China Computing Power Conference Claims Under Scrutiny: Why the 0.04dB/km Fiber Attenuation Figure Demands Independent Verification

September 12, 2026 marks the date attached to what sources describe as the "Annual Major Breakthrough Results" from the China Computing Power Conference. Based on industry records, the conference series includes 2023 Yinchuan and 2024 Zhengzhou editions under MIIT auspices. The 2026 designation requires independent cross-verification before any technical claims can be accepted at face value. What follows is a systematic deconstruction of the disclosed components from two listed A-share companies—ZTE Corporation (000063.SZ) and YOFC (601869.SH)—within a broader announcement reportedly containing 15 breakthrough achievements, of which only these two received substantive coverage. Code is law only if the audit trail is unbroken. The same principle applies to technology claims: without third-party verification, vendor self-assertions remain hypothesis, not evidence. The disclosed results span two fundamentally different technical tiers. ZTE's OEX architecture "super node" represents a systems engineering achievement—specifically, heterogeneous chip adaptation and integration efficiency. YOFC's hollow-core fiber, if the numbers hold, constitutes a material science breakthrough. Placing these side-by-side under a unified "major breakthrough" umbrella constitutes what I classify as narrative dimension reduction: combining a quarter-year integration milestone with a multi-year materials science quest under identical promotional weight. Starting with ZTE's claims: the "3-6 month chip adaptation cycle" requires immediate deconstruction. In CUDA-centric ecosystems, mainstream frameworks (PyTorch, TensorFlow) typically achieve basic operator coverage for new GPU hardware within days to weeks. A 3-6 month window for domestic chip adaptation reveals precisely the structural gaps that remain in compiler maturity, operator library completeness, and communication library (NCCL-equivalent) robustness. This metric functions as a two-sided indicator. On one face, it signals improved deployment feasibility for domestic silicon. On the other—and this is the face the announcement obscures—it documents continued ecosystem deficit that separates "deployable" from "production-efficient." The OEX architecture's "super node" nomenclature mirrors Huawei's CloudMatrix 384 and NVIDIA's GB200 NVL72 in terminology. Product naming in the supercomputing domain carries competitive signal: following established nomenclature suggests follower positioning rather than category definition. My 2017 due diligence experience—flagging three high-profile ICO failures through blockchain explorer cross-referencing—taught me that metrics without baselines are unverifiable. The "3-6 month" figure lacks a reference point.缩短 from what? A 12-month baseline tells a different story than a 6-month to 3-month compression. Without the baseline, this metric cannot enter any comparative framework. YOFC's hollow-core fiber announcement contains two quantitative claims: 0.04dB/km attenuation at 1550nm and 91.2km continuous drawing length. The attenuation figure requires the most stringent scrutiny. Current publicly documented world records for anti-resonant hollow-core fiber (ARF/NANF) cluster between 0.09-0.17dB/km at 1550nm, reported by teams including the University of Southampton and Microsoft since 2024. The 0.04dB/km figure, if accurate, would represent a substantial leap beyond all known published results. The physical mechanism supports the direction—air-core transmission eliminates Rayleigh scattering that limits silica fiber to approximately 0.14dB/km—but a 2-4x improvement over documented records demands reproduction under standardized measurement protocols. Critical questions remain unanswered. Is 0.04dB/km measured at a single optimal wavelength or across the full C-band? Does the figure represent cutback method measurements excluding connector losses? Is this a single best-sample result or statistical median across production? Has mode leakage-induced bending loss been accounted for? The measurement methodology and certifying institution are absent from all available documentation. The 91.2km drawing length similarly lacks context. Conventional silica single-mode fiber achieves hundreds to thousands of kilometers from a single preform. Hollow-core fiber's structural fragility makes continuous drawing more challenging, but without benchmarking against other hollow-core fiber manufacturers—Corning, Sumitomo, Lumenisity (now Microsoft)—this metric floats without comparative anchor. From a commercial standpoint, ZTE occupies the "domestic substitution enabler" position: system integrator leveraging external chip supply, targeting telecom operators and state-backed compute hub projects. The business model generates predictable revenue streams but faces margin compression from lacking chip-level vertical integration. Huawei's CloudMatrix 384, by contrast, couples Ascend silicon with full-stack cluster optimization—a structurally different competitive position that captures both higher margin and stronger customer lock-in. YOFC's hollow-core fiber targets the ultra-low-latency DCI (data center interconnect) market, where air-core propagation delivers approximately 1.5x speed advantage over silica, reducing cross-datacenter round-trip latency by roughly one-third. This specification matters for distributed training synchronization, financial high-frequency trading, and cross-region inference distribution. However, the announcement omits cost structure entirely. Industry estimates place current hollow-core fiber system costs at approximately 10x conventional single-mode alternatives. At that price premium, commercial deployment remains limited to latency-sensitive, high-value segments until manufacturing scale drives cost curves downward—a timeline measured in years, not quarters. The announcement's silence on compute utilization deserves particular attention. China's intelligent computing centers have faced documented utilization challenges since 2024-2025, with multiple regional facilities reporting below-expected load factors. The industry bottleneck is not "deployment speed" but "operational efficiency." Infrastructure achievements that emphasize construction capability without addressing utilization metrics miss the structural problem. The 13 cited "commercial and pilot" projects for YOFC's fiber similarly lack separation between actual commercial deployment and experimental installations—industry convention suggests pilots often comprise 80%+ of such combined statistics. Regarding competitive positioning: ZTE ranks second-tier in domestic compute clusters, behind Huawei's vertically integrated Ascend ecosystem. YOFC presents the more internationally relevant story. Hollow-core fiber remains an undefined competitive space globally, with players including Corning, Sumitomo, and Southampton spin-offs. YOFC's scale in traditional fiber manufacturing provides process engineering advantages for production scaling—a genuine source of competitive potential if the attenuation specification survives independent verification. For investors, the thematic catalyst potential differs between the two names. ZTE's scale limits near-term revenue contribution from compute cluster activities to single-digit percentage of total revenue. YOFC's smaller market capitalization and hollow-core fiber's novelty create higher beta to sentiment swings—but also higher risk of gap-fill corrections when fundamentals fail to materialize. Any price reaction to this announcement warrants separation between event-driven momentum and verifiable business development. The 15-results framework, of which two disclosed findings originate exclusively from listed companies, warrants independent verification across all remaining claims. Until计量机构 (measurement institute) certification or third-party laboratory reproduction confirms the 0.04dB/km specification, this figure remains a vendor assertion pending audit. The hollow-core fiber space merits continued monitoring—but with emphasis on cost-down trajectory and ITU-T standard participation as the genuine indicators of long-term positioning, not announcement-stage rhetoric. What to watch next: Independent verification of YOFC's attenuation measurement through documented reproduction by accredited laboratories. ZTE's actual deployment scale and customer composition in announced compute cluster projects. Any ITU-T or IEC standard contribution filings from YOFC, which would signal genuine competitive壁垒 (barrier) construction rather than product-stage marketing. The utilization rates reported by domestic intelligent computing centers over the next 12-18 months—these figures, not conference announcements, will determine whether China's compute infrastructure buildout represents genuine capacity addition or accelerated capital deployment against uncertain demand absorption.