When a tech company gives away a successful product, you can be sure they’re seeking profit from another part of the stack.

The tech industry’s strategic landscape is undergoing a seismic shift, with major players like Anthropic, Google, and OpenAI increasingly donating highly successful products to open-source foundations. This seemingly altruistic move, exemplified by Anthropic’s donation of the Model Context Protocol (MCP) and Google’s contribution of its Agent2Agent (A2A) protocol to the Linux Foundation’s Agentic AI Foundation, is not born out of pure generosity. Instead, it signals a calculated maneuver to commoditize one layer of the technology stack, thereby shifting competitive advantage to other, more lucrative areas. This phenomenon mirrors historical patterns seen in cloud computing and software development, where establishing open standards serves as a prelude to building dominant platforms.

The recent spate of open-sourcing initiatives within the AI sector points to a broader strategic realignment. Anthropic, a prominent AI safety and research company, made waves last year by donating MCP, a protocol that had achieved remarkable adoption, to the newly established Agentic AI Foundation under the umbrella of the Linux Foundation. At the time of its donation, MCP was reportedly garnering close to 100 million monthly SDK downloads, with over 10,000 active servers utilizing the protocol. The question naturally arises: why would a company relinquish control over such a demonstrably successful and widely adopted technology?

This pattern was prefigured by Google’s earlier decision, several months prior, to hand over its Agent2Agent (A2A) protocol to the Linux Foundation. This initiative garnered significant support from industry heavyweights, including AWS, Cisco, Microsoft, Salesforce, SAP, and ServiceNow, who joined as founding members. OpenAI, not to be left out of this developing ecosystem, has also integrated support for remote MCP servers within its Responses API. Furthermore, OpenAI holds a seat on the MCP steering committee and has contributed AGENTS.md to the same foundation, aligning itself with the open-source efforts of its closest rivals. This coordinated effort suggests a shared strategic vision among leading AI companies, even amidst intense competition.

This strategic choreography can be likened to a high-stakes game of chess, where the leading AI powers are orchestrating "regime change" through what appear to be acts of beneficence rather than direct confrontation. For seasoned observers of the technology industry, this playbook is entirely predictable. It echoes the strategic evolution witnessed in the cloud computing era and the deployment of on-premises servers, where platform companies routinely open-source technologies that are no longer central to their unique competitive advantage. Instead, they donate technologies that have become commoditized, allowing them to focus their resources and innovation on emerging battlegrounds where they hold a stronger strategic position.

The Shifting Sands of AI Dominance

The strategic rationale behind these open-sourcing moves becomes clearer when viewed through the lens of historical technological adoption and platform building. Google, for instance, has long been a significant contributor to the open-source community. As far back as 2017, analysis suggested that Google’s open-sourcing of foundational technologies like TensorFlow and Kubernetes was not merely an act of goodwill. Rather, it was a strategic move to create "on-ramps" to its Google Cloud platform. At a time when Google was actively competing with AWS and Microsoft in the burgeoning cloud market, then-Google product manager Martin Buhr articulated the company’s ambition: 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 aimed to commoditize the underlying infrastructure, making it easier for developers to build and deploy applications, while simultaneously drawing them into Google’s cloud ecosystem.

This approach underscores a fundamental principle of platform companies: they often commoditize one layer of the technological stack to establish a competitive advantage in areas where they possess a more significant strategic leverage. By making certain technologies freely available and widely adopted, they reduce the barriers to entry for developers and businesses, fostering a vibrant ecosystem. This, in turn, creates network effects and lock-in, making it more attractive for users to remain within their broader commercial offerings.

GitHub serves as another compelling illustration of this strategy. Git, the underlying version control system, is open-source, allowing anyone to host their repositories. Historically, various companies offered Git hosting services. However, GitHub emerged as the de facto standard for millions of software developers. While Git itself remains free, the vast majority of developers and organizations pay for GitHub’s services, which include collaboration tools, project management features, and enhanced security. This demonstrates how a foundational open-source technology can drive the adoption and profitability of a proprietary platform built upon it. The AI industry is now witnessing a similar phenomenon unfold.

Trading Contributions for Control: The New AI Frontier

The motivations behind Anthropic’s and OpenAI’s seemingly benevolent actions extend beyond a simple desire to benefit humanity. While these companies often articulate their goals in terms of advancing AI for the greater good, the aggressive pace at which they are open-sourcing critical components like MCP suggests a deeper strategic imperative. The core AI models themselves, while groundbreaking, have proven to be a less durable source of competitive advantage than initially anticipated. The rapid evolution of AI research means that leadership in frontier model quality is a fleeting advantage, with leaderboards changing almost weekly. This volatility makes it imprudent for enterprises to build their AI strategies on the assumption that any single vendor will maintain a perpetual edge in model performance.

