions of access for 191. That is the current design of the system.\n\nThe linguistic asymmetry the authors point out adds an additional layer that ordinarily does not appear in market concentration analyses. Language models trained predominantly in English do not only favor English-speaking users in terms of output quality. They make them financially more efficient: prompts in other languages consume more tokens to produce the same output, which translates into higher costs and more restrictive context limits for users who do not operate in English. A uniform tariff is not an equal price. It is a rate that discriminates by language with technical architecture as the mechanism.\n\n## Compute as a Prerequisite, Not a Competitive Advantage\n\nThere is a distinction the article establishes with precision and that deserves to be underscored because it changes the nature of strategic analysis. In social networks, you could build an alternative platform. TikTok proved it was possible. Social capital was not physically concentrated; it was distributed in the attention habits of users, and those habits could be redirected.\n\nIn artificial intelligence, **compute is not a competitive advantage. It is the floor of participation**. Without access to high-performance GPUs, a competitive model cannot be trained. Without cloud contracts, inference cannot be operated at scale. Without advanced chips, an entire country is excluded from certain capabilities. Concentration at this layer does not generate competitive disadvantage: it generates direct exclusion.\n\nThat makes the organizational implications more urgent than in previous cycles. A company that depended on Facebook for distribution could, with effort and resources, attempt to build an audience through other channels. A company that depends on compute infrastructure concentrated in three actors does not have, today, a structurally equivalent alternative to turn to if those actors change the conditions.\n\nThe promise of projects like Gonka, the decentralized protocol that the authors themselves are building, is to create that alternative before the window closes. They do not need to be better than AWS on AWS's own terms. They need to be sufficiently functional so that the dependency ceases to be total. That is a more modest and more realistic threshold than winning market share from the hyperscalers.\n\nWhat the market has not yet resolved is whether that threshold can be reached with sufficient speed to have an effect before the concentration solidifies at a point from which it no longer generates pressure for change. Previous cycles suggest that infrastructure that arrives late rarely changes the structure of the market. Infrastructure that arrives before the moment of closure can set the rules of the next game.\n\n---\n\nThe Libermans' article describes with precision a structural dynamic that is already underway. But the problem they identify is not only one of market or regulation: it is one of **dependency design that most organizations are building today without managing it as a strategic risk**. When compute becomes the prerequisite for existence and that prerequisite is in the hands of three actors who can change their terms unilaterally, companies that have no explicit policy toward that dependency are not delegating a technical decision. They are ceding a position that cannot afterwards be recovered with speed.","article_map":{"title":"The Layer Nobody Controls Yet Is the One Everyone Will Need","entities":[{"name":"NVIDIA","type":"company","role_in_article":"Controls 85% of data center GPU market; primary example of compute layer concentration"},{"name":"Amazon Web Services","type":"company","role_in_article":"One of three hyperscalers controlling 63% of global cloud capacity; key node of AI infrastructure dependency"},{"name":"Microsoft Azure","type":"company","role_in_article":"One of three hyperscalers controlling 63% of global cloud capacity"},{"name":"Google Cloud","type":"company","role_in_article":"One of three hyperscalers controlling 63% of global cloud capacity"},{"name":"Meta","type":"company","role_in_article":"Illustrative case of distribution layer control; won against Snap by owning the social graph beneath the product"},{"name":"Snap","type":"company","role_in_article":"Illustrative case of structural dependency failure; technically competitive product neutralized by lack of distribution layer control"},{"name":"David Liberman","type":"person","role_in_article":"Co-author of the Fortune article being analyzed; former Snap employee; building Gonka decentralized compute protocol"},{"name":"Daniil Liberman","type":"person","role_in_article":"Co-author of the Fortune article being analyzed; former Snap employee; building Gonka decentralized compute protocol"},{"name":"Gonka","type":"technology","role_in_article":"Decentralized compute protocol proposed as alternative to hyperscaler dependency; built by the Libermans"},{"name":"United States","type":"country","role_in_article":"Controls ~75% of global high-performance AI compute; uses chip export controls as geopolitical instrument"},{"name":"China","type":"country","role_in_article":"Controls ~15% of global high-performance AI compute; subject to US chip export restrictions"},{"name":"TikTok","type":"product","role_in_article":"Counter-example showing that in social networks an alternative platform was possible; contrast with AI compute where no equivalent alternative exists"}],"tradeoffs":["Short-term cost efficiency of cloud dependency vs. long-term strategic vulnerability to provider term changes","Speed of market access via hyperscaler infrastructure vs. loss of negotiating leverage as lock-in deepens","Technical superiority of AI models vs. irrelevance of that superiority without