In 1997 nobody could tell you why you'd want a gigabit connection. AI is in 1997.

The argument that everyday AI tasks will become ‘good enough,’ making advanced AI less valuable, measures the wrong thing by focusing on known tasks. Just as high-speed internet created demand for things not yet built, ‘good enough’ AI will likely spur new demands and create bigger jobs beyond simple task completion. The true value lies not in the one-time hardware purchase but in the recurring relationship and who controls it, with Apple indicating future fees for server-side AI features.

• The argument that everyday AI tasks will become ‘good enough’ overlooks how past technological advancements, like gigabit internet, created new demands rather than just speeding up existing ones.
• As AI becomes cheaper and more capable, it will lower the barrier for writing software, enabling new applications that reveal previously unknown user demands for AI.
• Apple’s hardware ownership provides an opportunity when a need arises, but it does not guarantee long-term dominance, as value created by cheap capacity differs from money retained by suppliers.
• Accumulated memory can lock users in, but a superior model can rebuild context from existing data rather than a reason that was never explicitly recorded.
  https://bender.layer3.press/articles/956010fa-5289-4fb4-831a-ba53647d1ddb