Green Computing - Hainan Free Trade Port AI Agent Association (Preparatory)

Compute-Power Synergy refers to the deep integration of computing infrastructure and power systems — allowing AI agent computing demand to follow the supply curve of clean energy, thereby enabling efficient green power utilization and significantly reducing compute costs. This is HNAIA\'s key technical proposition that distinguishes it from other AI organizations.

≤1.1

Target PUE
Lingshui Undersea Data Center

100%

Clean Energy Supply
Green Power Direct-Supply Node

A–G

Unified Energy Efficiency Rating
Mandatory Skill Label

0 Carbon

Zero-Carbon Skill Certification
Automatic Label

I. Lingshui Undersea Data Center

As the core node of the association\'s AI agent Skill library, the Lingshui Undersea Data Center uses seawater natural cooling to greatly reduce cooling energy consumption, targeting PUE ≤ 1.1 and accessing 100% clean energy supply. Member AI agent training and inference tasks can be preferentially scheduled to this node.

II. Hainan Power-Compute Unified Network

The Association promotes member AI agent computing to prioritize green power direct supply and participate in power-compute synergy scheduling: when wind/solar output is abundant and electricity prices are low, automatically scale up batch training and offline inference; when grid load is tight, automatically reduce or defer non-critical tasks.

III. Dynamic Scheduling Mechanism

Skills meeting association standards must implement a dynamic scheduling interface to receive grid load and price signals and autonomously adjust computing timing:

# Compute-power synergy scheduling interface example (Python)
from hnaia.grid import GridSignal, EnergyGrade

class SkillScheduler:
    def on_grid_signal(self, sig: GridSignal):
        # sig.price       current electricity price (CNY/kWh)
        # sig.green_ratio green power ratio 0~1
        # sig.load_level  grid load level low/mid/high
        if sig.load_level == "low" and sig.green_ratio > 0.7:
            self.scale_up(batch=64)          # abundant green power: scale up
        elif sig.load_level == "high":
            self.defer_non_critical(minutes=30)  # tight load: defer non-critical tasks

    def declare(self):
        return {"energy_grade": EnergyGrade.A, "green_preferred": True}

IV. Zero-Carbon Skill Certification

Skills running at the Hainan International Data Center / Lingshui Undersea Data Center and achieving energy efficiency grade A or B automatically receive the zero-carbon Skill certification label, displayed in the Skill market as Green Power Class A ⚡, and receive search/recommendation weighting under equal conditions.

V. A–G Energy Consumption Label System

Grade	Reference Energy per Call	Description & Incentives
Grade A	Extremely low (lightweight model / edge inference)	Eligible for zero-carbon certification, scheduling priority and recommendation weighting
Grade B	Low	Eligible for zero-carbon certification
Grade C–D	Medium	Regular scheduling, optimization encouraged
Grade E–F	Relatively high	Energy efficiency optimization plan required
Grade G	High (large-scale training class)	Scheduled only during green power low periods, separate filing required

🌿

Why Compute-Power Synergy Matters

The next bottleneck for AI is not algorithms but electricity. Turning compute demand into a schedulable flexible load can both lower operating costs for members and improve Hainan\'s clean energy consumption capacity — this is the technical anchor of the "green and low-carbon" core value.