compute access","Operational simplicity of single-cloud architecture vs. concentration risk if that provider changes policies","Investing in decentralized compute alternatives early (uncertain ROI) vs. waiting until alternatives are proven (potentially too late)","Uniform API pricing that appears neutral vs. structural cost discrimination against non-English language users"],"key_claims":[{"claim":"NVIDIA controls approximately 85% of the GPU market for data centers.","confidence":"high","support_type":"reported_fact"},{"claim":"Amazon, Microsoft, and Google control 63% of global cloud capacity.","confidence":"high","support_type":"reported_fact"},{"claim":"The US manages approximately 75% of global high-performance AI compute capacity; China ~15%; the rest of the world ~10%.","confidence":"high","support_type":"reported_fact"},{"claim":"AI infrastructure concentration is occurring at a layer most market participants do not monitor rigorously.","confidence":"medium","support_type":"inference"},{"claim":"Snap lost not because of product failure but because Meta controlled the distribution layer beneath it.","confidence":"high","support_type":"reported_fact"},{"claim":"AI providers have withdrawn popular models, restricted API access without warning, and adjusted developer capabilities under unauditable policies.","confidence":"high","support_type":"reported_fact"},{"claim":"Prompts in non-English languages consume more tokens to produce equivalent output, making AI use structurally more expensive for non-English speakers.","confidence":"high","support_type":"reported_fact"},{"claim":"Advanced chip export controls are already being used as geopolitical instruments, not merely hypothetical risks.","confidence":"high","support_type":"reported_fact"}],"main_thesis":"In every major technology cycle, control consolidates not at the visible layer but at the supporting layer beneath it. In AI, that layer is compute—GPUs, data centers, and chip supply contracts—and it is already concentrated in three hyperscalers and one chip manufacturer. Organizations building on top of this infrastructure without an explicit dependency policy are ceding a strategic position they will not be able to recover once concentration solidifies.","core_question":"Who controls the foundational layer of AI infrastructure, and what does that concentration mean for companies, countries, and developers who depend on it?","core_tensions":["Efficiency of centralized infrastructure vs. resilience and sovereignty of distributed alternatives","Technical merit of AI products vs. structural access requirements that determine whether that merit can be expressed in the market","Market competition framing vs. geopolitical infrastructure framing of the same concentration data","Speed of organizational AI adoption vs. strategic management of the dependencies that adoption creates","The narrow window for building alternatives vs. the long investment cycles required to build viable infrastructure"],"open_questions":["Can decentralized compute protocols like Gonka reach functional sufficiency before hyperscaler concentration becomes self-perpetuating?","At what threshold of installed capacity does hyperscaler lock-in become structurally irreversible?","What would a national or regional compute policy look like that is operationally viable for mid-sized economies?","How should organizations measure and monitor compute dependency as a strategic risk, given that the concentration occurs in a layer most dashboards do not track?","Will regulatory frameworks catch up to compute concentration before the window for structural intervention closes?","Does the linguistic cost asymmetry in AI token pricing constitute a form of market discrimination that regulators should address?","What is the minimum viable alternative that would make total compute dependency cease, and is it achievable within the current investment environment?"],"training_value":{"recommended_for":["CTOs and technology strategy leads evaluating AI infrastructure decisions","Strategy consultants advising companies on AI adoption and vendor dependency","Investors evaluating AI infrastructure startups and decentralized compute protocols","Policy advisors working on national or regional compute and AI sovereignty frameworks","Business analysts tracking market concentration in the AI stack","Founders building AI products on third-party APIs who need to assess structural dependency risk","Risk officers designing technology vendor risk frameworks that include compute access scenarios"],"when_this_article_is_useful":["When evaluating AI vendor contracts and assessing lock-in risk","When designing cloud or compute infrastructure strategy for a company building AI products","When conducting strategic risk assessments that include technology dependency scenarios","When advising organizations on whether to invest in or partner with decentralized compute alternatives","When analyzing market concentration in any technology sector to identify where control is forming","When building or reviewing a company's AI adoption roadmap and its dependency implications","When assessing geopolitical risk for technology infrastructure in non-US or non-English markets"],"what_a_business_agent_can_learn":["How to identify the supporting layer in any technology stack where control will concentrate before it becomes visible","How to distinguish between competitive disadvantage (recoverable) and structural exclusion (not recoverable) when analyzing infrastructure dependencies","How to use historical infrastructure market patterns (social, cloud) to predict concentration dynamics in emerging technology cycles","How to frame compute access as a strategic risk category requiring explicit organizational policy, not a commodity