As recent analyses have highlighted, the true value proposition for enterprises lies not solely in the sophistication of the AI models themselves, but in their seamless integration with existing enterprise data, workflows, and business processes. The "sexy" aspect of AI—the cutting-edge models—is increasingly giving way to the "dull reality" of practical implementation and value realization. AI companies are acutely aware of this. While they continue to invest billions in training increasingly powerful models, recognizing their importance in attracting developers, generating headlines, and opening doors to enterprise clients, they are also tacitly acknowledging that benchmark leadership alone is insufficient to build a sustainable and defensible platform.

The underlying principle is that developers and enterprises tend to gravitate towards environments where their existing tools, workflows, collaborators, and accumulated work reside. For developers, this means platforms that offer integrated development environments, seamless collaboration, and efficient deployment pipelines. For enterprises, this translates to systems that securely connect to their vast repositories of data, manage access controls and governance policies, and integrate with established business processes. Each new integration and workflow established within a particular platform strengthens its "gravity," making it increasingly difficult for users to migrate elsewhere. This is precisely the strategic objective behind the widespread adoption of protocols like MCP and A2A: to increase the "gravity" around AI models and the platforms that facilitate their use.

Every major AI company aspires to become the central hub for AI-assisted work. To achieve this, they are deploying legions of "forward-deployed engineers" and exploring various avenues to integrate legacy infrastructure with their cutting-edge models. Enterprise incumbents, conversely, are pursuing a similar goal but from a different angle. They do not necessarily need to own the frontier of AI research; instead, their primary objective is to connect these advanced capabilities to the robust and secure systems that already underpin their operations.

This dynamic leads to a degree of skepticism regarding predictions of a complete AI-driven overhaul of enterprise software. The historical pattern suggests a more nuanced evolution. While developers may thrive on the cutting edge of AI development, enterprises prioritize the integration of new capabilities with decades of accumulated applications, data, policies, and business processes. The newest AI model or agent framework, while significant, delivers limited business value until it is effectively connected to critical enterprise systems such as customer relationship management (CRM) platforms, financial systems, supply chain management tools, and human resources databases.

It is in this realm of integration and existing infrastructure that established enterprise players continue to wield immense "gravitational pull." Companies like Oracle, Microsoft, SAP, Salesforce, and ServiceNow, all active participants in the Agentic AI Foundation, firmly believe in this principle. Their strategy revolves around leveraging their existing customer base and deep integrations to facilitate the adoption of AI technologies. Ironically, the promotion of open protocols, rather than undermining their position, can actually strengthen it. When all AI models can communicate using a standardized protocol like MCP, and agents can interoperate through common standards, enterprises gain the flexibility to adopt the best frontier technologies available without the prohibitive cost of re-engineering every existing integration. The protocol, in essence, becomes interchangeable.

Open Standards: A Catalyst, Not a Replacement for Gravity

The rise of open standards like MCP and A2A does not diminish their significance. These protocols are succeeding because they address genuine, persistent problems faced by developers and enterprises. The need for developers to construct custom connectors for every instance an AI application requires access to a database or another business system is a significant friction point that MCP aims to alleviate. Furthermore, neutral governance of foundational infrastructure is paramount, as no organization wishes to cede control of essential technologies to a direct competitor. However, it is crucial to distinguish between open interfaces and open markets.

An enterprise might find it relatively straightforward to substitute one MCP-compatible model for another. Yet, this ease of substitution does not negate a deep-seated dependence on the platform where its prompts, evaluations, security policies, and employee work habits have become embedded. This scenario echoes past technological shifts. For example, Kubernetes has dramatically enhanced the portability of workloads, but it has not rendered cloud providers like AWS, Azure, and Google Cloud interchangeable. Similarly, SQL has been a standardized database query language for decades, yet relational databases remain fiercely differentiated and competitive markets. Standards effectively reduce friction, but they rarely eliminate competitive advantage; they merely relocate it.

In this ongoing evolution, Anthropic, Google, OpenAI, and other leading AI companies are actively standardizing the communication protocols between models, agents, tools, and enterprise systems. Their strategic calculus is that the communication layer itself will not be the decisive factor in determining market winners. Instead, they anticipate victory by becoming the preferred destination where AI-assisted work naturally accumulates and flourishes. This strategic play involves the generous distribution of code, which will undoubtedly boost developer productivity across the board, while simultaneously generating substantial financial returns for a select few. The game, as they say, is on.