NVDA Chip Obsolescence and AI Bubble Risks: Accounting Practices, Capex Sustainability, and Market Sentiment

On November 22, 2025, a Reddit discussion [7] focused on NVDA chip obsolescence and its implications for the AI industry. Key concerns included AI firms extending chip depreciation periods (Meta from 4.5→5.5 years [3]) to inflate earnings, with Michael Burry estimating top 5 AI firms will understate depreciation by $176B (2026-2028) [3]. Tech giants’ $1T AI infrastructure spending faces sustainability risks due to NVDA’s annual chip cycles [2], leading Barclays to cut AI firms’ 2025 earnings by up to10% [4]. NVDA dominates the market with 88.3% of FY2025 revenue from data centers [0], but the Tech sector lagged other gainers (up0.146% [6]) reflecting investor caution. Industry developments include Burry’s Substack warning comparing NVDA to Cisco (78% post-dot-com crash [1]), Amazon reversing its 6-year depreciation policy [3], and CoreWeave’s claim of 6-year chip life vs Burry’s 2-3 years [2].

1. Accounting practices (depreciation extensions) mask capex unsustainability, with Burry’s $176B understatement claim potentially triggering regulatory scrutiny [3].
2. A disconnect exists between analyst consensus (73.4% Buy [0]) and sector underperformance, indicating divergent institutional and retail sentiment.
3. NVDA’s annual chip cycles create an ‘upgrade trap’ for customers, as noted by Microsoft’s Nadella [5], while non-frontier workload optimization (inference) offers a path to extend chip life [7].
4. Burry’s Cisco comparison highlights systemic risks if AI demand wanes, contrasting with CoreWeave’s optimistic chip life estimate [2].

• AI Bubble
  : Unsustainable free adoption models and diminishing returns could lead to a bubble burst [7].
• Accounting Transparency
  : Regulatory scrutiny over depreciation practices may result in earnings restatements [3].
• Capex Sustainability
  : Annual chip replacements turn long-term investments into recurring costs [2].
• Obsolescence
  : Rapid innovation reduces ROI for AI firms [5].

• Workload Optimization
  : Using chips for non-frontier tasks (inference) extends their useful life [7].
• Diversification
  : NVDA expanding beyond data centers (gaming:8.7% of revenue [0]) mitigates risks.
• Regulatory Clarity
  : Standardized depreciation guidelines could resolve disputes over chip life [2].

• NVDA’s data center revenue:88.3% of FY2025 total [0].
• Analyst consensus:73.4% Buy with $250 target (+39.8% upside [0]).
• Burry’s understatement estimate: $176B (2026-2028 [3]).
• Tech sector performance: up0.146% vs Healthcare (+1.73% [6]).
• Meta’s depreciation extension:4.5→5.5 years [3].
• Barclays’ earnings cut: up to10% for AI firms [4].