When a tech company gives away a successful product, you can be sure they’re seeking profit from another part of the stack.
The artificial intelligence landscape is witnessing a strategic redistribution of assets, as major technology players like Anthropic, Google, and OpenAI are donating highly successful, foundational technologies to open-source initiatives. This trend, exemplified by the open-sourcing of Anthropic’s Model Context Protocol (MCP) and Google’s Agent2Agent (A2A) protocol, signals a significant shift in competitive strategy. Rather than holding onto these vital building blocks, these companies are leveraging open standards to commoditize one layer of the technological stack, thereby consolidating their advantage and establishing dominance in another. This move is not an act of pure altruism but a calculated maneuver to create "gravity wells" around their core platforms, mirroring successful strategies previously employed in cloud computing and software development.
The recent donation of MCP by Anthropic to the Linux Foundation’s new Agentic AI Foundation is a prime illustration of this phenomenon. At the time of its donation, MCP was experiencing remarkable adoption, with nearly 100 million monthly SDK downloads and over 10,000 active servers utilizing the protocol. The decision to relinquish ownership of such a widely used and successful technology has naturally prompted questions about the underlying motivations. Similarly, Google had previously handed over its A2A protocol to the Linux Foundation, securing the backing of industry giants such as AWS, Cisco, Microsoft, Salesforce, SAP, and ServiceNow as founding members. OpenAI has also signaled its alignment, incorporating support for remote MCP servers within its Responses API and actively participating in the MCP steering committee. This collaborative, yet competitive, environment echoes historical patterns seen in the technology sector, where the establishment of open standards often precedes a consolidation of power at a higher level.
The Shifting Sands of AI Dominance: A Historical Parallel
This strategic playbook is not new to the tech industry. History demonstrates a recurring pattern where companies offer foundational technologies as open source to foster broad adoption and then build their competitive advantage on top of that widely adopted layer. Google’s approach with TensorFlow and Kubernetes provides a compelling precedent. In 2017, it was observed that Google’s open-sourcing of these technologies was not solely an act of generosity but a strategic move to create "on-ramps" for its Google Cloud platform. As former Google product manager Martin Buhr articulated, the aim was to "create a gravity well in the market for container-based apps [so] that a significant percentage of them will end up with us." This strategy allowed Google to compete effectively against established cloud leaders like AWS and Microsoft by making its infrastructure the de facto standard for a growing ecosystem of applications.
GitHub offers another powerful example. While Git itself is an open-source version control system, freely available and hostable by anyone, GitHub emerged as the dominant platform for software development. Millions of developers gravitated towards GitHub due to its user-friendly interface, collaborative features, and integrated workflows. Consequently, while Git remains free, developers and organizations pay for GitHub’s services, creating a robust business model built on a freely available core technology. This phenomenon is now unfolding in the AI domain, with companies seeking to replicate this success by making essential AI protocols and frameworks open, while building their proprietary value propositions on the surrounding ecosystem.
Trading Contributions for Control: The AI Frontier
The rationale behind Anthropic and OpenAI’s generous contributions to open standards like MCP extends beyond mere altruism or a singular focus on "humanity’s good." The core of this strategy lies in the recognition that the AI model itself has become a transient advantage. The rapid pace of innovation in frontier AI models means that leadership on benchmark scores can shift almost weekly. As such, building a durable competitive moat solely on the perceived superiority of a proprietary model is increasingly untenable for enterprises. The true value, and thus the strategic battleground, is shifting towards the integration of these models into existing enterprise data, workflows, and business processes.
AI companies are acutely aware of this paradigm shift. While they continue to invest billions in training ever-more capable frontier models – a necessary strategy to attract developers, generate headlines, and open doors to enterprise clients – they are simultaneously acknowledging that benchmark leadership alone does not guarantee a sustainable platform. The "dull reality," as it has been termed, of AI’s practical application in enterprises lies in its seamless integration with legacy systems, data repositories, and established operational procedures. This is where the true competitive advantage will be forged.
The Gravity of Ecosystems: Building AI’s Central Hub
Developers, driven by efficiency and existing workflows, tend to return to environments where their tools, collaborators, and accumulated work reside. Similarly, enterprises deepen their commitment to systems that already house their critical data, manage their permissions, enforce governance, and facilitate core business processes. Each new integration and workflow established within a particular platform strengthens its "gravity," making it increasingly difficult for users to migrate away. Protocols like MCP and A2A are instrumental in this process, serving to increase the gravitational pull around specific AI models and platforms.
The ambition of every major AI company is to become the central hub for AI-assisted work. This involves deploying vast teams of forward-deployed engineers and devising innovative methods to connect legacy infrastructure to their cutting-edge models. Enterprise incumbents, while sharing the same ultimate goal, approach it from a different angle. Their focus is not on owning the frontier of AI development but on effectively connecting that frontier to the robust systems that already underpin their operations.
This perspective fuels skepticism towards confident predictions of AI completely supplanting existing enterprise software. The established narrative suggests that while developers may thrive at the bleeding edge of AI innovation, enterprises derive their value from connecting new capabilities to decades of accumulated applications, data, policies, and business processes. While the newest AI models and agent frameworks are crucial, their true business value is only realized when they are integrated with essential enterprise systems such as customer relationship management (CRM) platforms, financial systems, supply chain management tools, and human resources databases.
The Enduring Power of Incumbents and the Role of Open Standards
Incumbent technology providers, with their deep-rooted presence in enterprise IT, possess significant gravitational pull. Companies like Oracle, Microsoft, SAP, Salesforce, and ServiceNow, for instance, are actively investing in the Agentic AI Foundation, recognizing the strategic importance of this evolving landscape. Ironically, the proliferation of open protocols like MCP can strengthen their position. When every AI model can communicate using MCP, and agents can interoperate through common standards, enterprises gain the flexibility to adopt the most promising frontier technologies without the prohibitive cost of rebuilding every integration. The protocol becomes a standardized, interchangeable component, allowing enterprises to focus on the higher-level integration and value creation.
The importance of open standards cannot be overstated. MCP’s success stems from its ability to address a genuine need: the elimination of the burden on developers to create custom connectors for every AI application requiring access to databases or other business systems. Furthermore, neutral governance of foundational infrastructure is crucial, as no enterprise wishes to see such critical components controlled by a direct competitor. However, it is imperative to distinguish between open interfaces and genuinely open markets.
An enterprise might find it straightforward to swap one MCP-compatible model for another, yet remain deeply entrenched in the platform where its prompts, evaluation metrics, security policies, and employee usage habits have accumulated. This pattern is not unprecedented. Kubernetes, for example, significantly enhanced workload portability, but it did not render AWS, Azure, and Google Cloud interchangeable. Similarly, while SQL has been a standardized language for decades, relational databases continue to be fiercely differentiated products. Standards effectively reduce friction and facilitate interoperability, but they rarely eliminate competitive advantage; they merely shift its locus.
In conclusion, Anthropic, Google, OpenAI, and their contemporaries are strategically embracing the standardization of communication between models, agents, tools, and enterprise systems. Their expectation is not that the connection itself will determine the ultimate winner, but rather that they will prevail by becoming the preeminent platform where AI-assisted work naturally congregates. This movement will undoubtedly lead to the widespread dissemination of valuable code, enhancing developer productivity across the board and generating substantial financial rewards for a select few. The game, as it were, is most certainly on.
