# ![@DratchCap Avatar](https://lunarcrush.com/gi/w:26/cr:twitter::230912559.png) @DratchCap Matt Dratch

Matt Dratch posts on X about ai, $nvda, in the, $nbis the most. They currently have [-----] followers and [---] posts still getting attention that total [-----] engagements in the last [--] hours.

### Engagements: [-----] [#](/creator/twitter::230912559/interactions)
![Engagements Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::230912559/c:line/m:interactions.svg)

- [--] Week [------] -90%
- [--] Month [-------] +111%
- [--] Months [---------] +507%
- [--] Year [---------] +4,039%

### Mentions: [--] [#](/creator/twitter::230912559/posts_active)
![Mentions Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::230912559/c:line/m:posts_active.svg)

- [--] Week [--] -48%
- [--] Month [--] +46%
- [--] Months [---] +431%
- [--] Year [---] +482%

### Followers: [-----] [#](/creator/twitter::230912559/followers)
![Followers Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::230912559/c:line/m:followers.svg)

- [--] Week [-----] +0.83%
- [--] Month [-----] +15%
- [--] Months [-----] +645%
- [--] Year [-----] +1,089%

### CreatorRank: [-------] [#](/creator/twitter::230912559/influencer_rank)
![CreatorRank Line Chart](https://lunarcrush.com/gi/w:600/cr:twitter::230912559/c:line/m:influencer_rank.svg)

### Social Influence

**Social category influence**
[stocks](/list/stocks)  43.31% [technology brands](/list/technology-brands)  42.52% [finance](/list/finance)  25.2% [countries](/list/countries)  7.87% [currencies](/list/currencies)  3.15% [automotive brands](/list/automotive-brands)  1.57% [social networks](/list/social-networks)  1.57% [travel destinations](/list/travel-destinations)  0.79% [nba](/list/nba)  0.79% [fashion brands](/list/fashion-brands)  0.79%

**Social topic influence**
[ai](/topic/ai) 26.77%, [$nvda](/topic/$nvda) #2262, [in the](/topic/in-the) 9.45%, [$nbis](/topic/$nbis) 7.87%, [open ai](/topic/open-ai) 7.09%, [market](/topic/market) 7.09%, [nvda](/topic/nvda) #526, [$crwv](/topic/$crwv) 5.51%, [inference](/topic/inference) 5.51%, [investment](/topic/investment) 4.72%

**Top accounts mentioned or mentioned by**
[@sama](/creator/undefined) [@ser_kingslayer_](/creator/undefined) [@tshirtnjeans2](/creator/undefined) [@chamath](/creator/undefined) [@hedgeyecomm](/creator/undefined) [@gavinsbaker](/creator/undefined) [@mcuban](/creator/undefined) [@stevehou](/creator/undefined) [@openai](/creator/undefined) [@realjimchanos](/creator/undefined) [@benitoz](/creator/undefined) [@monkeymktplace](/creator/undefined) [@midnight_captl](/creator/undefined) [@elonmusk](/creator/undefined) [@signulll](/creator/undefined) [@evrgn11112231](/creator/undefined) [@benioff](/creator/undefined) [@unitreerobotics](/creator/undefined) [@hypertechinvest](/creator/undefined) [@amirpc](/creator/undefined)

**Top assets mentioned**
[NVIDIA Corp. (NVDA)](/topic/$nvda) [Nebius Group N.V. Class A Ordinary Shares (NBIS)](/topic/$nbis) [CoreWeave, Inc (CRWV)](/topic/$crwv) [Advanced Micro Devices (AMD)](/topic/$amd) [Hut [--] Mining Corp. (HUT)](/topic/$hut) [Alphabet Inc Class A (GOOGL)](/topic/$googl) [Microsoft Corp. (MSFT)](/topic/$msft) [Oracle Corporation (ORCL)](/topic/$orcl) [Metadium (META)](/topic/$meta) [Apple, Inc. (AAPL)](/topic/$aapl) [Amazon.com, Inc. (AMZN)](/topic/$amzn) [Iris Energy Limited (IREN)](/topic/$iren) [Adobe, Inc. (ADBE)](/topic/$adbe) [Tesla, Inc. (TSLA)](/topic/$tsla)
### Top Social Posts
Top posts by engagements in the last [--] hours

"Skepticism compounding faster than Ai is a feature. Be grateful. When GPT-5 was released some folks claimed AI progress was hitting a wall whereas others said progress would continue. GPT-5.2 was released [--] months ago. GPT-5.3-Codex was released [--] days ago and is twice as token efficient for coding. It's clear who turned out to be correct. https://t.co/hMV2TbE8Go When GPT-5 was released some folks claimed AI progress was hitting a wall whereas others said progress would continue. GPT-5.2 was released [--] months ago. GPT-5.3-Codex was released [--] days ago and is twice as token efficient for"  
[X Link](https://x.com/DratchCap/status/2020614951027745202)  2026-02-08T21:45Z [----] followers, [----] engagements


"I see a lot of discussion about how AI gives meh investing output this weekend. Its a basic bitch if you will. Duh. The quality of your questions also determines the quality of your output. Double duh. AI is NOT an imagination engine. Its trained on the distribution of whats already been said and thought. Imagining a future that isnt in the data yet and having feel (investment EQ) for what others will think and do remains inherently human. This is also why AI will ironically widen outcome gaps while it equalizes intelligence. Some will use it to compound faster. Others will outsource thinking"  
[X Link](https://x.com/DratchCap/status/2020570200492364022)  2026-02-08T18:47Z [----] followers, [----] engagements


"Interesting that when @sama made equivalently bold predictions people had such a different reaction. Narratives are powerful indeed. Dario predicts trillions in AI revenue likely before [----] and country of geniuses by [----]. I think no one can estimate what this really means. https://t.co/9Phuw78uIP Dario predicts trillions in AI revenue likely before [----] and country of geniuses by [----]. I think no one can estimate what this really means. https://t.co/9Phuw78uIP"  
[X Link](https://x.com/DratchCap/status/2022704203571818770)  2026-02-14T16:07Z [----] followers, [----] engagements


