Network Economics & Swarm Scalability

The Shardy Network is designed to thrive on “distributed weakness”—the collective power of millions of standard consumer devices. This document justifies the economic feasibility of low-power nodes and explains why a decentralized mesh allows for significantly lower compute pricing while remaining profitable for the average user.

1. The Economics of the Individual Node

Even if a node delivers only 2–3 GFLOPS (typical for integrated graphics or mobile browser tabs), the system maintains a positive economic loop. In the current runtime, admission for standard tasks requires a higher GFLOPS threshold, but lighter tasks (lite/super_lite) still benefit from weak-node participation.

Profitability Calculation (Case Study: EU/Spain)

Let’s calculate the earnings for a baseline node providing 3 GFLOPS ($0.003$ TFLOPS).

• Hourly Workload: $0.003 \text{ TFLOPS} \times 3600 \text{ sec} = 10.8 \text{ TFLOP}$ total work per hour.
• Hourly Revenue: At a network rate of $0.035$ per 1 TFLOP, the node earns: $10.8 \times 0.035 \approx \mathbf{$0.378}$ per hour.
• Daily Revenue: Calculated at 100% load: $\approx \mathbf{$9.07}$ per day.

Cost Comparison (Electricity)

In a high-cost energy market like the European Union (avg. €0.15–0.25 per kWh):

• Power Consumption: A laptop or PC under integrated GPU load consumes roughly 50–100W ($0.05–0.1$ kW).
• Electricity Cost: $0.1 \text{ kW} \times €0.20 = \mathbf{€0.02}$ per hour.
• Net Margin: $$0.37$ (Revenue) $-$ $€0.02$ (Cost) = High Positive Margin.

Conclusion: Participation remains highly profitable for the user even on “weak” hardware, as the value of AI-compute tasks significantly exceeds the utility cost of consumer electricity.

2. The Strategic Advantage of Scale

The “Weak Swarm” approach provides Shardy with structural advantages over traditional centralized data centers (H100/A100 clusters).

A. The Power of Many (Swarm Intelligence)

While a single node is weak, the network is massive.

• 1,000,000 nodes $\times$ 3 GFLOPS = 3 PetaFLOPS of total decentralized power.
• This matches the performance of the world’s most powerful supercomputers without the overhead of building massive facilities, cooling systems, or purchasing hardware.

B. Task Partitioning

Modern AI workloads can be partitioned into many independent tasks. In Shardy, each task is assigned one or more deliveries, and redundancy ensures result agreement.

C. Zero Barrier to Entry

By targeting 2–3 GFLOPS, Shardy eliminates the barrier to entry. We do not need 1,000 professional miners with expensive rigs; we attract 1,000,000 regular citizens who keep a browser tab open. This leads to viral, exponential network growth.

D. True Edge Computing (Geographic Proximity)

A million nodes spread across every continent reduces latency. Data is processed physically close to the end-user. For interactive AI applications, this geographic distribution provides a speed advantage that centralized clouds cannot match.

3. Justifying Disruptive Pricing

Traditional cloud providers must factor in hardware depreciation, real estate, and massive cooling costs into their pricing. Shardy bypasses these.

• Zero Capital Expenditure (CapEx): Shardy users provide their own hardware and cooling.
• Optimized Utilization: By utilizing “idle” cycles of devices that are already powered on, the network can offer competitive pricing while keeping contributors profitable.

4. Economic Mechanics (Aligned with Current Runtime)

• Rewards are distributed on task_verified events.
• Slashing applies on consensus_mismatch or invalid proofs.
• Stake-weighted rewards are computed in the state machine (computeStakeWeightedReward).