A fundamental question for investors in artificial intelligence businesses centres on whether the value created by an AI token surpasses its cost. Microsoft chief executive Satya Nadella recently underscored this as a key measure of success. ‘Tokens’, the units of data processed by AI models, are a relatively new concept in investment lexicon, signifying the nascent stage of this critical metric. Early AI deployments often prove inefficient, akin to a technology still finding its footing.
Improvement in token value is essential for AI to deliver its promised productivity gains, crucial for economic stability. However, companies face substantial AI costs, with modest tool usage potentially reaching A$7,200 per employee annually. Some initially encouraged maximum token usage, but many are now re-evaluating, recognising usage does not automatically equal value. Despite this, investors continue to back the entire AI value chain, from data centres to applications. Hyperscalers are projected to invest US$5 trillion in AI-related capital spending between 2026 and 2030, a 400 per cent increase, with significant debt funding leaving little margin for error.
For investors, an accelerating pace of technological cycles, particularly in chip design, poses a significant obsolescence risk. Today’s competitive advantages can quickly erode as more efficient technologies emerge, complicating the assessment of earnings durability and credit risk. This dynamic makes strong underwriting and clear visibility over long-term cash flows critical, especially in capital-intensive sectors. Ultimately, while AI offers enormous potential, the investment case may simplify to one basic calculation: does incremental token usage create real economic benefit or merely add cost?
