The surge in AI-assisted development has brought a new kind of vendor lock-in, but not the sort the industry has faced before. Instead of proprietary languages or rigid enterprise suites, the very tool that writes code is now the source of dependency. With over 70% of developers using AI coding tools, and research showing that productivity gains are real, the engineering world is becoming heavily reliant on paid external AI services.
This reliance raises a critical question: will development teams consume AI on their own terms, or on someone else's? The answer may lie in the lessons of open infrastructure. History shows that open standards, protocols, and foundations tend to win in the long run, from the early internet to cloud-native computing. The same pattern is emerging in the AI space.
The twin tracks of AI development
Current AI development travels two parallel paths. On one side, open-source AI is thriving. Thousands of open-weight models are available on community hubs, and new breakthroughs come from universities and labs using frameworks like OpenClaw. On the other side, many organizations are building their AI strategies on proprietary platforms offered by a handful of large providers. These platforms wrap open components in closed interfaces that trade short-term speed for long-term constraints.
The behavior of some providers has drawn attention: blocking access to models over policy violations, shutting off competitors, and limiting interoperability. Such moves do not bode well for an AI economy already built on shaky unit economics. As systems become less interoperable, organizations may be forced to standardize on a single stack across data pipelines, models, and decision logic, increasing exposure when platforms change direction, raise prices, or fall behind technically.
Why openness matters now
Platform lock-in creates direct business risk. Migrating enterprise software can cost upwards of $100,000, and closed systems make those costs unavoidable. Open standards, interfaces, and infrastructure provide a necessary hedge. They allow enterprises to swap out models, agents, data, hardware, and orchestration as needed, without re-architecting entire AI systems.
Openness also addresses concentration risk at scale. If AI replaces human labor en masse, economic value concentrated in a handful of companies could create an order of magnitude bigger problem than in any previous wave of computing. Open foundations ensure that value remains distributed and that adaptability is built into the system.
Growing momentum toward openness
Industry momentum toward open AI infrastructure is building. Donations of protocols like the Model Context Protocol (MCP) and agents to foundations, as well as the establishment of new vendor-neutral homes for open-source AI technologies, signal a push for interoperability. These moves help ensure long-term support and care, much like the early internet benefited from standards bodies.
However, some veterans note that the AI ecosystem lacks the same level of organized stewardship that the internet had. All hope is that the industry will adapt and create a durable standards layer for AI.
Key layers for open infrastructure
Several layers of the AI stack stand out as especially important for openness. First is the model itself: open-weight models allow for trust, inspection, and customization. Second, the connective tissue between AI components—open APIs, metadata standards, identity and policy frameworks, and protocols for how models and agents communicate—enables flexibility. Third, cloud-native architectural standards, such as Kubernetes, provide a vendor-neutral deployment option for data, workloads, and AI components.
Experts agree that MCP has become the critical link between AI agents and the broader API ecosystem. Getting that protocol right could unlock the same level of interoperability that the Web 2.0 era enjoyed. A gap remains around standard API conventions for models, but calls for commitments from providers may eventually be answered.
Historical precedent
The pattern of open infrastructure winning is not new. Linux became the default operating system because it offered a common, vendor-neutral foundation. Docker and Kubernetes revolutionized cloud-native computing by providing open standards. The history of the internet itself, built on open protocols, shows that closed ecosystems eventually lose to flexible, collaborative ones.
While it is too early to say whether AI will develop a similarly durable standards layer, growth of open-source AI models and tools suggests a positive trajectory. The current dominance of proprietary platforms may be temporary. As systems become more distributed and agent-driven, closed ecosystems will struggle to scale. Open infrastructure offers the control, portability, and choice that enterprises need to survive market corrections and adapt to evolving conditions.
Source: InfoWorld News