Honestly, Acurast staking is completely different from what I initially thought PoS was.
It’s not about locking up tokens to earn interest. Instead, you use ACU directly as collateral to guarantee how much real computing power you can deliver. Operators must declare on-chain how much CPU, memory, and storage they can provide, then stake the corresponding amount of ACU. Can’t deliver? You get slashed in proportion to the shortfall.
What’s interesting is that they don’t penalize everything right away. The network has an epoch roughly every 90 minutes, during which it measures the computing power phones can actually provide.
If you made a commitment but fell short, only the undelivered portion is penalized; you still receive rewards as usual for what you did deliver.
A little fluctuation won’t get you wiped out, and anyone who spots a shortcoming in a node can trigger a slash. After a slash, 90% of the ACU is burned, reducing the circulating supply.
People actually running it say they usually set their commitments at around 80% of measured capacity as a safety buffer. Otherwise, an ordinary phone dropping offline briefly or experiencing performance fluctuations could easily get penalized.
Put simply, this system ties uptime and verifiability directly to the token economy, without relying on SLAs from centralized cloud providers.
For developers, backends, AI inference, oracles, and other workloads can all run on phones around the world, with an economic layer backing the quality of delivery.
It’s not about locking up tokens to earn interest. Instead, you use ACU directly as collateral to guarantee how much real computing power you can deliver. Operators must declare on-chain how much CPU, memory, and storage they can provide, then stake the corresponding amount of ACU. Can’t deliver? You get slashed in proportion to the shortfall.
What’s interesting is that they don’t penalize everything right away. The network has an epoch roughly every 90 minutes, during which it measures the computing power phones can actually provide.
If you made a commitment but fell short, only the undelivered portion is penalized; you still receive rewards as usual for what you did deliver.
A little fluctuation won’t get you wiped out, and anyone who spots a shortcoming in a node can trigger a slash. After a slash, 90% of the ACU is burned, reducing the circulating supply.
People actually running it say they usually set their commitments at around 80% of measured capacity as a safety buffer. Otherwise, an ordinary phone dropping offline briefly or experiencing performance fluctuations could easily get penalized.
Put simply, this system ties uptime and verifiability directly to the token economy, without relying on SLAs from centralized cloud providers.
For developers, backends, AI inference, oracles, and other workloads can all run on phones around the world, with an economic layer backing the quality of delivery.