[GUEST ACCESS MODE: Data is scrambled or limited to provide examples. Make requests using your API key to unlock full data. Check https://lunarcrush.ai/auth for authentication information.]

[@superhoogie](/creator/twitter/superhoogie)
"Current market rental rates confirm this: A100 spot/batch: $0.50-$1.50/hr A100 on-demand: $1.30-$1.80/hr Hyperscaler rates: $3-$5/hr If A100s were truly obsolete rental rates would drop to the marginal electricity cost (400W x XXX PUE x $0.20/kWh = 10c/kwh). Instead they are being rented out at 15-25x higher rates because they generate real value"  
[X Link](https://x.com/superhoogie/status/1976756393869787370) [@superhoogie](/creator/x/superhoogie) 2025-10-10T21:07Z XXX followers, XX engagements


"Jim when evaluating the GPU market which matters more: the depreciation schedule on a GAAP income statement or the actual cash these assets generate I think the answer to that question reveals whether we're looking at a bubble or a misunderstood cash-flowing asset class. I'm open to being wrong on this what am I overlooking Full analysis here:"  
[X Link](https://x.com/superhoogie/status/1976756396512198786) [@superhoogie](/creator/x/superhoogie) 2025-10-10T21:07Z XXX followers, XX engagements


"Jim let me walk through why I respectfully disagree with the "GPU bubble" thesis. The math tells a different story than the depreciation schedules. Start with unit economics: A100 GPU: XXX kW XXX PUE = XXXX kWh/hour At $0.20/kWh: 10/hour in power At $0.50/hour rental: 40/hour gross profit The breakeven power price is $1.04/kWh. USA WAVG wholesale power prices are 4-5c/kWh"  
[X Link](https://x.com/superhoogie/status/1976756380510855347) [@superhoogie](/creator/x/superhoogie) 2025-10-10T21:07Z XXX followers, XXX engagements


"NHK Spring (TSE: 5991): AI data-center storage supplier in a global duopoly trading at XXXX forward EPS AI is no longer just compute. It is a content factory. That is the driver of an expanding nearline HDD TAM and NHK Spring sits at the choke point. Key drivers fast: X. Multimodal content explosion video audio and 3D raise retained data per model by orders of magnitude. X. Drive-level density OEM moves to Ultra-SMR and HAMR plus multi actuator designs push suspensions per drive from XX toward XX. X. Scale-out nonlinearity campus builds increase capacity far faster than unit counts creating"  
[X Link](https://x.com/superhoogie/status/1978223389148926299) [@superhoogie](/creator/x/superhoogie) 2025-10-14T22:16Z XXX followers, XXX engagements


"Beholdthe last human still using their actual brainrefusing the AI slop tsunamino ChatGPT essaysno algorithmic drivelno predictive textjust pure unfiltered occasionally wrong organic thoughtsure everything takes longerand yes they forget things Google could answer instantlybut theyre swimming upstream against the endless deluge of generated listiclessynthetic postsand those soulless XX productivity hacks articlesa digital heroor just stubborneither wayrespect the commitment to sufferingeven when their brain autocorrects nothingand blames no one"  
[X Link](https://x.com/superhoogie/status/1978999310550470913) [@superhoogie](/creator/x/superhoogie) 2025-10-17T01:39Z XXX followers, XX engagements


"A $225bn Question for NVIDIA Bulls and Bears TSMC CEO C.C. Weis response during Q&A on the 2Q25 earnings call implies a 45%+ CAGR for TSMC AI GPU production capacity through 2029 (1.454.25 = 4.9x current run rate). If we flow this through to NVIDIA: Current LTM revenue: $165bn Implied 2028 revenue: $550bn FactSet consensus for 2028 sits at $325bn. In a supply-constrained market TSMCs capacity roadmap is essentially NVIDIAs production ceiling. If TSMC builds it NVIDIA ships it. This is a supply-chain signal not an explicit forecast. But a $225bn gap between what TSMC is building toward and"  
[X Link](https://x.com/superhoogie/status/1979253254857900282) [@superhoogie](/creator/x/superhoogie) 2025-10-17T18:28Z XXX followers, XX engagements


"Nidec with XX% share of the global HDD spindle motor market is a good barometer for HDD unit builds. Their 1Q data already shows nearline volumes above XX% of total and growing guiding for higher = very supportive of expanding HDD suspensions volumes for NHK Spring"  
[X Link](https://x.com/superhoogie/status/1978820478924898466) [@superhoogie](/creator/x/superhoogie) 2025-10-16T13:49Z XXX followers, XX engagements


"1/ The AI bubble debate is missing the demand side. Everyone talks about Nvidia GPUs and hyperscaler capex. Nobody's tracking how fast token consumption is actually growing. So I looked at the data. The numbers are wild. 🧵"  
[X Link](https://x.com/superhoogie/status/1980188738870460860) [@superhoogie](/creator/x/superhoogie) 2025-10-20T08:26Z XXX followers, XX engagements


"7/ Multimodal is only XX% of prompts today - mostly images. Video hasn't scaled yet. And a 10-min video costs 177K tokens just to LOAD. Meanwhile: Context windows expanding to millions Vision models adding "perception tokens" Deep research prompts 20x higher than a year ago"  
[X Link](https://x.com/superhoogie/status/1980188756708901074) [@superhoogie](/creator/x/superhoogie) 2025-10-20T08:26Z XXX followers, XX engagements


"8/ CC Wei (TSMC) guided 45%+ of AI GPU demand is sustainable. If quarterly token additions keep jumping from 1Tn to 40Tn if agentic workflows consume 4-15x more compute if video scales. XX% might be too low"  
[X Link](https://x.com/superhoogie/status/1980188759812723125) [@superhoogie](/creator/x/superhoogie) 2025-10-20T08:26Z XXX followers, XX engagements