Rope a Dope

2 min read

"Is that all you got, George?" Muhammad Ali faced the devastating punching power of the younger and stronger George Foreman in 1974.

"That don't hurt. I thought you were supposed to be bad," Ali taunted. Foreman swung harder, incredulous that fear of his firepower had no effect. Foreman was not supposed to beat Ali; he was supposed to pulverize him. But Ali executed what we now call "Rope a Dope" - losing every round, absorbing the punishment, but exhausting Foreman until by round 8 the young man had nothing left.

Knockout.

Muhammad Ali standing over George Foreman after the eighth-round knockout at the Rumble in the Jungle, 1974

Let's talk about Apple.

Apple is the laughingstock of the AI race, but may well win this war. Yesterday at WWDC they short-circuited the panic of rising token costs, announcing free inference on-device and even on their private cloud, for smaller developers. They are flexing the unique architecture of Apple Silicon and matching it with Google's Small Language Model advancements (see Gemma 4) to create the largest AI cloud in the world.

Made of iPhones.

That you pay for, happily.

Take a look at the CapEx run rates for some hyperscalers:

  • 🔵 Microsoft: ~$31.0B
  • 🔴 Alphabet: $35.7B
  • 🟠 Amazon: $44.0B
  • 🍎 Apple: $1.97B

Rope a Dope.