Acurast turns smartphones into TEE-powered cloud servers. Explore its DePIN model, ACU tokenomics, traction, risks, and growth potential.
Author: Akshat Thakur
Acurast is a DePIN compute network that turns ordinary smartphones into a distributed serverless cloud. Developers can deploy APIs, webhooks, agents, LLM inference and confidential workloads. Phones execute those jobs inside hardware TEEs and prove that the code ran unmodified. Providers receive ACU for supplying compute.
The pitch is not more GPUs. Traditional DePIN compute still relies heavily on data centers, racks and operators who control the hardware. Acurast takes a different approach. It uses the secure hardware already built into billions of smartphones.
The network can use idle or recycled devices as compute units. Acurast’s docs put the fleet at roughly 300,000 compute units. Its public X bio cites more than 301,000 phones, 1 billion onchain transactions and over 626,000 deployments across Android, iOS and Solana Mobile.
The timing also matters. Confidential compute and on-device AI have moved from emerging concepts into active infrastructure themes. AI agents need cheap and schedulable runtimes, while enterprises increasingly want execution they can verify without exposing sensitive workloads.
Acurast is already past its launch phase. Mainnet and the ACU token went live around late 2025 and early 2026. ACU trades around $0.12, with a market cap ranging roughly from $26 million to $45 million depending on the circulating-supply figure used. CoinMarketCap shows around 217 million circulating, while other figures are higher. FDV sits near $119 million against roughly 1 billion genesis supply.
The opportunity is therefore not a future launch. Acurast already has the network and token.
$ACU enters a crowded decentralized-compute market.
Akash, Render, io.net, Nosana and Gensyn compete for GPU and CPU workloads, particularly AI training and inference. They already have stronger mindshare around high-performance compute.
Phala, Oasis, Secret and Arcium approach the problem from confidential computing, offering TEE or MPC infrastructure at the protocol layer. Hyperscalers such as AWS, GCP and Azure remain the benchmark for enterprise workloads because they offer established infrastructure, regions and service-level agreements.
Phone-based DePIN networks also compete for idle-device participation. The risk is that these networks become reward farms rather than useful infrastructure.
Acurast does not win by offering more raw compute. A smartphone will not replace an H100 cluster for a large AI training workload.
Its narrower advantage is the combination of commodity phones, hardware security and distributed supply. Jobs can run inside the device’s TEE, preventing the phone owner from directly reading the workload. Developers can deploy without operating their own data center.
That creates a potential niche for webhooks, oracles, light inference, agent loops and confidential micro-jobs.
The threat model is also different from conventional cloud infrastructure. Hyperscalers can offer confidential VMs, but Acurast is built around federating large numbers of consumer devices rather than centralized servers. GPU DePIN networks can add confidential execution, but they still depend on purchased accelerators and physical infrastructure.
The distinction matters. Acurast is not competing with NVIDIA on performance.
Its real proposition is attested serverless computing on phones.
The important metric will therefore be paid deployments and the quality of those workloads, not the number of phones connected to the network.
Acurast has a doxxed Swiss team with a background that fits its mobile-security thesis.
Alessandro De Carli is a co-founder and serves as board president and CEO. He previously worked on mobile security at Credit Suisse and co-founded Papers AG. He also helped build AirGap, an air-gapped wallet that uses a spare phone as an offline signer.
His background follows a consistent theme. The smartphone is treated as a security module rather than simply a consumer device. He also has a University of Zurich MIS background and early Lightning-related research experience.
Pascal Brun is Acurast’s other co-founder. His background is more focused on systems engineering, operations and business development. He also came through Papers, the Swiss studio that incubated AirGap and Acurast.
The project operates through the Acurast Association in Zug and has a public team in the mid-teens.
The team also publishes runtime upgrades, GitHub releases and security post-mortems. In 2026, it disclosed double-claim and conversion-replay issues that were patched. The Association covered roughly 71,000 ACU in excess rewards related to the incidents.
That level of operational disclosure provides a stronger signal than an anonymous DePIN operation.
The more important proof is what the team has already shipped. AirGap is an established product, while Acurast has built a live network with hundreds of thousands of compute units and hundreds of thousands of deployments.
The main question is no longer whether the team can build mobile-security infrastructure.
It is whether that experience can translate into developer adoption against GPU marketplaces that already dominate AI compute mindshare.