"Natural human content will be like organic food 5yrs. (H/t my good friend Slavka 😉) *TIKTOK'S CHINESE PARENT HAS AN APP TO REPLACE HOLLYWOOD: WSJ *TIKTOK'S PARENT HAS AI MODEL TO PRODUCE VIDEO W/ STORYLINE: WSJ *TIKTOK'S CHINESE PARENT HAS AN APP TO REPLACE HOLLYWOOD: WSJ *TIKTOK'S PARENT HAS AI MODEL TO PRODUCE VIDEO W/ STORYLINE: WSJ"  
[X Link](https://x.com/DratchCap/status/2022739160247996429)  2026-02-14T18:26Z [----] followers, [---] engagements


"Sometimes the Simulation makes things too obvious"  
[X Link](https://x.com/DratchCap/status/2023131044611015052)  2026-02-15T20:23Z [----] followers, [----] engagements


"I think the consensus is wrong on GPU depreciation. People are modeling these assets like a self-storage facility run by some jabroni in Jersey (static boxes with no ecosystem or terminal value beyond the life of the first lease). But thats not the case. If this is the first time youve seen some of the below points you should deeply consider your sources. 1) The residual value debate is really a macro call on AI demand. - If you assume Ai slows token growth lags hardware gains and supply demand then no sh*t Sherlock residual pricing on old GPUs will suck and short lived asset lives make"  
[X Link](https://x.com/DratchCap/status/1989201665455542736)  2025-11-14T05:20Z [----] followers, 44.1K engagements


"This type of garbage is getting tiring. WSJs headline implies Nvidia is bailing on OpenAI but the article itself says the original $100B / 10GW framework was nonbinding and never a finalized contract (old news Jensen didnt like the $amd deal either - duh). It also says the latest talks still include a large Nvidia equity investment in OpenAIs current funding round. This is click-first framing that fuels FUD and gets eagerly amplified by haters (especially of @sama) and shorts. The actual news buried in paragraph [--] seriously NVDA is still putting tens of billions into the round. Headlines like"  
[X Link](https://x.com/DratchCap/status/2017392886359138696)  2026-01-31T00:22Z [----] followers, 30.2K engagements


"Keep Away is nigh. $hut (and all its compute / power friends) I dont think people appreciate how close we are to Hunger Games: AI Labs Edition for everything compute related. Marginal ROI on compute is increasing not decreasing. If one or two incremental use-cases take off the way Claude Code/Clawdbot did the near-to-medium term I dont think people appreciate how close we are to Hunger Games: AI Labs Edition for everything compute related. Marginal ROI on compute is increasing not decreasing. If one or two incremental use-cases take off the way Claude Code/Clawdbot did the near-to-medium term"  
[X Link](https://x.com/DratchCap/status/2018402714682597529)  2026-02-02T19:14Z [----] followers, 12.5K engagements


"Ahead of potentially a mega raise by @sama I present this chart without comment. $nvda. $orcl $crwv"  
[X Link](https://x.com/DratchCap/status/2018460331698516196)  2026-02-02T23:03Z [----] followers, 60.4K engagements


"The New Circularity: Ai is compressing software terminal values. But to displace software you need enormous compute memory and power. The lost value *should* accrue to the physical stack. And the market isnt pricing in nearly enough of that migration. The problem is that we remain in a shoot first regime. Few participants have first principles for how AI works. DeepSeek in 1Q25 was the template: incomplete understanding turns every headline into a FUD crisis. A super BOOM with micro busts if you will. This dynamic may be here to stay since nobodys sure exactly whos swimming naked. As my old"  
[X Link](https://x.com/DratchCap/status/2018715965710590450)  2026-02-03T15:59Z [----] followers, 18.8K engagements


"False binary. Software wont vanish. The risk is disintermediation: agents become the UI stitch workflows across vendors and make apps swappable backends (switching costs fall). That shifts pricing to usage/outcomes pressures margins/NRR (increased competition) and drives multiple compression. And voila welcome to value stocks land"  
[X Link](https://x.com/DratchCap/status/2019045023539396717)  2026-02-04T13:47Z [----] followers, 56K engagements


"We havent seen a blowout capex number like this since insert any quarter $googl $hut"  
[X Link](https://x.com/DratchCap/status/2019158145839337630)  2026-02-04T21:16Z [----] followers, [----] engagements


"All this questioning of AIROIC on a pre-tax basis ignores some pretty big tax code changes. With OBBB-style immediate expensing you pull the tax shield forward. At a 7% discount rate a $100 deduction is worth $100 now vs $82 over [--] years vs $33 over [--] years i.e the DC shell benefit is massive. For a real cash taxpayer at 21% that can make a $1B build feel like $790M after-tax (and far worse under normal 40-year depreciation). This is a material accelerator for the AI boom and the associated capex. Tax something less and you get more of it duh. :) $nvda $googl $msft $meta $amzn"  
[X Link](https://x.com/DratchCap/status/2020180574674018487)  2026-02-07T16:59Z [----] followers, 15K engagements


"@signulll Its hard to switch though = weve got an agent for that 👀👀"  
[X Link](https://x.com/DratchCap/status/2020266238543573248)  2026-02-07T22:39Z [----] followers, [---] engagements


"I think a few things are going on w/ $NVDA (and its mega cap brethren) 1) Lag [--] are all sources of funds 2) burdened by its own success / size 3) AI bulls can get more juice / asymmetry elsewhere given [--] + [--] The other narratives are so conflicting in context of the broader market I find them less compelling. Punish the spenders on ROI yet AI disruption is decimating entire sectors 🤔. I suspect theres a reflexive loop of quant vigilantism (press Ai losers) HF pain and narrative backfill. A little time and maybe some pre NVDA eps FOMO probably helps."  
[X Link](https://x.com/DratchCap/status/2022486314709295490)  2026-02-14T01:41Z [----] followers, [---] engagements


