The Montgomery Summit 2026• Mar 16, 2026• 43:36ConferencePanel

AI Built in the Open: Systems, Scale, and the Future of Innovation

From The Montgomery Summit · 2026

Yvonne Wassenaar(Moderator, Arista, Braze, JFrog, Rubrik)•Charles Zedlewski(Panelist, Together.ai)•Robert Brennan(Panelist, OpenHands)•Nader Khalil(Panelist, NVIDIA)

Executive Summary

The business model for commercial open source in AI has shifted from selling licensed features for installable software to providing managed cloud services (APIs), where the competitive moat is operational efficiency and performance.
Open-weight models are seeing significant adoption, particularly by AI-native companies who switch from proprietary models at scale to improve performance and control, driving hundreds of millions in revenue for infrastructure providers like Together AI.
The entire AI industry, both open and proprietary, is fundamentally constrained by the availability of compute, capital, and energy, which are the primary limiters on growth.
Open source is foundational to the entire AI stack, with even closed-model providers relying on open source inference engines, indicating a hybrid future where open and closed systems coexist and interoperate.

11 quotes

Concerns Raised

The entire AI industry is fundamentally constrained by the availability of compute, capital, and energy.
Commercializing open source is challenging, requiring companies to find product-market fit twice and risking community backlash.
Accountability and support models are still maturing for when issues arise within complex enterprise open source AI stacks.

Opportunities Identified

The massive shift of AI-native companies to open-weight models at scale creates a multi-hundred-million-dollar market for specialized AI cloud providers.
AI coding agents have only penetrated 20-30% of the developer market, indicating significant room for growth and productivity gains.
Building a competitive moat through systems engineering and operational efficiency in providing AI services, as the underlying models become commoditized.

Key Themes

The New Commercial Open Source Playbook

The discussion highlights a fundamental shift in monetizing open source. The old model of holding back licensed functionality for on-premise software has been replaced by offering a managed cloud service where the open source components are accessed via an API. The company's intellectual property and value-add lie in the operational excellence, performance, and efficiency of the service they run, not in the code itself.

This theme is critical for investors and founders in the AI space, as it defines the modern, durable business model for open source companies, moving the competitive battleground from features to operational performance.

The Economic Power of Open-Weight Models

Speakers assert that open-weight models are not just a niche but a major economic force. AI-native companies are spending hundreds of millions on inference and are increasingly shifting to fine-tuned open-weight models to achieve superior performance for specific tasks. This trend is driving massive demand for AI-native cloud platforms that specialize in serving and customizing these models.

This demonstrates a clear market validation for open-weight models beyond ideological preference, showing they are a pragmatic choice for performance-driven, high-growth companies, creating a significant market opportunity.

Finding Product-Market Fit Twice

A key challenge for commercial open source companies is that they must achieve product-market fit twice. First, they need to build a vibrant community around their open source project. Second, they must find a separate product-market fit for a commercial offering that provides distinct value, typically for teams and enterprises (e.g., collaboration, security, integrations), without crippling the open source version.

This provides a crucial strategic framework for building a sustainable open source business, emphasizing the need for a deliberate and distinct commercialization strategy from day one, rather than simply trying to monetize a popular project later.

Compute as the Ultimate Constraint

A recurring point is that the primary bottleneck for the entire AI industry's expansion is not a lack of models or software innovation, but a physical shortage of compute, capital, and energy. This scarcity dictates the economics of the industry and shapes the competitive landscape, as access to and efficient use of infrastructure is paramount.

This reframes the strategic focus for the AI industry, suggesting that long-term winners will be those who can secure and efficiently manage physical infrastructure, making it a more fundamental moat than the models themselves.

The Symbiotic AI Stack

The panel argues against a binary view of open vs. closed AI. Instead, they describe a symbiotic stack where open source is the foundational layer for everyone. For instance, nearly all major model providers, whether open or closed, use an open source foundation for their inference engines. This interoperability is key, allowing developers to mix and match components to build the best possible applications.

This insight is important for developers and enterprise architects, showing that the future is not about choosing one camp but about intelligently integrating open and proprietary technologies to optimize for performance, cost, and control.

Get started free

Topics

AI Infrastructure Open Source Business Models Commercial Open Source (COSS)Foundation Models Large Language Models (LLMs)Open-Weight Models AI Agents Coding Agents Inference Engines Model Training Fine-Tuning Reinforcement Learning Developer Productivity Cloud Services Compute Constraints

Processed Apr 19, 2026 yt-dlp + mlx-whisper + Gemini