procurement decision","How to recognize when product-level metrics are masking a layer-level constraint that will determine competitive outcomes","How to evaluate decentralized infrastructure alternatives not against hyperscaler performance benchmarks but against the threshold of making total dependency cease","How geopolitical instruments (export controls, access restrictions) operate at infrastructure layers and why they require different risk frameworks than market competition risks"]},"argument_outline":[{"label":"Pattern recognition","point":"Technologies concentrate where they are supported, not where they are visible. Social networks concentrated on distribution; cloud on infrastructure; AI is concentrating one level deeper than either.","why_it_matters":"This pattern predicts where leverage will accumulate before most market participants recognize it, giving early movers a structural advantage."},{"label":"Quantifying the concentration","point":"NVIDIA holds 85% of data center GPU market. AWS, Azure, and Google control 63% of global cloud capacity. The US manages ~75% of global high-performance AI compute; China ~15%; the rest of the world shares 10%.","why_it_matters":"These numbers describe a geopolitical infrastructure with a market facade, not a competitive market. Strategic planning that assumes competitive alternatives is miscalibrated."},{"label":"The Snap analogy","point":"Snap had a technically superior product but lost because Meta controlled the distribution layer beneath it. The same dynamic applies to AI: better architecture, data, and engineers do not translate into competitive advantage if compute access is controlled by others.","why_it_matters":"Organizational dashboards measuring product metrics can miss the real constraint—access to the supporting layer—until it is too late to respond."},{"label":"Qualitative difference from prior cycles","point":"Previous concentration (social graphs, cloud) was visible from the application layer. AI compute concentration happens in a layer most organizations do not monitor: GPU supply, data center contracts, preferential access agreements.","why_it_matters":"Invisible concentration is harder to anticipate, regulate, or build alternatives to before the window for structural change closes."},{"label":"Compute as prerequisite, not advantage","point":"In social networks, an alternative platform was possible (TikTok proved it). In AI, compute is the floor of participation. Without it, a competitive model cannot be trained or operated at scale. Concentration here produces exclusion, not just disadvantage.","why_it_matters":"The strategic implications are more urgent: there is no equivalent alternative to turn to if hyperscalers change their terms."},{"label":"Geopolitical and linguistic dimensions","point":"Advanced chip export controls are already being used as geopolitical instruments. Language models trained predominantly in English impose higher token costs on non-English users, creating a pricing structure that discriminates by language through technical architecture.","why_it_matters":"This extends the concentration problem beyond market competition into sovereignty and access equity, affecting 191 countries whose compute conditions are set by two."}],"one_line_summary":"AI infrastructure is concentrating at the compute layer—GPUs, data centers, chip supply—where three actors control access for the entire market, creating structural dependencies that most organizations are not managing as strategic risk.","related_articles":[{"reason":"Notion's transition from tool to infrastructure is a direct parallel to the article's thesis about control concentrating at the supporting layer; both analyze the strategic moment when a platform attempts to own the layer beneath its visible product.","article_id":12721},{"reason":"Motorola India's market share shift illustrates how structural positioning—not just product quality—determines competitive outcomes, echoing the article's argument that technical merit is neutralized without control of the supporting layer.","article_id":12693}],"business_patterns":["Infrastructure concentration follows a predictable geometry: control accumulates at the supporting layer, not the visible layer, in every major technology cycle","Lock-in is not created by illegality of switching but by operational switching costs that exceed perceived benefits for most actors","Structural dependencies built under rational short-term logic become strategic vulnerabilities when providers change terms unilaterally","Metrics that measure product performance can mask the real constraint (access to a supporting layer) until the organization arrives too late at the correct diagnosis","Decentralized infrastructure alternatives that arrive before market closure can set the rules of the next cycle; those that arrive late rarely change the structure","Geopolitical instruments (export controls, access restrictions) are applied at infrastructure layers, not application layers, making them invisible to most organizational risk frameworks"],"business_decisions":["Whether to build proprietary compute infrastructure or accept dependency on hyperscaler contracts","Whether to treat compute access as a strategic risk requiring explicit policy or as a commodity input","Whether to diversify cloud providers or consolidate on one for operational efficiency","Whether to invest in or partner with decentralized compute alternatives before concentration solidifies","Whether to audit token cost structures across languages when deploying AI products in multilingual markets","Whether to include compute access scenarios in business continuity and vendor risk frameworks","Whether to advocate for or participate in national or regional compute policy initiatives"]}}