#AI股持续上涨还有哪些投资机会
AI is 60% cheaper—doubling usage isn’t enough
On September 22, Anthropic $GOOGLB released Opus 5.5.
Compared with its own Fable 5.1, the per-million pricing for input/output tokens dropped from $10/$50 to $4/$20—both 60% lower.
Let’s do the math: assume a batch of business workloads are fully migrated to the new model, keeping the input/output ratio and discount the same. If usage doubles, revenue is still 20% lower; only when usage reaches 2.5 times the original does revenue break even.
This is just a replacement scenario, not actual operating results yet. But it adds a credibility check to AI valuation: when call volume surges, does revenue keep up?
Lower prices first benefit application companies that use paid AI model services. Whether model vendors benefit depends on new demand—and whether unit costs can fall faster.
Going forward, when you see “usage hits a new high,” ask one more question: for each unit of usage, how much can they still charge?
#AI投资 #technology stocks
AI is 60% cheaper—doubling usage isn’t enough
On September 22, Anthropic $GOOGLB released Opus 5.5.
Compared with its own Fable 5.1, the per-million pricing for input/output tokens dropped from $10/$50 to $4/$20—both 60% lower.
Let’s do the math: assume a batch of business workloads are fully migrated to the new model, keeping the input/output ratio and discount the same. If usage doubles, revenue is still 20% lower; only when usage reaches 2.5 times the original does revenue break even.
This is just a replacement scenario, not actual operating results yet. But it adds a credibility check to AI valuation: when call volume surges, does revenue keep up?
Lower prices first benefit application companies that use paid AI model services. Whether model vendors benefit depends on new demand—and whether unit costs can fall faster.
Going forward, when you see “usage hits a new high,” ask one more question: for each unit of usage, how much can they still charge?
#AI投资 #technology stocks
