{
    "action": "final",
    "content": "# AI Supercomputers Dataset: Column Reference\n\nThis dataset contains **61 columns** across **482 rows** tracking AI supercomputer systems worldwide. Here's a comprehensive breakdown:\n\n## Identification & Status\n- **Name** (VARCHAR) — Unique identifier for each supercomputer system\n- **Status** (VARCHAR) — Current operational status (all entries show \"Existing\")\n- **Certainty** (VARCHAR) — Confidence level in the data (\"Confirmed\" or \"Likely\")\n- **Single cluster?** (BOOLEAN) — Whether the system is a single unified cluster (all TRUE)\n\n## Hardware Specifications\n- **Chip type (primary)** (VARCHAR) — Main GPU/accelerator model (36 distinct types)\n- **Chip quantity (primary)** (DOUBLE) — Number of primary chips (64 to 200,000)\n- **Chip type (secondary)** (VARCHAR) — Secondary chip type if applicable (11 types)\n- **Chip quantity (secondary)** (BIGINT) — Count of secondary chips (1 to 50,000; 464 nulls)\n- **Total number of AI chips** (DOUBLE) — Combined chip count (64 to 230,000)\n- **GPU Supplier (primary)** (VARCHAR) — Manufacturer of primary GPUs (11 suppliers: AMD, NVIDIA, etc.)\n- **GPU supplier (secondary)** (VARCHAR) — Secondary GPU manufacturer\n- **Hardware note** (VARCHAR) — Additional hardware details (38 distinct notes)\n\n## Performance Metrics\n- **Max OP/s** (measure) — Maximum operations per second\n- **Max OP/s (log)** (DOUBLE) — Log-scale performance (15.3 to 20.7)\n- **H100 equivalents** (DOUBLE) — Performance normalized to NVIDIA H100 units (1 to 275,796)\n- **16-bit OP/s** (measure) — 16-bit floating point operations per second\n- **16-bit OP/s (log)** (measure) — Log-scale 16-bit performance\n- **32-bit OP/s** (measure) — 32-bit floating point operations per second\n- **8-bit OP/s** (measure) — 8-bit integer operations per second\n- **Energy Efficiency** (measure) — Performance per watt\n- **Energy Efficiency (log)** (DOUBLE) — Log-scale efficiency (10.08 to 12.46)\n\n## Power & Cost\n- **Power Capacity (MW)** (DOUBLE) — Total power capacity in megawatts (0.11 to 352.35 MW)\n- **Reported Power Capacity (MW)** (measure) — Officially reported power capacity (402 nulls)\n- **Calculated Power Capacity (MW)** (measure) — Estimated power capacity (65 nulls)\n- **Hardware Cost** (DOUBLE) — Total hardware expenditure ($3.5M to $7.1B)\n- **Calculated Cost** (measure) — Estimated total cost (98 nulls)\n- **Reported Cost** (measure) — Officially reported cost (454 nulls)\n- **Reported Cost (Inflation adjusted)** (measure) — Cost adjusted for inflation (454 nulls)\n- **Cost Quote** (measure) — Source quote for cost information\n\n## Location & Ownership\n- **Country** (VARCHAR) — Nation where system is located (44 countries)\n- **Location** (VARCHAR) — Specific address or facility (158 distinct locations)\n- **Owner** (VARCHAR) — Organization that owns the system (122 distinct owners)\n- **Users** (VARCHAR) — Organizations using the system (118 distinct user groups)\n- **Sector** (VARCHAR) — Industry sector (4 categories: Public/Private, etc.)\n- **latitude** (spatial) — Geographic latitude coordinate\n- **longitude** (spatial) — Geographic longitude coordinate\n\n## Temporal Information\n- **First Operational Date** (DATE) — When system became operational (2010-11-15 to 2025-07-22)\n- **First Operational Date Note** (VARCHAR) — Additional date context (242 distinct notes)\n- **Decommissioned Date (if applicable)** (temporal) — When system was retired\n\n## Relationships & History\n- **Builds Upon** (VARCHAR) — Previous system this one extends (25 distinct predecessors)\n- **Superseded by** (VARCHAR) — Newer system that replaced this one (31 distinct successors)\n- **Possible Duplicate** (VARCHAR) — Flag for potential duplicate entries\n- **Possible Duplicate Of** (VARCHAR) — Reference to possible duplicate\n\n## Rankings & Context\n- **Rank when first operational** (metric) — Position in global rankings at launch (37 nulls)\n- **% of largest cluster when first operational** (metric) — Size relative to largest system (37 nulls)\n- **Largest existing cluster when first operational** (VARCHAR) — Name of largest system at the time\n\n## Data Quality & Analysis\n- **Exclude** (categorical) — Flag to exclude from analysis\n- **Include in Standard Analysis** (categorical) — Flag for standard analysis inclusion\n- **Noteworthy** (categorical) — Special significance flag (473 nulls)\n- **Certainty Note** (VARCHAR) — Details on data confidence (45 distinct notes)\n\n## Documentation\n- **Note** (VARCHAR) — General notes and details (76 distinct entries)\n- **Quote** (VARCHAR) — Source quotations (226 distinct quotes)\n- **Source 1, 2, 3, 4** (VARCHAR) — Reference sources for the data\n\n## Key Insights\n- **Temporal coverage**: 2010–2025 (15 years of AI supercomputer evolution)\n- **Geographic spread**: 44 countries represented\n- **Scale range**: Systems from 64 chips to 230,000 chips\n- **Cost range**: $3.5M to $7.1B in hardware costs\n- **Power range**: 0.11 MW to 352 MW capacity\n- **Data quality**: Most core fields (Name, Country, Owner, Chip type) have zero nulls; cost and power metrics have more missing values (98–454 nulls depending on field)"
}