What are AI products really selling: computing power, or time saved? These two needs shouldn’t be conflated: renting a machine is merely a supply of resources; users getting stable, usable results is much closer to product value. There’s currently no information confirming how many customers, how much revenue, or what computing power utilization each of these three networks has, so I won’t turn the industry’s interest in AI into a claim that they already have orders.
Rolling 24 hours, $TAO -3.32%. Rolling 24 hours, $RENDER -1.36%. Rolling 24 hours, $FET -3.96%. These are the respective changes in three spot prices, not task volume, customer demand, or computing power utilization, and there’s no evidence that the declines stem from the same cause. Price isn’t an order receipt, and a steeper drop doesn’t necessarily mean demand has weakened.
Providers bear the costs of equipment, bandwidth, and maintenance. Network incentives may attract resources to come online, but that’s an expense on the network side, not proof that customers have paid. A machine being online doesn’t mean it’s received a task, and completing a task once doesn’t prove reliable delivery or that customers are willing to buy again. To say that demand has been validated, we’d at least need to know who is buying, what they’re getting, and whether they’re paying on an ongoing basis. These are still things to watch.
Token value needs to be assessed separately: Do customers have to hold or pay with the token? Where do service fees ultimately go? If the product can settle payments without the token, increased usage may not create demand for the token; rewards paid to resource providers don’t count as customer revenue either. People using a product doesn’t automatically mean the token captures its usage value.
I’m interested in AI computing power, but more than a list of resources, I want to see ongoing tasks, reliable delivery, and repeat payments. Product demand, the supply and demand for computing power, network incentives, and token value are related, but they aren’t the same ledger. Until we know who is paying, why they keep paying, and how fees flow through to the token, an AI label isn’t enough.
Rolling 24 hours, $TAO -3.32%. Rolling 24 hours, $RENDER -1.36%. Rolling 24 hours, $FET -3.96%. These are the respective changes in three spot prices, not task volume, customer demand, or computing power utilization, and there’s no evidence that the declines stem from the same cause. Price isn’t an order receipt, and a steeper drop doesn’t necessarily mean demand has weakened.
Providers bear the costs of equipment, bandwidth, and maintenance. Network incentives may attract resources to come online, but that’s an expense on the network side, not proof that customers have paid. A machine being online doesn’t mean it’s received a task, and completing a task once doesn’t prove reliable delivery or that customers are willing to buy again. To say that demand has been validated, we’d at least need to know who is buying, what they’re getting, and whether they’re paying on an ongoing basis. These are still things to watch.
Token value needs to be assessed separately: Do customers have to hold or pay with the token? Where do service fees ultimately go? If the product can settle payments without the token, increased usage may not create demand for the token; rewards paid to resource providers don’t count as customer revenue either. People using a product doesn’t automatically mean the token captures its usage value.
I’m interested in AI computing power, but more than a list of resources, I want to see ongoing tasks, reliable delivery, and repeat payments. Product demand, the supply and demand for computing power, network incentives, and token value are related, but they aren’t the same ledger. Until we know who is paying, why they keep paying, and how fees flow through to the token, an AI label isn’t enough.
