Is AI Like the Internet? Where the Analogy Helps—and Where It Breaks

AI is like the internet in one useful sense: both are enabling technologies that become valuable through the products, workflows, infrastructure, and institutions built around them. The analogy becomes misleading when it implies that AI adoption will follow the same path or create the same economics.

Similarity: the infrastructure matters before the applications mature

The early web needed networks, browsers, servers, payment systems, search, and standards before many familiar businesses became practical. AI applications likewise depend on models, compute, data pipelines, evaluation, identity, security, and integration into existing software.

In both cases, infrastructure can look overbuilt or underused before complementary applications catch up.

Similarity: adoption is not the same as productivity

Giving an organization a new technology does not automatically improve results. Workflows, incentives, training, data quality, process ownership, and measurement must change too. A company can buy AI tools and still make the old process slower by adding poorly designed review steps.

Similarity: new distribution creates new businesses

The web lowered the cost of distributing information and software. AI lowers the cost of producing or transforming some information and software. That can make previously uneconomic services viable while also making generic output easier to copy.

Difference: AI produces answers, not just connections

The internet primarily connected people, documents, services, and machines. Generative AI can synthesize, classify, predict, write, generate images, and take actions through tools. That moves the technology closer to the work product itself.

Difference: outputs can be probabilistic and wrong

A network packet either arrives or does not; an AI-generated answer can sound plausible while being incorrect. That makes evaluation, provenance, human review, and constrained tool permissions important parts of the system rather than optional polish.

Difference: capability can be concentrated

Building frontier-scale models requires substantial compute, data-center capacity, engineering, and capital. Applications can still be decentralized, but important layers of the stack may be controlled by a smaller number of providers than the open-web analogy suggests.

Difference: AI can automate parts of knowledge work directly

The internet changed where work happened and how information moved. AI can also perform parts of drafting, coding, analysis, support, design, and operations. That means job change can happen inside a role: some tasks become cheaper, others become more important, and new review/accountability tasks appear.

What the internet analogy is good for

• Expect complementary infrastructure and organizational change to matter.
• Do not confuse early hype with the final set of useful applications.
• Assume business models will change as distribution and production costs change.
• Expect standards, regulation, and trust mechanisms to evolve after adoption begins.

What the analogy cannot tell you

It cannot tell you which AI company wins, how many jobs change, how fast productivity appears, or where regulation settles. Those are empirical questions, not consequences of historical rhyme.

Use the internet as a lens for asking better questions—not as a script for predicting the future.