Warehouse-Scale Buildings, Humming

Looking back from the confrontation brewing in late 2018, the planet’s digital nervous system had crystallized into a distinct architectural pattern by 2017. Vast regions—Northern Virginia, Dublin, Osaka, São Paulo—were dotted with concentrations of warehouse-scale buildings, humming with activity and linked by undersea cables and fiber arteries. These were not isolated data centers but integrated nodes within a few monolithic grids: Amazon Web Services, Google Cloud Platform, Microsoft Azure. Their service catalogs, accessible through a console or an API call, listed hundreds of offerings—virtual machines, object storage, machine learning models, message queues, and databases—not as discrete products but as facets of a unified utility.

Software development for enterprises had largely shifted from building and maintaining infrastructure to composing services rented from these catalogs. The industrial vision of computing as a ubiquitous, on-tap utility, long forecast, was now the operational reality.

The generative core of this utility, however, its foundational software, was not minted by the utility owners. It was the accumulated output of the global open-source commons, poured into the cloud’s molds.

This was the embrace: total, efficient, and increasingly suffocating for those who had created the raw material.

Descending from this aerial view to a specific conference room in late 2017 reveals the human and corporate tension within this system. On a quarterly earnings call, executives of a publicly traded open-source company are addressing financial analysts. The numbers show growth, but the questions circling are not about product innovation; they are about competition from a cloud provider one hundred times their market capitalization. The cloud provider offers a managed service based on the very open-source project this company created and maintains. The executives articulate a strategy of differentiation—superior features, deeper expertise, closer alignment with the upstream community. Yet the subtext of the exchange is a stark economic asymmetry. The cloud provider views this service as a feature of its platform, a low-margin utility to drive consumption of core compute and storage. For the open-source company, it is the product, the revenue lifeline.

This scene, repeated with variations for companies like Elastic, MongoDB, and Confluent, crystallized the pivotal and paradoxical moment of 2015 to 2018. The hyperscale cloud providers completed their assimilation of open-source software as a core utility, triggering a fundamental crisis for the commercial open-source business model. By offering proprietary, fully-managed versions of popular open-source projects, they severed the traditional link between widespread adoption and sustainable funding for the original stewards. This chapter details that crisis—the culmination of an industrial reorganization where open source’s success in creating universal infrastructure led to its capture by the industry’s most concentrated capital, reshaping the very meaning of “open” in the process.

To grasp how profound this departure was, recall the lineage it disrupted. In the early days of computing, particularly during the 1950s and 1960s, programmers and developers commonly shared software to learn from one another and advance the field. Early systems such as Unix even provided users with access to their source code, allowing collaboration and modification. This was a pragmatic and cultural norm in an era before software became a standalone commercial product.

The rise of proprietary software in the 1970s and 1980s erected walls around that commons. The open-source movement of the 1990s and 2000s revived and institutionalized the collaborative ethos, using legal licenses like the GPL and Apache License to create a protected, production-grade commons. By the 2010s, surveys indicated that more than 90 percent of companies used open-source software as a component of their proprietary software—a pragmatic decision driven by quality, cost, and agility. The commons had become the standard software stack.

The cloud providers then made a pivotal realization: this thriving, community-built commons functioned as a free, global, and extraordinarily effective research and development department. Their own business model was not to sell software licenses but to sell convenience, scale, and integration—compute, storage, and above all managed services. Their strategic masterstroke was to repackage the fruits of this communal labor as proprietary, fully-managed offerings within their walled gardens. Critics gave this practice pointed names: “strip-mining” or “open-source laundering.”

The metaphor of extraction was apt. Cloud giants could harvest value from the open-source commons, refine it into a proprietary service offering, and return little of the resulting revenue or code to the upstream community that sustained the project. From their perspective, the economic efficiency was devastating. Amazon Web Services executed this strategy most aggressively. In 2015 AWS launched Aurora—a cloud-native database service advertised as MySQL- and PostgreSQL-compatible—using open-source wire protocols while keeping Aurora itself proprietary to AWS infrastructure. For users it offered managed experience with promises of higher performance; for open-source database communities it diverted potential customers who might otherwise have paid for enterprise support or managed services from commercial sponsors.