Acurast has raised capital publicly and has named backers behind the project.
The disclosed funding stack includes roughly $11 million announced ahead of TGE across grants and token rounds. A CoinList sale in May 2025 raised around $5.4 million at $0.09 per ACU, implying roughly $90 million FDV at the time. Earlier 2023 and 2024 seed and strategic rounds added roughly $2 million in reported capital.
Some aggregators show higher lifetime figures when later public and secondary activity is included, so the cleanest reference is the approximately $11 million officially announced figure alongside the CoinList sale.
The investor base includes Gavin Wood, peaq and Leonard Dörlochter, the Tezos Foundation, Web3 Foundation, HV Capital, Scytale Digital, Sigma Capital, MN Capital and Ogle.
That gives Acurast a credible mix of Polkadot, DePIN and crypto infrastructure support, alongside the public CoinList distribution.
Funding is no longer the primary question. Mainnet and ACU are already live, so the project has moved beyond the stage where capital simply needs to keep development alive.
ACU also has an active role in the network. It funds compute payments and provider rewards, supports staked compute and participates in governance. Current documentation puts inflation at 5%, with 70% of inflation directed toward the staked-compute pool.
The more important issue is supply and demand.
Acurast has built a substantial supply of mobile compute. The network now needs developers to use that capacity for valuable workloads.
If paid deployments grow alongside the roughly 300,000-device network, the funding story looks increasingly credible. If workloads remain small or experimental, additional capital will not solve the core problem.
Acurast has raised enough to build and launch.
Now it needs to prove that developers will pay for what it built.
Acurast is already live. Genesis mainnet and $ACU launched in the late-2025 to January-2026 window, so this is no longer an incentivized-testnet story.
Users can run a Processor Lite on an everyday phone or Processor Core on a dedicated, factory-reset device. After onboarding, providers earn ACU for attested uptime. Developers can deploy through the CLI, SDK, Hub UI or Deploy Agent, which supports USDC payments on Base through x402 without requiring an ACU account.
The supported workload class includes APIs, cron jobs, Telegram bots, Puppeteer, Wasm, WordPress through Cargo and light LLM inference. Users can also stake or delegate ACU into Staked Compute and access hardware benchmarks covering CPU, RAM and storage.
The software is still actively shipping. Acurast’s runtime releases reached v0.26.x in September 2026, including security patches.
That does not make it an AWS replacement. Phones cannot match H100s for demanding AI workloads. Device churn, congestion and TEE limitations restrict what the network can handle.
The project has also experienced real bugs. Protocol notes document double-claim staking rewards and conversion-message replay issues that were patched. The Acurast Association covered roughly 71,000 ACU in excess rewards.
That is an important distinction. Acurast has a functioning network, consumer hardware and developer tooling, but it remains a young L1 and distributed-device cloud.
The product is real.
The open question is whether the workloads running on it become economically important enough to justify the network’s growing compute supply.
Acurast’s traction needs to separate phones that exist from jobs that generate economic demand.
The network has reached roughly 300,000 compute units. Its public figures cite more than 301,000 phones, 1 billion onchain transactions, 626,000+ deployments and coverage across 175+ countries. Testnet figures had already reached 149,000 phones and 492 million transactions before mainnet.
The hardware commitment is meaningful. Processor Core requires a dedicated phone running continuously, while Processor Lite still requires a real device and uptime. Multi-processor setups are also possible. This is closer to operating infrastructure than joining a Discord server.
The developer side is harder to validate. Acurast supports paid deployments through ACU and USDC via x402, with Cargo, WordPress and agent tutorials designed to make the network useful beyond simple testing.
Funding provides another signal. The CoinList sale raised roughly $5.4 million at $0.09, alongside earlier seed and strategic funding. The disclosed funding stack heading into TGE was around $11 million.
Acurast has also cited more than $200 million in assets touched across Bitcoin, Ethereum, Tezos, Polkadot and peaq. That figure should be treated as protocol-adjacent security usage rather than AUM.
The weaker signals are follower counts and raw deployment or transaction numbers. More than 113,000 X followers do not prove developer demand. Neither does a large device graph if most phones remain idle and collect emissions.
The key metric is revenue from actual developer workloads versus ACU emitted to providers.
A sold-out token sale proves demand for the token. A 300,000-phone network proves supply-side incentives worked.
Neither proves that developers need Acurast.
