New Industries and Business Models

Economic disruption always has two sides. The same wave that erodes old industries deposits the sediment from which new ones are built. Steam displaced sail-powered shipping but created the railroad industry. The internet displaced travel agencies and record stores but created e-commerce, cloud computing, and the creator economy. AI is already producing new industries, new business models, and new economic activity that did not exist a decade ago — and the pace of that creation is accelerating.

AI-Native Industries

An AI-native industry is one that could not exist — or could not exist at meaningful scale — without AI as a core capability. Several are already established and growing rapidly. AI infrastructure: The companies that train and serve foundation models (OpenAI, Anthropic, Google DeepMind, Mistral, xAI) and the cloud providers that sell the compute those models require (AWS, Google Cloud, Microsoft Azure) constitute an industry that barely existed in 2017 and was worth hundreds of billions of dollars by 2024. Chip manufacturers, particularly Nvidia, have seen valuations multiply by factors of ten or more as demand for AI-specific processors has surged. AI-assisted drug discovery: Firms like Isomorphic Labs, Recursion Pharmaceuticals, and Exscientia use AI to identify drug candidates, predict protein interactions, and design clinical trials at speeds and costs that were previously impossible. Traditional drug discovery takes 10-15 years and costs over a billion dollars per approved drug. AI-native firms claim they can compress multiple stages substantially, though clinical validation remains slow. AI-native vertical SaaS: Dozens of startups have built specialized AI tools for narrow professional domains — AI for contract review (Ironclad, Harvey), AI for medical documentation (Nuance, Suki), AI for code review (Cursor, Codeium), AI for financial analysis (Kensho, AlphaSense). These firms do not sell general AI — they sell AI deeply integrated into the workflows of a specific profession, commanding premium prices because the AI handles tasks that previously required expensive human time.

Incumbent industries are also being transformed from within, producing new business models inside existing sectors. These are not new industries but new ways of competing within old ones. Retail and logistics: Amazon and Walmart now use AI at nearly every step of their operations — demand forecasting, warehouse robotics, delivery route optimization, personalized recommendation engines, and dynamic pricing. Firms that cannot match this AI-driven efficiency face structural cost disadvantages. Financial services: Algorithmic trading already accounts for the majority of U.S. equity market volume. AI-driven credit scoring models have expanded lending access while also raising concerns about discriminatory proxies. Insurance underwriting is shifting from actuarial tables to continuous behavioral data. The speed and cost advantages are real; so are the systemic risks — AI trading systems interacting can produce flash crashes and correlated failures. Healthcare delivery: AI triage tools are deployed in emergency departments. AI-read mammograms and chest X-rays operate at diagnostic accuracy comparable to specialist radiologists in controlled studies. Remote patient monitoring systems use AI to flag deterioration before clinical staff notice. Healthcare AI does not eliminate physicians — it changes which physicians are needed, in what numbers, and what tasks they spend time on. Media and creative industries: Generative AI models produce commercial-quality images, music, video, and written content at near-zero marginal cost. Stock photo agencies have seen revenue decline as generated images displace licensed photography. News organizations are experimenting with AI-written earnings reports. These changes create genuine economic value — and genuine displacement of creative workers.

New Economic Models: Platform AI and the API Economy

One of the most significant economic structures AI has produced is the foundation model API economy. Companies like OpenAI, Anthropic, Google, and Mistral train powerful models and then sell access to those models as an API — a programmable interface through which any developer or business can incorporate the model's capabilities into their own products, paying per query or per token. This creates a two-sided platform dynamic. On one side: AI companies competing on model quality, safety, cost, and reliability. On the other: thousands of application builders who layer industry-specific logic on top of the base capability. The application builder does not need to train a model — they simply access capability as a service. This dramatically lowers the barrier to building AI-powered products. The result has been an explosion of new products: AI tutoring apps, legal research tools, code assistants, mental health chatbots, customer service agents, and scientific writing aids — most of them built by small teams with no AI research capability of their own. The API economy functions similarly to the app store model that followed the smartphone: a platform with strong infrastructure advantages enables a long tail of specialized applications. The risks of this structure are real: if one or two foundation model providers control the API layer, they can set prices, alter terms, and potentially cut off competitors who build on their infrastructure — a dependency that makes the application builder ecosystem fragile. Vertical integration (a foundation model company building application-layer products) creates direct conflicts of interest with the developers who build on their APIs.