Google Cloud Platform and Microsoft Azure rapidly adopted this playbook because they saw it as strategically necessary: Google launched Cloud Bigtable (heavily influenced by HBase) plus managed PostgreSQL/MySQL; Microsoft Azure leveraged its pivot toward Linux/open source plus its enterprise sales channels to build its own portfolio of managed database services.

Companies like Elastic (search and analytics), MongoDB (document database), and Confluent (data streaming, built on Apache Kafka) had pioneered business models predicated on monetizing massive adoption. The typical approach was “open core”: a robust, freely available open-source core product drove adoption, fostered community contribution, and established a de facto standard. The company would then monetize by selling proprietary enterprise extensions, premium support, and—increasingly critical—its own managed cloud service. Their sustainability depended on converting a small fraction of a vast, global user base into paying customers. This model thrived when the primary alternative was self-hosting the complex open-source software. Now, they faced competition from entities orders of magnitude larger, who offered a managed version of the very core open-source product they had created and sustained. The competition was asymmetrical in every dimension: capital reserves, marketing reach, sales forces, and physical infrastructure.

The cloud providers’ primary strategic goal was to increase lock-in and consumption of their core compute and storage resources; they could afford to treat these managed database services as loss leaders or low-margin utilities designed to keep workloads tethered to their platform. For the dedicated open-source companies, these services were often the central pillar of their revenue model. The clouds could undercut on price, bundle the service seamlessly with other offerings, and leverage pre-existing relationships with virtually every large enterprise.

The concrete financial pressure manifested relentlessly in the quarterly rituals of public markets. On earnings calls, executives from open-source companies faced analysts’ probing questions about competitive displacement by cloud providers. Their responses consistently framed a battle for differentiation: the claimed superiority of their own managed services (MongoDB Atlas, Elastic Cloud, Confluent Cloud) in terms of advanced features, faster access to the latest upstream innovations, and direct support from the original engineering teams. They spoke of “true elasticity,” “native integration,” and “developer experience.” Yet the economic asymmetry was palpable and often reflected in valuation multiples.

The cloud providers had successfully turned the open-source companies’ greatest strategic asset—a widely adopted, standard-setting project—into a commoditized component they could resell, often at a lower upfront price point due to their scale advantages.

This dynamic provided a concrete, real-world illustration of the Openness Stack, the layered model where each level of software exhibits a strategically chosen degree of legal and technical openness. In the cloud era, the stack was being re-architected by the dominant platform players. The foundational layers—physical infrastructure, virtualization, and core networking—were proprietary cloud utilities, entirely closed. The middle layers—runtimes, data platforms, and orchestration—were increasingly defined by open-source cores (like Kubernetes, MySQL, Elasticsearch). However, the most lucrative and user-facing manifestation of these layers was often a proprietary, managed service wrapper offered by the cloud provider. The application layer at the top remained where independent developers and software companies could still capture unique value, but even that layer was increasingly reliant on and often integrated with the cloud providers’ proprietary middle-layer services.

The clouds aimed to own and control the profitable managed service layer, using the open-source core beneath it as a de facto standard that ensured compatibility and drew users in, but without assuming the obligation to be the primary sustainer of its development.

A deterministic explanation for this outcome exists and must be addressed. From this perspective, open source’s ascendance was always leading toward modular, commoditized components; its superior networked engineering efficiency made this inevitable. The cloud providers, according to this view, were simply the natural, efficient consolidators of this modularity, the logical endpoint of a trend toward utility computing. The struggles of open-source companies were therefore superficial epiphenomena, the fading protests of an intermediary layer being disintermediated by a more efficient economic model.

While the efficiency gains delivered by cloud services are undeniable and drove their adoption, this deterministic view overlooks the deeply contested institutional and legal renegotiation at play. The cloud providers did not invent a new software stack from whole cloth; they captured and enclosed an existing, thriving commons built under a specific set of legal licenses (designed for software distribution) and social norms (community-driven collaboration).

The cloud providers did not invent a new software stack from whole cloth; they captured and enclosed an existing, thriving commons built under a specific set of legal licenses (designed for software distribution) and social norms (community-driven collaboration).