"We havent seen this kind of timing since @mcuban collared his stock https://www.wsj.com/business/deals/onlyfans-in-talks-to-sell-majority-stake-to-investment-firm-61f9971fgaa_at=eafs&gaa_n=AWEtsqfI2cj0cZGpc3m8jQ6IH9VLxm97CeJUeXxfq5hQzcRAsMA2-JZmJBXZanSIzCo%3D&gaa_ts=6990c3ab&gaa_sig=FN304vaKGfYqx2gEpceIFPVPiQlU9vh-rZOzwAWe7NRpXsF4CP_fY3xkTVZL7T1yEc7xqHtx9n1OvCI3D2KrjQ%3D%3D In the spirit of narrative consistency I have to ask is this good or bad for OnlyFans 🤔"  
[X Link](https://x.com/DratchCap/status/2022741560933167159)  2026-02-14T18:35Z [----] followers, [---] engagements


"@evrgn11112231 @GavinSBaker Forgot to mention we are still punishing the spenders on ROI questions at the same time 🤯"  
[X Link](https://x.com/DratchCap/status/2023257440796463216)  2026-02-16T04:45Z [----] followers, [---] engagements


"AIs Shadow Output Gap While Washington obsesses over debt and inflation AI is already ushering in an age of abundance (Part 1) The political and economic establishment cant stop talking about deficits debt and the CPI. Capitol Hill hearings FOMC minutes and financial news all pulse to the same beat. Yet this fixation ironically coincides with the arrival of the most powerful productivity engine in human history: generative AI. Its impact is creating a shadow output gap an invisible but rapidly widening expansion of supply-side capacity. Policymakers especially at the Federal Reserve act as if"  
[X Link](https://x.com/DratchCap/status/1942372440727867746)  2025-07-07T23:57Z [----] followers, 74.9K engagements


"Tower of Babel vibes running hot on my feed this weekend"  
[X Link](https://x.com/DratchCap/status/2020637286405492752)  2026-02-08T23:14Z [----] followers, [----] engagements


"Narrative violation: noticing co-pilot for outlook is *much* improved. Can summarize entire folders remember your portfolio / watchlist & search for relevant info under constraints (summarize anything on my names from MS and xyz analysts). Helpful for those of us trapped inside a sandbox. $msft https://twitter.com/i/web/status/2020648766471012732 https://twitter.com/i/web/status/2020648766471012732"  
[X Link](https://x.com/DratchCap/status/2020648766471012732)  2026-02-08T23:59Z [----] followers, [----] engagements


"Market today https://www.youtube.com/watchv=DlB0xi65q2k https://www.youtube.com/watchv=DlB0xi65q2k"  
[X Link](https://x.com/DratchCap/status/2021615620287086929)  2026-02-11T16:01Z [----] followers, [---] engagements


"I think I preferred debating ROI your moat is AIs opportunity. Just sayin. $NVDA"  
[X Link](https://x.com/DratchCap/status/2021630222614114406)  2026-02-11T16:59Z [----] followers, [---] engagements


"I have found the next great AI stock. It hits all the key themes even fear of the unknown $TLT"  
[X Link](https://x.com/DratchCap/status/2021967748880851451)  2026-02-12T15:20Z [----] followers, [----] engagements


"Welcome to the Known Unknown Regime. Rolling recessions are now rolling corrections $nvda"  
[X Link](https://x.com/DratchCap/status/2021980543357964335)  2026-02-12T16:11Z [----] followers, [---] engagements


"You can Lag [--] all [--] now. What a time to be alive. $nvda $amzn $msft $googl $tsla $meta $aapl"  
[X Link](https://x.com/DratchCap/status/2022329262947418172)  2026-02-13T15:17Z [----] followers, [----] engagements


"Mr Regulation is also Mr Responsible how convenient. And of course actions words always: TPUs = GPUs. Panic bidding power. Upsizing capital raise. Starting beauty pageant for IPO. 🤔🤔 Dwarkesh asks Dario a fantastic question relating to how he is so bullish on AGI yet so conservative on data center build out - Dario has an amazing take on this: Dario Amodei details the staggering financial risk of the AI race explaining that if growth continues at 10x a https://t.co/Ed99X88zwS Dwarkesh asks Dario a fantastic question relating to how he is so bullish on AGI yet so conservative on data center"  
[X Link](https://x.com/DratchCap/status/2022468954611208285)  2026-02-14T00:32Z [----] followers, [----] engagements


"Compute = table stakes. I spent last night with Andrew Strominger and Alex Lupsasca two of the top physicists in the world They just released a paper co-authored with OpenAi that seems to me like ASI Andrew who helped develop string theory told me that a year ago his view was that he didnt know https://t.co/wwKPAskcOd I spent last night with Andrew Strominger and Alex Lupsasca two of the top physicists in the world They just released a paper co-authored with OpenAi that seems to me like ASI Andrew who helped develop string theory told me that a year ago his view was that he didnt know"  
[X Link](https://x.com/DratchCap/status/2022469456446808361)  2026-02-14T00:34Z [----] followers, [----] engagements


"AI regulation is climate [---]. New establishment policy will run entirely through this prism. And China is the new attack on democracy imo. Its a frame that tries to make opposition radioactive for folks who disagree. If preventing China from accessing our chips and winning Ai (whats that mean btw) is so existential shouldnt the govt be doing way more at home to advance the cause Yet one mention of a govt backstop and people had a melt down. Why is Dario Amodei so weirdly hawkish on the Chinese https://t.co/DvXZzNhoxK Why is Dario Amodei so weirdly hawkish on the Chinese https://t.co/DvXZzNhoxK"  
[X Link](https://x.com/DratchCap/status/2022492406768185483)  2026-02-14T02:05Z [----] followers, [----] engagements


"Havent seen a $TLT pitch this good in [--] days. your gentle reminder there are like zero economists or ppl in general who know how to reason about what happens when near zero cost human level intelligence gets woven into the fabric of the economy at scale this fast. this scenario has never remotely been in the possibility your gentle reminder there are like zero economists or ppl in general who know how to reason about what happens when near zero cost human level intelligence gets woven into the fabric of the economy at scale this fast. this scenario has never remotely been in the possibility"  
[X Link](https://x.com/DratchCap/status/2022672857810243680)  2026-02-14T14:02Z [----] followers, [----] engagements


