“Dan Ives states that for widespread AI adoption, the cost of tokenization for large language models must continue to decrease.”