不过，对于该指数的走软亦有较乐观的解读。虽然Token的单价自2023年以来已下跌超过90%，但市场的整体支出却比去年翻倍，意味著便宜的Token反而扩大了整体市场规模。若这仅是市场的短暂消化，辉达等晶片与数据中心厂商的看涨前景依然成立。
相反地，若消费者付费意愿已达瓶颈，加上美欧等国近日加强对Anthropic及OpenAI等尖端模型的监管，使得企业出于合规与部署成本考量而转用低阶模型，那么这波高昂的硬体资本支出将面临危机，因为支撑此波繁荣的是企业的定价权而非单纯的晶片需求。
As markets grow uneasy about whether the massive investments in artificial intelligence will pay off, the Silicon Data LLM Token Expenditure Index has dropped nearly 20% from its peak. This decline alerts stock investors that AI companies may be losing pricing power among cost-sensitive customers, with rumors of OpenAI delaying its IPO further highlighting profitability challenges.
However, a more positive interpretation suggests this decline represents a healthy digestion period, as token prices have plunged over 90% since 2023 while total expenditure has doubled, thereby expanding the market. If this is merely a temporary demand shift, the bull case for GPU manufacturers, memory makers, and data centers remains valid.
Conversely, if customer willingness to pay has peaked and recent regulatory pressures from the US and EU prompt CFOs to route workloads to cheaper models, the high-priced AI trade could crack. This suggests the sector's long-term sustainability hinges on corporate pricing power rather than just hardware availability.