"I can teach to the test better than you Bytedance is putting pressure on the American Frontier Labs. China is gaining ground every day. And DeepSeek v4 hasn't even been released yet. Bytedance hasn't just recently revolutionized text-to-video; they're also catching up in regular LLMs. https://t.co/xOmAIOFuGu Bytedance is putting pressure on the American Frontier Labs. China is gaining ground every day. And DeepSeek v4 hasn't even been released yet. Bytedance hasn't just recently revolutionized text-to-video; they're also catching up in regular LLMs. https://t.co/xOmAIOFuGu"  
[X Link](https://x.com/DratchCap/status/2022739464867471363)  2026-02-14T18:27Z [----] followers, [---] engagements


"In the spirit of narrative consistency I have to ask is this good or bad for OnlyFans 🤔 Looks great Looks great"  
[X Link](https://x.com/DratchCap/status/2022741090399392035)  2026-02-14T18:33Z [----] followers, [----] engagements


"We havent seen this kind of timing since @mcuban collared his stock In the spirit of narrative consistency I have to ask is this good or bad for OnlyFans 🤔 In the spirit of narrative consistency I have to ask is this good or bad for OnlyFans 🤔"  
[X Link](https://x.com/DratchCap/status/2022743361455820974)  2026-02-14T18:42Z [----] followers, [----] engagements


"@stevehou Hmm. GOOGL could do $225m EBITDA in [--] leaving it $30-35bn in FCF and they could cut capex in 1/2 and repay all debt in around [--] quarters just for context. Not really scary to have 1/2 turn of leverage during investment cycle. 🤷🏻♂"  
[X Link](https://x.com/DratchCap/status/2023108871414583639)  2026-02-15T18:55Z [----] followers, [---] engagements


"Ive written about a shadow output gap over the last few months AI deflating the price of intelligence in ways the data doesnt capture yet. I still think thats real. But Ive been stuck on that framing and all I see around me is doomerism about job loss. Recently I forced myself to think about the demand side of the equation and the irony is hard to miss. Its 1yr since DeepSeek and people still dont really understand Jevons paradox. The market has learned layer [--] (tokens). It hasnt fully learned layer [--] (thinking) where cheaper cognition makes 10x more projects experiments and entire categories"  
[X Link](https://x.com/DratchCap/status/2023528663438827863)  2026-02-16T22:43Z [----] followers, [----] engagements


"The reflexivity coefficient between markets and the real economy is approaching 1.0market swings directly impact real-world behavior faster than ever (consequence of household equity exposure / high information flow) Housekeeping data point: A friend rolling up accounting firms shared this weekend Were seeing broad-based caution from SMB clients uniformly across our accounting business. CEOs feeling cautious heading into Q1 earnings have every incentive to temper expectations especially with stocks already down. But AI could be a notable exception. Crucially big AI players report later"  
[X Link](https://x.com/anyuser/status/1901285718330093658)  2025-03-16T14:53Z [----] followers, [----] engagements


"$NVDA. Why the sleepy response post GTC Some thoughts. 1) GTC historically is a boondoggle for nerds who get excited by new acronyms that most people will never (or need to) understand 2) Jensen is burdened by his own meteoric success -- heavy wears the crown of impossible expectations 3) He is asking people to imagine an exponential future while most investors think linearly and are naturally skeptical because of #2 (this must be I think his focus on economics and token math is an attempt to help start exponential thinking. What I hear: - reasoning requires a lot more tokens (more compute);"  
[X Link](https://x.com/anyuser/status/1902495232320811012)  2025-03-19T22:59Z [----] followers, [---] engagements


"@Benioff @UnitreeRobotics That robot costs 50-100k per the website"  
[X Link](https://x.com/DratchCap/status/1903492672343478626)  2025-03-22T17:03Z [----] followers, [----] engagements


"@HyperTechInvest LOL"  
[X Link](https://x.com/DratchCap/status/1903621512550138211)  2025-03-23T01:35Z [----] followers, [----] engagements


"The current administration needs a broader definition of trade. Its not just goodsits also financial assets. In that sense the U.S. runs a massive trade surplus: We import goods they buy our bonds and stockslowering our cost of capital and reinforcing USD as the global reserve. That system powered the U.S. economic machine: It let companies like Amazon burn cash for a decade reach escape velocity and change the worldwith low rates and deep capital markets. Ironically the admin doesnt acknowledge the trickle-down value of that Even while pushing deregulation capex incentives and corporate tax"  
[X Link](https://x.com/DratchCap/status/1906922636363362534)  2025-04-01T04:12Z [----] followers, [---] engagements


"Almost every AI signal lately has been bullish but the markets latched onto one cautious datapoint from MSFT (even though GS raised MSFT [----] capex recently after meeting with them). Flip the scenarioMSFT bullish everyone else soft would the market be excited Of course not. This is a Rorschach test. People want to sell so theyre cherry-picking reasons. Belief will return when the macro vol fades. $NVDA"  
[X Link](https://x.com/anyuser/status/1907242701398876251)  2025-04-02T01:24Z [----] followers, [----] engagements


"@chamath @amirpc Theres multiple teams at Wedbush with multiple analysts and they likely speak to 100s of analysts management teams and expert calls every week given the number of people. That would be a research shop doing its job. Lol Dead"  
[X Link](https://x.com/anyuser/status/1908643262531014703)  2025-04-05T22:09Z [----] followers, [----] engagements


"@AravSrinivas TikTok for finance. 30-60 second clips with research analysts"  
[X Link](https://x.com/DratchCap/status/1910870583153131562)  2025-04-12T01:40Z [----] followers, [----] engagements


"@zerohedge *explanation not exclamation"  
[X Link](https://x.com/DratchCap/status/1912567578926460996)  2025-04-16T18:03Z [----] followers, [----] engagements


