AI Value Shifts to 'Intelligence Stack'
As foundational AI models become commoditized, value is reportedly accruing higher up in the 'intelligence stack'. This new paradigm suggests market winners will be platforms that differentiate on orchestration, workflow integration, compliance, and verticalization, rather than on the underlying models themselves.
- The global vertical AI market, which focuses on industry-specific solutions, was valued at over $10 billion in 2024 and is projected to grow at a CAGR of over 21%, indicating strong demand for specialized AI applications. - Microsoft CEO Satya Nadella has stated that foundational models are becoming commoditized, suggesting that a lasting competitive edge comes from integrating AI into a "full system stack and great successful products" rather than model supremacy alone. - AI orchestration platforms like Domo, Apache Airflow, and IBM watsonx Orchestrate act as a "conductor" for enterprise AI, coordinating diverse models and systems to work together and automate complex, multi-step workflows. - The cost to access leading AI models has dropped significantly, with OpenAI's token pricing for its GPT series falling by over 80% from 2023 to 2024 due to increased competition and efficiency gains. - AI governance platforms have emerged as a critical component of the stack, with tools from companies like OneTrust, Credo AI, and Fiddler AI helping organizations manage model risk, ensure compliance with regulations like the EU AI Act, and reduce ethical bias. - A recent MIT report found that 95% of generative AI pilots fail to move into production, often due to a lack of orchestration, siloed data, and no clear connection to business outcomes. - Investors are increasingly rewarding companies that embed AI into their technology stacks with higher valuations; AI-focused companies can command revenue multiples of 25x or more, compared to around 7x for traditional SaaS companies. - Google Cloud's VP of Global Startups, Darren Mowry, has warned that AI startups focused on simple "wrappers" around large language models or acting as model aggregators face extinction as hyperscalers like AWS and Azure build similar multi-model orchestration capabilities directly into their platforms.