$ACU has already launched, so there is no TGE opportunity left. Investors are entering a mid-unlock, inflationary DePIN economy.
Genesis supply is 1 billion ACU, with annual inflation currently set at 5% and adjustable through governance. Around September 2026, ACU trades near $0.12, with market cap estimates ranging from roughly $26 million to $45 million depending on the circulating-supply figure used. FDV sits around $119 million to $124 million.
The token reached roughly $0.34 in January 2026 and fell to around $0.065 in June.
The next scheduled unlock is around September 20, 2026. Third-party calendars estimate roughly 27 million ACU, or about 2.7% of supply and roughly 6% to 7% of market cap. That is material for a token of this size.
The official allocation is more important than the headline circulating number. Team and advisors hold 24%, with a six-month cliff followed by 36 months of linear vesting. Community Treasury holds 24%, while Operational Funds hold 11.5%, both subject to three-month cliffs and 24-month vesting. Liquidity receives 10%, the CoinList public allocation 6.5%, early backers 6.5%, and early compute providers 6.5%.
Roughly 21.5%, or 215 million ACU, unlocked at TGE across the liquid public, listing and LP sleeves.
The token has genuine utility. Users spend ACU for compute, stake or commit it to improve provider rewards, participate in network security and use it for governance. Fees can also be burned as gas under the network’s architecture.
But inflation remains the key issue. Around 70% of inflation goes toward Staked Compute, with additional emissions for base compute rewards, the treasury and collators.
Without developer fees growing faster than inflation and unlocks, ACU risks becoming a work token driven more by provider rewards than genuine demand.
The structure is healthier than a 4% circulating supply at a multibillion-dollar FDV. It is still an inflationary token with multi-year unlocks.
The main airdrop seasons and CoinList window are closed. Early participation now means operating the network, deploying workloads or buying ACU on the secondary market.
The most direct route is running a Processor. Processor Lite works on an everyday phone, while Processor Core uses a dedicated Android device. Devices must meet the network’s requirements, including a non-rooted setup and locked bootloader. Providers earn rewards based on benchmarks and can increase their rewards through ACU staking.
Users can also stake or delegate ACU into Staked Compute. Rewards are distributed each epoch, while cooldown periods and hardware scores affect the eventual allocation.
Developers have another path. They can deploy real workloads through the CLI, SDK, Cargo or x402. Paying for compute creates demand on the opposite side of the network from providers earning ACU.
Buying ACU is the simplest route, but it is no longer an early allocation. The token already trades on the market, and the September 20 unlock is a supply event rather than an entry opportunity.
Existing recipients of MIST or Cloud Rebellion airdrop allocations can continue receiving vested tokens over 24 months. New users should not spend time chasing old point seasons.
The economics also need to be realistic.
Running processors requires hardware, electricity and uptime. The rewards need to justify those costs after accounting for ACU inflation and future unlocks.
Operating 20 dedicated phones to capture 5% annual inflation is not passive farming. It is a small infrastructure business.
Acurast’s early phase is therefore over. What remains is a choice between becoming a network provider, generating genuine compute demand or taking a liquid-market position in ACU.
The airdrop is over. The economic question is now simpler: is the return from running phones worth the hardware and electricity costs, or is ACU attractive at roughly $120 million FDV?
For new users, the phone-farming case is weak. Cloud Rebellion and MIST seasons have already paid out, with 10 million ACU vesting over 24 months. Running Processor Lite solely for rewards means entering after the early incentives have passed.
The economics make more sense if you already own compatible Android devices and have cheap power. Buying phones specifically to farm ACU turns the strategy into a hardware and electricity business. ACU has also fallen roughly 65% from its January 2026 high near $0.34.
Buying the token is a different decision. ACU trades around $0.12, with market cap estimates between $26 million and $45 million and FDV around $119 million to $124 million. The next major unlock is around September 20, with roughly 27 million ACU entering the market.
Compared with Render, Akash and io.net, Acurast is cheaper and offers a distinct mobile-security angle. But its valuation becomes harder to justify if the 300,000-device network remains mostly idle.
The expected-value case depends on one metric: paid compute demand.
If developers increasingly pay through ACU or USDC/x402 and that revenue grows faster than 5% inflation and scheduled unlocks, the network can support a stronger token thesis. If providers remain dependent on emissions, ACU is still primarily a work token with a strong TEE narrative.
