The compute AI lacks isn’t all on the training side. Whether an email should be escalated, which queue a support ticket should go into, whether content violates the rules, whether an on-chain operation should execute automatically—these tasks need compute that is verifiable, private, and paid for per call, not another data center to rent.
Phones are on the other end of the spectrum. There are billions of them, their chips are capable, and they have hardware isolation. Most of the time, they’re just sitting around plugged in. Acurast pools this idle compute into a decentralized cloud. It’s not another phone-mining pitch.
Install Processor on a phone and it becomes a node. Developers submit code, scheduling instructions, and resource requirements. The network assigns tasks based on performance; once they’re done, it produces a hardware-signed proof, and payment is settled on-chain. Tasks run in an isolated environment, so in theory, even the person holding the phone can’t see the plaintext.
Containers can be deployed, and small-model inference works too. The API is also compatible with OpenAI. Rewards are calculated based on single-core and multi-core performance, memory, and storage. Around 70% of inflation goes to staked compute, with devices actively doing work getting a little extra. It’s suited to small tasks and use cases where privacy and verifiability matter; competing with data-center GPUs on training isn’t realistic.
The mainnet and $ACU launched on January 20. By late September, more than 313,000 phones had connected; reports put the number at around 318,000, across more than 175 countries, with roughly 1 billion on-chain transactions.
Saying it can replace AWS is a bit of a stretch. A phone can’t match a training cluster for compute, cooling, or uptime.
It’s more like it fits into a gap: cloud providers can’t be bothered to serve it properly, while GPU-based DePIN projects aren’t good at it. Phones have the advantages of widespread hardware isolation, distributed nodes, and low marginal electricity costs. Recruiting people is easy; finding work isn’t. If phones are mostly earning inflation, the more phones there are, the less each one gets.
$ACU has a total supply of 1 billion and 5% annual inflation. The public sale price was $0.09; around October 2026, it was about $0.13, below the January peak. Put simply, Acurast has shown that a phone can be a verifiable compute node. It hasn’t yet shown that this network can become a cloud people are willing to pay for monthly. That gap is where phone-based compute remains an unfilled niche in AI and DePIN.
Phones are on the other end of the spectrum. There are billions of them, their chips are capable, and they have hardware isolation. Most of the time, they’re just sitting around plugged in. Acurast pools this idle compute into a decentralized cloud. It’s not another phone-mining pitch.
Install Processor on a phone and it becomes a node. Developers submit code, scheduling instructions, and resource requirements. The network assigns tasks based on performance; once they’re done, it produces a hardware-signed proof, and payment is settled on-chain. Tasks run in an isolated environment, so in theory, even the person holding the phone can’t see the plaintext.
Containers can be deployed, and small-model inference works too. The API is also compatible with OpenAI. Rewards are calculated based on single-core and multi-core performance, memory, and storage. Around 70% of inflation goes to staked compute, with devices actively doing work getting a little extra. It’s suited to small tasks and use cases where privacy and verifiability matter; competing with data-center GPUs on training isn’t realistic.
The mainnet and $ACU launched on January 20. By late September, more than 313,000 phones had connected; reports put the number at around 318,000, across more than 175 countries, with roughly 1 billion on-chain transactions.
Saying it can replace AWS is a bit of a stretch. A phone can’t match a training cluster for compute, cooling, or uptime.
It’s more like it fits into a gap: cloud providers can’t be bothered to serve it properly, while GPU-based DePIN projects aren’t good at it. Phones have the advantages of widespread hardware isolation, distributed nodes, and low marginal electricity costs. Recruiting people is easy; finding work isn’t. If phones are mostly earning inflation, the more phones there are, the less each one gets.
$ACU has a total supply of 1 billion and 5% annual inflation. The public sale price was $0.09; around October 2026, it was about $0.13, below the January peak. Put simply, Acurast has shown that a phone can be a verifiable compute node. It hasn’t yet shown that this network can become a cloud people are willing to pay for monthly. That gap is where phone-based compute remains an unfilled niche in AI and DePIN.