NVDABullunverifiable
Meet Kevin“you could simultaneously think that hardware or stocks can do well over the next 6 to 12 months and also think that it's all going to come crashing down into a glorious bubble pop uh you know in 2 to 5 years”
Thesis at the time
MixedKevin breaks down a bullish AI-compute thesis piece and pushes back on several of its assumptions, arguing that token costs are rapidly depreciating and that open-weight models from firms like Alibaba will compress margins for frontier labs and hardware demand. He says he remains bullish on AI hardware and stocks in the near term but believes the buildout could end in a bubble pop within a few years.
Key arguments
- Token costs are a rapidly depreciating asset, undermining the 'two S-curve' bull argument for exponential compute demand
- Open-weight/Chinese models (e.g., Alibaba's Qwen) are up to 100x cheaper than frontier lab tokens, which will compress margins
- Software, not infrastructure or tokens, will capture most of the long-term value because it maintains high gross margins
- The bull piece's comparison to Qualcomm vs MediaTek is flawed because token models can be swapped instantly while phone chips cannot
- Assumptions of 40% CAGR revenue growth justifying 100 gigawatts of compute by 2030 are overly optimistic
Counter-arguments acknowledged
- The bull case argues AI usage involves two multiplying exponential curves (more users x more tokens per user), unlike the flat-fee internet buildout
- The piece argues Anthropic's revenue slowdown was due to compute constraints and anti-distillation crackdowns, not fundamental demand issues
- The report contends closed-source frontier labs can maintain a moat because open-source models are harder to optimize
Hedges and caveats (from the video)
- Host acknowledges understanding the bull case argument for exponential growth via dual S-curves
- Analysis is exploratory and questions-focused rather than definitive
- Host notes this is a summary and commentary on a third-party analysis, not original research
- Membership and course promotions included in video suggest potential conflicts of interest
The call
- Date said
- Aug 28, 2026
- Timeframe
- over the next 6 to 12 months
- Confidence
- medium
- Specificity
- vague
How it resolved
- Status
- unverifiable
Why this resolved this way(resolution audit)
Full rules: docs/resolution-spec.md.
Confidence Reasoning
Direct statement of near-term bullishness but softened by 'could simultaneously think' framing rather than a firm commitment; no specific price target given.
Source
Is the AI Trade DEAD? The FULL AI Stock Thesis Break Down.
Said on Aug 28, 2026Open on YouTube ↗