↳ CAF WATCHER // CONTINUOUS PHYSICAL AUDIT LEDGER
Baidu (ERNIE & Kunlunxin) Grade A
China's foundational search and AI infrastructure pioneer: proprietary Kunlunxin ASICs and the Yangquan green megacampus · Category: Asian AI Frontier
Executive Physical Feasibility & Delivery Audit
Institutional Baseline: CAF Watcher evaluates stated accelerator capabilities against empirical physical reality (interconnection queues, electrical substations, liquid cooling heat transfer, and optical network fabrics). Missing public evidence is classified as unknown—it is never treated as zero capacity. Operator telemetry strengthens assessments, but participation is never mandatory.
CAF INTERVAL
[0.5 – 0.83]
DELIVERY GAP SCORE
30 pts
OPERATING POWER
0.99 GW
vs 0.45 GW announced
EVIDENCE CONFIDENCE
MEDIUM (65%)
Physical Coupling Pillars
-
Power & Grid Substrate (E): 90/100. Baidu built and owns its physical facilities, with dedicated 110kV and 220kV high-voltage substations connected to regional grid hubs in Shanxi and Hebei.
-
Thermal Dissipation & CDU Limits (T): 92/100. Yangquan megacampus achieved record PUE 1.08 using zero-water evaporative cooling during winter and high-efficiency direct liquid cold plates during summer.
-
Cluster Topology & Optical Fabric (N): 88/100. Baidu's Scorpio rack architecture standardizes power, cooling, and networking across hundreds of thousands of nodes with centralized power shelves.
-
Silicon & Advanced Packaging (O): 84/100. Kunlunxin has deployed tens of thousands of proprietary AI accelerators across Baidu search and autonomous driving (Apollo), delivering genuine silicon sovereignty.
Attributable Public Evidence Ledger
Every claim evaluated by CAF Watcher is grounded in public regulatory filings, utility interconnection disclosures, environmental permitting reports, or peer-reviewed engineering papers.