Their actions forced a harsh economic reckoning and a consequential redefinition of what “open” meant when the dominant consumption model shifted from distributing code to providing services. This was not an inevitable, frictionless endpoint but a specific power dynamic—the concentration of infrastructural control meeting the diffuse production of software—that triggered a crisis severe enough to demand a radical response.

The conflict escalated sharply as the cloud providers expanded and deepened their offerings, moving beyond basic managed hosting to full API-compatible reimplementations. A stark turning point arrived in late 2017 when AWS previewed Amazon DocumentDB (with MongoDB compatibility). Announced in early 2018, the service was described as a “fast, scalable, highly available MongoDB-compatible database service.” It was crucial to note what it was not: it was not the MongoDB open-source software. It was a proprietary AWS re-implementation of the MongoDB API.

For MongoDB Inc., this was the crystallization of an existential threat. AWS was leveraging the entire ecosystem, developer mindshare, and toolchain that MongoDB had spent a decade building—the very value created by its open-source project—while contributing nothing back to its development and siphoning away potential customers. Similarly, AWS’s Elasticsearch Service continued its expansion, frequently marketed with an emphasis on its tighter integration with other AWS monitoring and security services than Elastic’s own official cloud offering could possibly achieve, leveraging the inherent advantage of controlling the entire platform.

The pressure thus transitioned from competitive to existential. The venture-backed open-source company model, which had fueled so much innovation in the previous decade, appeared to be hitting a structural ceiling. The clouds could move faster, scale cheaper, and sell more efficiently against the very creators of the software. These creators found themselves trapped in the paradoxical position where their greatest success—establishing a ubiquitous open standard—invited its own commoditization by larger players who viewed the underlying R&D as a free public good. The traditional open-source licenses provided no defense.

The GNU General Public License (GPL) and the permissive Apache License were designed for a world of software distribution. They guaranteed the freedom to use, modify, and redistribute the code, but they contained no provisions to prevent a cloud giant from using that code to run a highly profitable service without sharing any of the modifications or revenue generated by that service. The legal framework had not anticipated the service-based enclosure of the commons.

Financial markets internalized this risk, applying a discount to the valuations of pure-play open-source companies. The path to robust, defensible profitability seemed narrow when a dominant competitor with infinite resources could offer a “good enough” version of your core product as a cheap, integrated add-on to its broader suite.

The cloud’s embrace was warm and welcoming for end-users, who gained reliable, simplified services and often lower initial costs. For the original creators and stewards, however, that same embrace felt like a constricting squeeze, threatening to eviscerate the economic model that funded ongoing innovation. The period from 2015 to 2018 thus represents the decisive industrial reckoning.

The unresolved economic contradiction from the prior era of commoditization—where Docker had standardized the container and Kubernetes the orchestrator—found its logical conclusion in the cloud utility model. The efficiency of containers and standard APIs had lowered operational friction so completely that packaging open-source software into scalable, reliable services became an increasingly trivial engineering task for those who controlled the underlying platform. The promise of the earlier cloud era—that anyone could build a service on top of open source—was revealed to have a corollary: so could the infrastructure owners, with overwhelming advantages.

By late 2018, for the leadership of companies like MongoDB and Elastic, abstract crisis had hardened into concrete, inescapable reality. The collaborative ideals of the open-source ethos had given way to a stark confrontation over economic survival. The existing licenses were obsolete armor in this new war.

The only conceivable path forward appeared to be a radical re-tooling of the legal framework itself—to engineer new licenses that could specifically address the novel threat of cloud service provision without breaking the broader norms of open collaboration. This would not be a retreat from openness, its advocates would insist, but a defense of the sustainable openness that funded long-term innovation.

The existential threat had become so tangible that a legal counterattack evolved from a theoretical topic in community forums to a board-level necessity for corporate survival. In October 2018, MongoDB Inc. acted decisively, replacing its open-source license with the new Server Side Public License (SSPL). The license explicitly sought to require cloud providers who offered MongoDB as a service to open source the entirety of their service infrastructure—a “share-alike” proviso extended to the service layer. It was a direct, deliberate shot across the bow of the cloud giants. The gauntlet was thrown. The industrial squeeze had forced a new kind of war onto the legal battlefield.