"Random thought: OpenAIs first mover advantage + new unlimited memory = contextual lock-in. Just like Microsoft shipped the PC OS six months earlier and rode that head start into a multi-trillion-dollar monopoly OpenAI is quietly building the most valuable memory graph in history: your data your preferences your interactionsforever. $nvda @OpenAI"  
[X Link](https://x.com/anyuser/status/1913594351541784714)  2025-04-19T14:03Z [----] followers, [---] engagements


"$GOOGL I wonder if we should ditch the ad market valuation (search is screwed) and value it as the only big tech company that owns end-user context the model and real custom silicon (Jensen will tell you he respects TPU but not the others). That combination is likely to increase in value as the AI Flywheel spins faster Also YouTube is an index of cultural memory: [---] hrs/min uploaded the only proprietary video RLHF corpus at global scale. Competitors LLMs may have to license that data eventually for true multi-modal capabilities. Oh it also trades at a notable discount to S&P and its bad biz"  
[X Link](https://x.com/anyuser/status/1915926045120774271)  2025-04-26T00:28Z [----] followers, [---] engagements


"I can see it now: DeepSeek R2 crushes compute needs Huge breakthrough The Great @DanielTNiles linked a presentation by another guy named Niles which claimed to analyze all the latest DeepSeek papers to preview their amazing breakthroughs. $nvda @OpenAI @sama Great headlinesback to reality oh there goes gravity 👇 [--]. Sparse-expert tricks arent so special Only 1-in-20 experts fire per token yet RAM & network pipes stay busy so real-world speed gains are way lower [--]. Training is still pricey Cheaper forward passes dont spare the trillion-token data consumption or months of reward-tuning runs."  
[X Link](https://x.com/anyuser/status/1920971370910884224)  2025-05-09T22:37Z [----] followers, 57.4K engagements


"You are confusing fair value thinking with cost-basis GAAP. You have apparently never heard that you should study accounting before commenting on it :). $CRWV doesnt take a mark to market / impairment because secondary market chips trade cheaper to carrying value. Under GAAP GPUs sit in PP&E at historical cost less accumulated depreciation in this case straight line depreciation. The accounting firm will only require a write-down if an impairment test shows the entire cash generating asset cant earn back its carrying value (revenue has to comfortably cover book value). Also the impairment"  
[X Link](https://x.com/anyuser/status/1924095473889010125)  2025-05-18T13:31Z [----] followers, [---] engagements


"All of the hot takes on the coming US power shortage in response to the Senate bill solar / wind subsidy changes are far too alarmist. An alternative view: - We are short GPUs today NOT power. Power in the US is a merchant market when demand shows up prices spike new capacity comes on. You can see this in PJM mkt dynamics. When the reserve cushion got thin last year (7x prior price) [--] GW of quick start capacity instantly moved to construction. A greenfield CCGT (combined cycle gas) can be built in [--] months and there are *DOZENS* of 100-MW-plus brownfield sites SITTING EMPTY today waiting for"  
[X Link](https://x.com/anyuser/status/1939416908887110122)  2025-06-29T20:13Z [----] followers, [---] engagements


"I view the housing as the last covid remnant. Existing supply at 40yr lows because of the great refi at zero in 2020-2021. This created fiscal dominance for the first time ever in the USA. Housing prices UP into high rates diluted the impact of the monetary channel on aggregate demand. Interestingly this also means cutting rates may provide less stimulus than in past (housing may hold steady or decline rather than go up)"  
[X Link](https://x.com/anyuser/status/1949098094735368310)  2025-07-26T13:22Z [----] followers, [----] engagements


"@HedgeyeComm @FishtownCap The legacy biz is growing slower than peers bc its 25% larger AI is growing triple digits for Amzn and still below 10%. If you can show their Ai business is growing slower than others great"  
[X Link](https://x.com/anyuser/status/1951136065823134016)  2025-08-01T04:21Z [----] followers, [---] engagements


"Theres a lot of discussion about the Chinese energy advantage (3x the power grid) so I think its time to set the record straight. The TL/DR is that *Token per grid MWh* is the only measurement that matters and the Chinese system is an *ORDER OF MAGNITUDE* behind despite their total size. Our advantage is also compounding (faster chip innovation from a higher base) making it difficult to overcome. $NVDA (Note some people will debate about the exactness of numbers like anything its imperfect but the framework and direction are right). A quick thread:"  
[X Link](https://x.com/anyuser/status/1952571403737120817)  2025-08-05T03:24Z [----] followers, [---] engagements


"SemiAnalysis note post mkt highlighting nvda cgx. My summary: $NVDA out with a new leap forward in compute in the last 24hrs called $NVDA CGX. Its a monster leap for inference efficiency and will be available as part of Rubin (after Blackwell). Its 67x improvement and unlocks extremely long context windows by splitting inference into [--] parts pre-fill and decode. In current architectures the model has to re-read the book so to speak at multiple points in an interaction slowing things down including throughput. In this case one accelerator indexes the book (pre-fill) and then all decode can"  
[X Link](https://x.com/anyuser/status/1965886883021390234)  2025-09-10T21:15Z [----] followers, [---] engagements


"1/ The $NVDA / OpenAI deal tells me were closer to the beginning of the AI boom than the end. And todays *meager* stock reactions across the ecosystem suggests market skepticism or ignorance. Either way its opportunity. (1/5)"  
[X Link](https://x.com/anyuser/status/1970331055895818403)  2025-09-23T03:35Z [----] followers, [---] engagements


"4/To all the Private Credit guys whining about having too much GPU exposure Jensen just debunked your biggest fear. Hes providing credit support by putting in $100Bn equity capital BEHIND YOU. Perhaps actually you have *too little*"  
[X Link](https://x.com/DratchCap/status/1970331484298059990)  2025-09-23T03:36Z [----] followers, [---] engagements


"Everyone is upset about the $NVDA ponzi / circular deals Id love to hear their thoughts on this other big one called fractional banking. Supposedly that one is *really* something"  
[X Link](https://x.com/anyuser/status/1970536106530865459)  2025-09-23T17:09Z [----] followers, [---] engagements


