---
id: "claim-capability-depreciation"
type: "claim"
source_timestamps: ["§ The U.S. Strategy: Compete on Capability", "§ The Long Game"]
tags: ["market-dynamics", "rd-strategy", "commoditization"]
related: ["concept-capability-competition", "quote-today-leader-tomorrow-scrambler"]
confidence: "high"
testable: true
speakers: ["Yuanyuan Gina Cui", "Patrick van Esch", "Jan Kietzmann"]
enrichment_status: "directionally-supported; specific figures unverified"
sources: ["attention"]
sourceVaultSlug: "hbr-seg-attention"
originDay: 4
articleStem: "hbr-tier2-07-chinese-ai-firms-habits"
sourceUrl: "https://hbr.org/2026/06/lessons-from-chinese-ai-firms-on-owning-customers-habits"
sourceTitle: "Lessons from Chinese AI Firms on Owning Customers’ Habits"
---
# AI capability advantages depreciate on a six-week cycle

## Claim: Capability advantages depreciate on a six-week cycle

**Confidence: high · Testable: yes**

The authors assert that in the current AI race, **no U.S. company can hold its technical lead for long**. Every benchmark advantage is temporary, and competitors match marginal improvements within weeks.

### Evidence cited
- Rapid succession of **"Code Red" memos** between [[entity-google-d7]] and [[entity-openai-d7]] (Google's Gemini 3, Nov 2025, triggered a Code Red at OpenAI).
- Enterprise market-share volatility: [[entity-openai-d7]] fell from **~50% (2023) → 27% (2025)**; [[entity-anthropic-d7]] rose from **12% → 40%**; [[entity-google-d7]] climbed to **21%**.

**Conclusion:** An advantage depreciating on a **six-week cycle is not a strategy — it is a "holding cost."** This is the empirical backbone of [[concept-capability-competition]] and the reason the authors advocate a [[concept-habit-moat]].

> Anchoring quote: [[quote-today-leader-tomorrow-scrambler]].

**Enrichment / external validation:** The **directional** claim — that frontier advantages erode quickly and aren't durable moats — is consistent with observed leapfrogging and competitive product launches. However, the **specific "six-week cycle"** and the **exact market-share percentages** are *not corroborated by public research*; treat them as **author extrapolations / proprietary estimates**, i.e., an interpretive heuristic rather than validated fact. Some analysts also note that **distribution/platform control** (default status in Office, Android, iOS) can sustain advantages longer than benchmark parity alone.