0xrelogic

Analyzed Q3 2025 earnings data from UBS and found China's AI story is actually two separate cycles with 12-18 month lag between them.

• Mainland hardware suppliers (optical transceivers, servers): +34-57% earnings growth NOW

• Innolight hit $3.3B revenue (+122% YoY) from datacenter buildout

• Hong Kong internet platforms (Alibaba Cloud, Tencent): still in capex phase, monetization projected 2H26

• Alibaba e-commerce AI already break-even with +12% ROAS

The gap between these cycles matters for positioning. Hardware is paying off today, cloud is a 2026 call option.

Happy to discuss the optical transceiver data or enterprise adoption timelines.

I wrote this after Jensen Huang's Nov 7 statement that Nvidia now has "zero share" of China's AI datacenter market, down from 95%. The geopolitical split is creating three distinct compute ecosystems. - China revenue impact: Dropped from 21.4% of Nvidia's total (FY2023) to 13.1% (FY2025), with ~$15B in annual datacenter revenue now offline

- Huawei Ascend 910C performance: Delivers roughly 60% of H100 throughput for inference workloads, but training large-scale models remains challenging. The bigger gap is software CUDA's 15-year ecosystem vs Huawei's CANN framework

- Google TPU positioning: While Nvidia Blackwell has 12-18 month lead times and China is locked out, Google's TPU becomes more attractive for specific workloads (Transformer training, inference at scale) especially for GCP customers

- China's 30% mandate: State-funded datacenter projects under 30% completion must remove all foreign AI chips or lose funding (Reuters reporting from this week)

The article traces the export control timeline from Oct 2022 initial rules through Oct 2023 tightening, with all revenue figures verified against Nvidia's public financials and multiple sources.

Three parallel tracks are forming

1. Nvidia Blackwell (premium, supply-constrained, no China)

2. Huawei Ascend (domestic alternative, 60% H100 performance, ecosystem gap)

3. Google TPU (available hedge for workload-specific optimization)

Happy to discuss the geopolitics, technical performance comparisons, or supply chain implications. Also interested if anyone has hands-on experience with Ascend 910C vs H100 workloads.

Sources include Reuters, BIS.gov export control documents, Nvidia's FY2025 financials, and verified tech specs from multiple outlets.

Last week I wrote about Meta's $75B AI infrastructure spending. Seven days later: $30B bond offering (largest since 2023), stock dropped 11% (worst day in 3 years), and Wall Street analysts who were bullish 48 hours earlier all downgraded.

What changed? Nothing fundamental. Meta beat earnings on every metric. The only new info: 2026 capex will be "significantly larger" than 2025's $70-72B.

Key findings: - Oppenheimer downgraded from $696 target to no target, citing "Metaverse spending" concerns (the exact bear case they'd been dismissing) - 5 major firms cut price targets 3.6-10% within 24 hours - Fed Chair Powell had to address "circular financing" concerns - CoreWeave's $9B acquisition of Core Scientific collapsed same day

The irony: My Oct 28 article predicted bubble concerns, off-balance-sheet debt becoming a theme, and ROI timeline issues. Bloomberg published "Meta Starting Trend for Off-Balance Sheet Debt" on Oct 31.

Wall Street doesn't predict the future. It narrates the present.

Happy to discuss the analyst flip-flops, the $21.6B in hidden debt (Blue Owl SPV), or whether Zuckerberg's "better to overbuild" strategy is genius or gambling.

I tested Next.js 16 on a 200+ component project. The Turbopack numbers are real:

Development startup: 1083ms (Webpack) → 603ms (Turbopack) = 44% faster

Production builds: 45s (Webpack) → 17s (Turbopack) = 2.6x faster

This isn't a micro-optimization. It's a noticeable difference in daily workflow.

The article also covers: • Astro 5.15 deployment skew protection (solves version mismatch bugs)

• Node 22 LTS native proxy support (useful for enterprise environments)

• CSS View Transitions now Baseline (all major browsers)

• Vite+ announcement (unified JavaScript toolchain, preview early 2026)

October was packed with releases. Happy to discuss any of these updates.