"China wants $NVDA chips; they just want the real thing not the hand-me-downs. Theyre saying they dont to save face in the event the US doesnt cave. In modern Seinfeld they would likely call this stealing the breakup (I dumped you first). 4/4"  
[X Link](https://x.com/anyuser/status/1970685085981528412)  2025-09-24T03:01Z [----] followers, [----] engagements


"If the Fed were public itd be a multi-trillion-dollar entity controlling liquidity (M2).$NVDA plays the same role in AI central bank of compute.It issues the base currency (GPUs) and does QE (indirectly buys its own stuff) so the ecosystem can expand and multiply. The same core rule also applies: dont fight the Fed"  
[X Link](https://x.com/DratchCap/status/1971941633169281105)  2025-09-27T14:14Z [----] followers, 16.5K engagements


"Deepminds new paper on Chain of Frames (reasoning for video models)= dawn of the *multi-modal token ramp*. Veo3 is basically Gpt3 and as we move to CoF in future models compute demand will explode. This is part Jensens compounding exponential. Every extra second of video = 10s of thousands of tokens adding reasoning (multiple passes) involves another 2-4x. Think about robotics training. A 60s clip at [--] frames per second & 256k tokens / frame = 460k tokens (60x a text prompt). With [--] reasoning passes were at 920k tokens. One synthetic robot training session = hundreds of such clips i.e."  
[X Link](https://x.com/anyuser/status/1972040994041258024)  2025-09-27T20:49Z [----] followers, [---] engagements


"Imagine China buying $NVDA GPUs made by $INTC in the USA. Beijing secures $NVDA supply Trump gets a trade win and $INTC gets the demand boost it needs to scale AI chips and help de-risk US reliance on Taiwan. Everybody wins :) 🤔"  
[X Link](https://x.com/anyuser/status/1972735545785847951)  2025-09-29T18:49Z [----] followers, [---] engagements


"Ironically if the productivity boom from Ai plays out at scale we will all have much more time to create and watch AI slop. A self fulfilling loop #skynet $nvda"  
[X Link](https://x.com/anyuser/status/1973518747207082220)  2025-10-01T22:41Z [----] followers, [---] engagements


"Even if every existing GPU on earth magically became a Blackwell GB300 tomorrow wed still be 10-30x short on compute for *consumer inference alone* at saturation. The other TAM verticals are large (er) and compound the problem. This is why the big guys are spending. A quick overview and a few things to note: The genesis of this thought experiment is the recent compute demand surge data points / Jensens *compounding exponentials* from BG2. The goal is to imagine the art of the possible consumer inference TAM + the necessary compute to get a sense for what AI inning were in. 1) this is a SWAG"  
[X Link](https://x.com/anyuser/status/1975002656037806473)  2025-10-06T00:58Z [----] followers, [----] engagements


"For the wheres the ROI Dinosaurs Doomers and Top callers ZOOM OUT: Tokens are the work units the dividend is TIME. AI turns clicks actions. Assistants buy / book / fix / teach so humans dont. That creates a *shadow consumer surplus* (time saved money not spent) that P&Ls wont show at first (solows paradox is back). Rails get built; yield shows up later as inference-driven revenue. Monetization is things like merchant-funded conversion lift affiliate/booking fees payments/interchange action ads per-action fees.and perhaps most importantly opex deflation when agents (robots) do $50/hr work for"  
[X Link](https://x.com/anyuser/status/1975263183150473521)  2025-10-06T18:13Z [----] followers, [---] engagements


"It is *highly* likely Jensen blessed AND *likes* the OpenAI / $AMD deal. It is in $NVDAs interest for the AI flywheel to *spin faster* rather than have OAI slowdown to wait for Jensens *limited* cowos supply. Let the LO pre $AMD analyst day chase commence"  
[X Link](https://x.com/anyuser/status/1975302894611275798)  2025-10-06T20:51Z [----] followers, [---] engagements


"@FundaBottom @theinformation They dont get clicks for that. BUT they did us a favor anyone who panicked / believed it showed us they are swimming naked and dont understand a key part of the AI ecosystem at all. Keep the receipts. $nvda $amd $nbis"  
[X Link](https://x.com/anyuser/status/1975727059692626041)  2025-10-08T00:56Z [----] followers, [---] engagements


"Perhaps actually you have too little exposure as h100 pricing goes UP. 4/To all the Private Credit guys whining about having too much GPU exposure Jensen just debunked your biggest fear. Hes providing credit support by putting in $100Bn equity capital BEHIND YOU. Perhaps actually you have *too little* 4/To all the Private Credit guys whining about having too much GPU exposure Jensen just debunked your biggest fear. Hes providing credit support by putting in $100Bn equity capital BEHIND YOU. Perhaps actually you have *too little*"  
[X Link](https://x.com/DratchCap/status/1976070902636011800)  2025-10-08T23:43Z [----] followers, [---] engagements


"Macro musings #1 Tough to be a dollar bear (even with cuts) when NVDA GPUs (and stock) are priced in USD (note: its a metaphor) #Computedollars $nvda $nbis $crwv $amd"  
[X Link](https://x.com/anyuser/status/1976446526282137968)  2025-10-10T00:35Z [----] followers, [---] engagements


"Musing #4: a disciplined and simple follow Jensen policy since [----] approximated the investment outcome of a multi-year AI deep dive $nvda"  
[X Link](https://x.com/anyuser/status/1976476688298193340)  2025-10-10T02:35Z [----] followers, [---] engagements


"The best hedge is often to make money somewhere else. Perhaps (not so) rare earth stocks are interesting then. If theres no art in the deal the US gov goes full throttle to build domestic capacity (operation warp speed part 2); if theres a deal Ai stocks are back on the train to #ripcity and youre happy to lose on the rocks. You also own the option rare earths work either way and you win on both 🫣 $USAR (lot of daylight between its market cap and peers 🤷🏻♂)"  
[X Link](https://x.com/anyuser/status/1976814023397032000)  2025-10-11T00:56Z [----] followers, [---] engagements