The verdict is therefore straightforward.
Running new hardware is generally not worth it. ACU can justify only a small speculative position, not a core holding.
The September unlock and a potential return toward the June low around $0.065 should be treated as real downside scenarios, not impossible outcomes.
Acurast’s biggest risks are not missing another developer tutorial. They are weak utilization, token emissions, competition and regulation.
Akash, Render and io.net already own much of the decentralized-compute narrative and have GPU supply. Phala, Oasis and Arcium compete in confidential computing, while AWS offers confidential infrastructure with established enterprise procurement.
Acurast therefore needs to win a narrower market: cheap, attested micro-jobs that benefit from mobile TEEs. If those workloads remain on conventional servers, its large phone fleet becomes mostly ornamental.
The same applies to the DePIN narrative. A network with 300,000 phones sounds impressive until utilization remains low. Daily-driver devices can generate heartbeats and reward activity without creating meaningful production demand.
Regulation adds another major risk. Consumer devices processing third-party workloads create questions around data protection, export controls and cloud-provider obligations. Acurast is a Swiss association operating a global device network.
The token also combines compute payments, staking and 5% inflation. That creates another layer of regulatory and economic complexity. TEE protection does not automatically remove liability if a device in someone’s home processes third-party production data.
App-store restrictions could also damage the supply side quickly if Processor applications were removed or limited.
The token structure carries its own pressure. Team and advisors hold 24%, while treasury and operational allocations total another 35.5%. Those tokens vest into a market with a sub-$50 million market cap and ongoing 5% inflation.
Execution remains another concern. Device churn, Android attestation failures, iOS limitations and runtime bugs can all reduce reliability. Acurast has already dealt with double-claim and replay issues, although both were patched.
The core model is also familiar. Hardware joins the network, tokens incentivize supply and developers are expected to arrive later.
TEE attestation is the genuinely differentiated component.
Without paid jobs, however, Acurast remains another DePIN network trying to convert emissions into real utilization.
Acurast has already passed TGE and launched mainnet. The next catalysts are therefore about network usage and token supply.
The most immediate dated event is the September 20, 2026 unlock. Roughly 27 million ACU is expected to enter circulation, making it important for price regardless of new product announcements.
Beyond supply, the important catalysts are paid workloads. Runtime upgrades, Cargo and agent deployments, and USDC payments through x402 can demonstrate that developers are using Acurast for actual production jobs.
A named third-party customer would be an especially strong signal. It would show that the network is generating demand beyond internal testing, tutorials or ecosystem incentives.
The biggest positive catalyst would be sustained paid compute revenue exceeding provider emissions. That would begin shifting Acurast from an emission-funded DePIN toward an economically self-sustaining compute marketplace.
The downside catalysts are equally clear. Another rewards exploit, removal of Processor applications from stores or continued 5% inflation without meaningful utilization would weaken the thesis.
There is no reason to act because Acurast is supposedly still “early.” That phase is over.
For users considering ACU, waiting three months provides a useful test. If developer fees begin to exceed emissions and real workloads become visible, the network will have stronger evidence behind its valuation.
Trading ACU now can make sense as a liquid DePIN-beta position. Running new phones is different. That is a labor decision with hardware and electricity costs, not a simple token trade.
WATCH
Acurast is one of the more credible phone-based DePIN networks because the team has relevant mobile-security experience, the mainnet is live, TEEs are part of the architecture and the network has reached roughly 300,000 compute units.
The scale is interesting. Acurast cites more than 1 billion transactions, 626,000+ deployments and a network spanning more than 175 countries.
But the token is a different story.
ACU carries roughly $120 million FDV, 5% inflation, a 24% team allocation and ongoing unlocks. The network has demonstrated substantial supply-side participation, but it has not yet demonstrated that developer fees can consistently outrun emissions.
That distinction matters.
300,000 phones are not 300,000 customers.
The milestone that would change the rating is sustained, disclosed paid compute revenue from third-party workloads exceeding provider inflation for more than one quarter. ACU or USDC/x402 invoices would both count.
Until that happens, Acurast is a real network with an interesting TEE architecture, but its token remains an emission-heavy work asset.
WATCH.
All the opinions in this article are that of the author and in no way are financial advice. Our Crypto Talk and the author always suggest you do your own research in crypto and to never take anything as financial advice that you read on the internet. Check our Terms and conditions for more info.
