Own or Be Owned: Why Every Company Needs Its Own AI Model (Yash Patil, Co-Founder & CEO of Applied Compute)

2026-06-02 · Source: The Generalist · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Intermediate, extended

Summary

Yash Patil, Co-Founder & CEO of Applied Compute, advocates for companies to develop and own custom AI models, citing the "own or be owned" thesis. His \$1.3 billion company trains specialized models on proprietary business data, leveraging open-weight models and advanced post-training techniques. This approach addresses issues like frontier model guardrails, anti-competitive concerns, and the high cost of general-purpose AI, which Patil likens to "cooking with a blowtorch." Applied Compute's method focuses on optimizing for specific tasks, cost, latency, and data ownership, enabling clients like DoorDash to achieve superior performance for niche applications. Patil also predicts AI's economic rollout will span decades due to change management and data readiness challenges within large organizations.

Key takeaway

For Directors of AI/ML or AI Product Managers evaluating their AI strategy, relying solely on large frontier models risks dependency, high costs, and limited differentiation. You should explore custom model development on your proprietary data to gain control over capabilities, optimize for specific business tasks, and build a competitive advantage. Prioritize internal data readiness and invest in robust ML infrastructure to support continuous model improvement.

Key insights

Companies must own custom AI models to control capabilities, optimize costs, and differentiate from competitors.

Principles

• Frontier models are often the wrong tool for specific business tasks.
• Proprietary evals define model value and differentiation for enterprises.
• Continuous training loops are essential for model adaptation and improvement.

Method

Applied Compute trains custom, task-specific models on customer data using open-weight models, serving them at scale, and continuously improving them via production usage data and advanced ML infrastructure.

In practice

• Develop task-specific models to optimize for cost, latency, and performance.
• Codify internal procedures into proprietary evals for model training.
• Implement asynchronous sampling and training for efficient GPU utilization.

Topics

• Custom AI Models
• Open-Weight Models
• Post-Training
• ML Infrastructure
• AI Strategy
• Data Ownership
• Compute Optimization

Best for: Investor, CTO, VP of Engineering/Data, Director of AI/ML, AI Product Manager, Entrepreneur

Related on AIssential

• To Train or Not to Train — Post-training specialized models on open-weights bases offers AI application companies economic, latency, and differentiation advantages, despite rapid base model evolution.
• Adaptive ML boss: To win in AI, businesses must build with open source — Open-source AI models offer businesses a strategic advantage over proprietary solutions.
• Your Company Doesn’t Need an AI Strategy — Companies must build AI learning systems that compound human and token capital, not just rely on vendor-centric AI strategies.
• Microsoft’s AI chief on the greatest game of catchup ever played — Microsoft is aggressively building proprietary AI models, chips, and integrated software to achieve cost leadership and deep product integration, aiming to surpass competitors.
• Why companies build their own model on top of someone else’s? — The provided content indicates an error page from Medium, stating "Apologies, but something went wrong on our end." It suggests refreshing the page, checking Medium's site status, or finding other content to read.
• Custom model training. Domain-specific language models. | Mistral - Mistral AI — Mistral AI enables enterprises to build fully-owned, domain-specialized LLMs through collaborative customization and flexible deployment.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Generalist.