"BTW racks optics power distribution cooling all have their own cycles with multi year lead times. They have to upgrade as well to handle new hardware. So no you cant just compress the hardware cycle to any time frame you want *at scale*. People are still getting Blackwell and figuring out how to make it work 1yr after initial delays. Also software improvements are not discussed and they can materially change the inference efficiency for old chips over time. Better to do own work than just read seeking alpha pieces. I just brought up multiple things on the fly that are critical to the analysis"  
[X Link](https://x.com/anyuser/status/1976844759369998609)  2025-10-11T02:58Z [----] followers, [----] engagements


"@chamath Behind the meter to a DC campus near you"  
[X Link](https://x.com/anyuser/status/1976866042069889089)  2025-10-11T04:22Z [----] followers, [---] engagements


"@DanielSLoeb1 20x levered 24/7 $NVDA 🤞"  
[X Link](https://x.com/anyuser/status/1977031083305951720)  2025-10-11T15:18Z [----] followers, [---] engagements


"@firstadopter When the bubble stories stop is the tell its a bubble. $nvda $nbis"  
[X Link](https://x.com/anyuser/status/1977412564410929653)  2025-10-12T16:34Z [----] followers, [---] engagements


"@WarrenPies I heard the assets of our time GPUs & QQQ are priced in USD. Debase that $nbis $nvda"  
[X Link](https://x.com/anyuser/status/1978240495848485052)  2025-10-14T23:24Z [----] followers, [---] engagements


"Any specialist in the sector knows you cant compress the hardware cycle fast enough which the most important counter and you did not mention. So while Jensen can put out new stuff every year we cant make the rest of the infrastructure fast enough for 1-2yr full replacement cycles. Jensen will tell you this himself. @PikerCapital @OutsiderInfo76 @RihardJarc BTW racks optics power distribution cooling all have their own cycles with multi year lead times. They have to upgrade as well to handle new hardware. So no you cant just compress the hardware cycle to any time frame you want *at scale*."  
[X Link](https://x.com/anyuser/status/1978495247123366330)  2025-10-15T16:16Z [----] followers, [----] engagements


"Most Neoclouds that rent by the hour also batch / time slice / partition inference workloads which means the $/hr x hrs /day x utilization is the wrong formula. $2/hr headline rate could for example yield effective $510/hr realized per GPU depending on throughput & time slicing. ROI is as much about scheduling as it is sticker price. Many seem to not know how the cloud businesses actually work but EVERYONE thinks they know the depreciation schedule. 🥱. $NVDA $NBIS $CRWV @GavinSBaker"  
[X Link](https://x.com/DratchCap/status/1980108486571704560)  2025-10-20T03:07Z [----] followers, [----] engagements


"@GavinSBaker @FT They also miss that most GPU renters time-slice and partition I.e. sell the same GPU hr multiple times over for inference. Effective yield of a GPU can be higher than sticker price x hours x utilization ☺ $nbis"  
[X Link](https://x.com/DratchCap/status/1980109537072877824)  2025-10-20T03:11Z [----] followers, [----] engagements


"Weekend musing: Anchor AI with a Govt floor bid build the interstate let the market race on top. Perhaps we can call it Sovereign Awareness 😉 *Imagine* for example the U.S. takes a stake in OpenAI with profits capped like defense. This replaces (fake) circularity concerns instantly with a sovereign spine. The result is a lower ecosystem WACC which leads to more behind the meter power faster interconnects turbines and fabs that actually get built (producers dont worry about a demand rug pull later). By de-risking the ecosystem private capital is *crowded in* rather than crowded out. AI-winter"  
[X Link](https://x.com/anyuser/status/1982573194247553481)  2025-10-26T22:20Z [----] followers, [---] engagements


"The biggest and most important company in history just told everyone their revenue estimates are 40% low and the stock is only up 5% Oh. My. God. $NVDA"  
[X Link](https://x.com/anyuser/status/1983341473061093559)  2025-10-29T01:13Z [----] followers, [---] engagements


"@kashyap286 @potawang Macro expert who not surprisingly has never modeled a company. SG&A doesnt scale even close to linearly for a powered shell or cloud. This is a concept called operating leverage :) Also I love that both of you very clearly used ChatGPT and cant do your own analysis"  
[X Link](https://x.com/anyuser/status/1985410906415681880)  2025-11-03T18:17Z [----] followers, [---] engagements


"@mikealfred $NBIS owns the NJ (MSFT) datacenter outright as well as Finland. And they are buying land for new sites actively. CRWV also has as [---] GW of capacity open. Hmmm"  
[X Link](https://x.com/anyuser/status/1985420199433306511)  2025-11-03T18:53Z [----] followers, [----] engagements


"$Msft had existing relationship with Nscale at $2.70/hr and then onboarded a new relationship that doesnt have proven execution at $2.90/hr $iren. AND with a big prepayment. = GPU / cloud market tightening. $nbis $crwv"  
[X Link](https://x.com/anyuser/status/1985893819812839445)  2025-11-05T02:15Z [----] followers, [---] engagements


"Lets get the party started already. $nvda $nbis $crwv"  
[X Link](https://x.com/DratchCap/status/1986197314646552941)  2025-11-05T22:21Z [----] followers, 12.7K engagements


"Asking for a lower WACC is about speed. Cheap capital accelerates platform build-outs (duh). We did this in the last decade it was called ZIRP 🤣😉. And it helped finance a segment known today as mag7 while they burned cash. Seems to have worked out well (see: modern technology ecosystem) $nvda $nbis $amd $iren"  
[X Link](https://x.com/anyuser/status/1986212136511574192)  2025-11-05T23:20Z [----] followers, [---] engagements


"Bullish Ai datapoints are now so bullish they are bearish which is of course bullish. Circularity indeed $nvda $nbis $amd $hut"  
[X Link](https://x.com/anyuser/status/1986220140388630965)  2025-11-05T23:52Z [----] followers, [---] engagements


"USA has to choose: we cant say we are not sending our best chips to China bc of national security and then do nothing to lubricate the capital machine to build out what is clearly a nationally security priority. No one would give a damn if spaceX ipod but it would not exist without material financing from the US govt via NASA"  
[X Link](https://x.com/anyuser/status/1986230501158625767)  2025-11-06T00:33Z [----] followers, [---] engagements


