ECLIPSE — POWERED BY APERION
What you use. What you earn.
What you carry forward.
Illustrative economics · final rates may differ
REUSE FIRST · MEASURE THE DIFFERENCE
Community Compute starts with compatible equipment people already own. Our goal is to make better use of that investment, extend useful equipment life, and reduce pressure to manufacture more hardware and build more infrastructure for suitable workloads.
The IEA’s 2026 analysis estimates global data-center electricity consumption at 485 terawatt-hours in 2025. Its central projection reaches around 950 terawatt-hours in 2030. These figures cover all data-center workloads, not AI alone.
Source: IEA, Key Questions on Energy and AI (2026)Berkeley Lab estimated that U.S. data centers directly consumed about 66 billion liters of water in 2023. Electricity generation adds an indirect water footprint. Impacts depend on cooling, location and power sources; they are not one fixed amount per AI conversation.
Source: Berkeley Lab, 2024 U.S. Data Center Energy Usage Report, pp. 55–57Compare the same work at comparable quality and response time. Include host electricity, cooling, data transfers, coordination and retries—not just the processor’s rated power.
Track whether participating hardware is already owned, remains useful longer, and actually replaces a need for new capacity. Being connected alone does not prove avoided manufacturing or construction.
Use measured cooling data and location-specific power information where available. Clearly label estimates and unknowns when comparing water and emissions impacts.
Owners choose resource and availability limits. Account for electricity costs, heat, wear and household use alongside earnings, so participation serves both the contributor and the network.
This is the environmental goal and evaluation plan for a network under development. Existing hardware still consumes energy, and a modern data center can be more efficient for some jobs. Aperion has not yet demonstrated network-wide energy, water or emissions savings. Publish measured comparisons before claiming reductions.
THE MONTHLY BILL
These examples use a $9.99 subscription. Hardware and the one-time $0.99 app purchase are separate. Included usage is part of the subscription; it is not charged again.
Use earnings to reduce what you pay.
$6.00 covered · $3.99 remaining
Your earnings can cover the bill and leave credit.
$9.99 applied · $4.01 carried forward
Additional usage adds cost only after approval.
$6.00 covered · $6.99 remaining
INCLUDED USAGE & APPROVALS
For illustration, a plan includes $5.00 of metered usage. After $3.50 is used, $1.50 remains.
70% used. The $3.50 is within the allowance, so it adds $0 to the subscription bill.
The next paid action should show its price or an agreed maximum before you approve it. Declining keeps that additional paid action from starting.
Example of a cost to review before approval.
Different plans can include different allowances. A balance does not mean every task costs the same: workload, model, hardware, and runtime affect the cost.
Credits are internal accounting units. Dollars are the customer view. The $5.00 allowance is an example, not a finalized plan entitlement. Allowance rollover and overage rates must be stated in the applicable plan terms.
TRY THE NUMBERS
Change the amounts to explore the model. This calculator is an illustration; it does not access your account, approve usage, or charge you.
YOUR EXAMPLE
CAPACITY ON YOUR TERMS
The proposed host model matches work to processor capacity, RAM, and graphics memory. A job needs to fit the machine that runs it. Larger capacity can support a wider range of suitable jobs; it does not guarantee higher earnings.
Approved model files could be kept in a limited work area for reuse. You would control storage limits and availability. Contributing storage would not authorize access to personal documents or agent memories.
Resource controls, job verification, and payments must be tested before enrollment. The calculator estimates bill credits only; it does not benchmark your hardware or predict demand.
CONTRIBUTOR ECONOMICS
Contributors offer eligible capacity. Suitable jobs are matched to it, and accepted work is measured against an agreed rate. Availability alone does not guarantee paid work.
Assume 1,000 accepted compute minutes priced at $0.01 per minute:
1,000 × $0.01 = $10.00
With an illustrative 80% contributor share:
$10.00 × 80% = $8.00 earned
The remaining 20%, or $2.00, is platform revenue. These rates are examples; there is no promise that every device or job earns the same amount.
Suppose that work earns $8.00 and uses 0.8 kWh at $0.20 per kWh:
0.8 × $0.20 = $0.16 electricity
$8.00 − $0.16 = $7.84
That leaves $7.84 before equipment wear, cooling, internet costs, taxes, and any other expenses. Hardware efficiency and local electricity rates change the result. The assumed energy use is illustrative.
A contributor minute is a priced unit for a defined workload and hardware class—not a universal unit of AI capability. Published rates, verification, rejection rules, fees, and settlement terms must define what counts as payable work.
CARRY-FORWARD EXAMPLE
$14.00 in accepted earnings covers the $9.99 fee and leaves $4.01 in credit.
With no new earnings or extra usage, the next $9.99 bill uses that $4.01 credit:
$9.99 − $4.01 = $5.98 to pay
This explains the proposed billing-credit model. Bill credit is not automatically cash you can withdraw. Any cash payout option, expiry, refunds, taxes, and settlement timing require separate published terms.
THE NETWORK BEHIND THE NUMBERS
A subscription supports access to network services and the included usage allowance. Real costs include compute providers, external services where used, coordination, security, operations, and support.
Efficient routing and useful existing capacity can help reduce costs. Subscription revenue is not automatically profit, and we do not claim that 100% is paid directly to compute contributors.
Choose eligible hardware, availability, and limits. Accepted work can offset your network bill. Earnings depend on demand, suitability, verification, and the agreed rates.
More compatible, available compute can strengthen the network. More customers also add demand; customer count alone does not create processing capacity.
Providing compute does not grant access to your Key, saved memories, files, microphone, or personal tools. Those require separate permissions.
Work must follow an appropriate privacy route. Encryption in transit and at rest does not by itself conceal data from the machine processing it.
Compute earnings compensate accepted work. Membership or governance rights need explicit terms; earning a compute payment does not automatically grant equity or voting rights.
Before participation, contributors should be able to see the rate, platform share, resource limits, acceptance rules, and how earnings will be applied.