"@iamgingertrash He doesnt say for them specifically and hes talking about the ecosystem. For example the govt might backstop debt against turbine facilities to lower cost for companies unwilling to take a risk on building (long cycle) capacity in the USA"  
[X Link](https://x.com/DratchCap/status/1986789376928751695)  2025-11-07T13:34Z [----] followers, [----] engagements


"There is a false narrative that AI will shrink the knowledge gap across the population (everyone has ubiquitous access). But humans are the same across time and space. It is more likely AI widens the gap as power users run fast ahead and while other rely on the AI lazily. This is also likely to mean we experience more Financial Market Influence Operations where a nefarious actor (state group etc) perpetuates market / sector FUD online. Case in point the OpenAI CFO bailout commentary. It is clear from watching the actual video below that the essence of her point was NOT that OAI needs a"  
[X Link](https://x.com/anyuser/status/1986794029703377038)  2025-11-07T13:53Z [----] followers, [----] engagements


"There is more discussion about a new Chinese model called Kimi K2 that is benchmarking like frontier models. Some thoughts: 1) who cares The models are goodharting the tests and the low cost models are distilled off much larger industry capex 2) what matters is distribution inference latency real context windows and the resulting feedback loop of recursive improvement. 3) you (or enterprises) sign up for Gemini / GPT bc of the *full stack*: compute offering reliability networking tool use AND integrated ecosystem upgrades. 5) cost per token MUST be put into relative context. At what latency"  
[X Link](https://x.com/DratchCap/status/1987175966376906992)  2025-11-08T15:10Z [----] followers, [---] engagements


"$crwv customers scrambling to re-lease old chips [--] quarters early at 95% original price. Also in an ironic twist Claude can now update @RealJimChanos depreciation estimates for him automatically 😂"  
[X Link](https://x.com/anyuser/status/1988044753791852769)  2025-11-11T00:43Z [----] followers, [----] engagements


"Will the @RealJimChanos please stand up #depreciation :)"  
[X Link](https://x.com/DratchCap/status/1988245370187469309)  2025-11-11T14:00Z [----] followers, [---] engagements


"No not correct If energized MW are the scare asset then you dont rip out old GPUs for free you only do it if: Incremental rev from the new chip (lost rev during downtime + retrofit capex + financing cost). That threshold wont always clear especially for inference / mid-tier work where old GPUs are good enough. AND. If you truly believe theres nowhere to plug them in then anyone who already has energized MW with GPUs in racks is sitting on a call option on future demand and will have pricing power. Most likely theyll run a tiered fleet (new chips to premium tokens old to cheap inference). That"  
[X Link](https://x.com/anyuser/status/1989456042938478814)  2025-11-14T22:10Z [----] followers, [---] engagements


"$ORCL 5yr CDS: A *theory* on how a boring working-capital plumbing trade can reflexively morph into a credit signal. **Receivable puts** enter stage left. 1.Stargate is so enormous / cap-intensive / long-cycle that suppliers are using receivable puts to monetize AR and avoid being liquidity-hostages to Oracles payment terms. If anything this lesser-known form of factoring is a *scale tell* (NOT repayment fear) and is logical / expected. 2.Those flows can nudge CDS wider and **the widening becomes the story**. Desks pitch / sell it to big AI-long equity holders as an asymmetric synthetic put"  
[X Link](https://x.com/DratchCap/status/1989726363264180389)  2025-11-15T16:05Z [----] followers, [----] engagements


"Everyone assumes @sama cant fund a $1T buildout. Im not so sure Id bet against his likely side hustle: hyper-personal AI girlfriend/boyfriend experiences incl emotional support late-night chats (video too) and custom for your eyes only content with your own avatar. 🫣 The infinite monetization loop A repeatable scalable highly addictive and potentially countercyclical bet on total human depravity 😂 (historically hard to bet against). You just laughed but deep down you just got a little less uncertain. Sorry but someone had to say it. #darkbuttrue"  
[X Link](https://x.com/DratchCap/status/1989876254892331481)  2025-11-16T02:00Z [----] followers, [---] engagements


"I think this confuses catalyst with cause and would go much simpler. Fed made Dec live and people have a lot of profits to protect. Ai happens to be where the profits are so the @sama narrative (and other things) are the backfill for explaining the risk mgmt everyone already wanted to do. And I think too many people ignore the *IF* qualifier from @sama and collapse the ecosystem buildout he is lobbying for with OAIs contractual obligations (not the same). The spend is a *plan* for *if* they can exectute. Price action = Narrative Ignoring Sams communication style the token demand trajectory"  
[X Link](https://x.com/DratchCap/status/1989890367987109999)  2025-11-16T02:56Z [----] followers, [----] engagements


"This is the wrong way to think about Ai monetization. That analyst needs to think a little deeper. The real TAM is a take-rate on actions and opex savings across the entire economy. Capex justification comes from creating a platform that allows you to tax the ecosystem vs own the end customers. So the question is what is the ultimate value to the activity that runs through AI stacks and what % can they get. Being bigger / earlier is big advantage in that race. Consumers or subs dont need to pay GPT per month for it to work they could for example pay other services that pay GPT (directly take"  
[X Link](https://x.com/anyuser/status/1989918528493715896)  2025-11-16T04:48Z [----] followers, [----] engagements


"Recently spent time updating an older analysis on where AI demand is actually going and came away still thinking were massively short compute (850x short) on consumer inference alone. Big range (future is humbling) but even the low end makes the point. I dropped a link to the fuller write-up for anyone inclined over a slow week. It also hits a few popular debates / my steelman AI bear case. Some of this may be optimistic (or wrong). Im a dreamer so be kind :) Consumer is easiest to parameterize. If were massively short just on that you start to understand why the biggest players are building"  
[X Link](https://x.com/DratchCap/status/2003517655555440669)  2025-12-23T17:26Z [----] followers, 135.5K engagements


"IF is doing all the work here. The report does not substantiate its claims (see AI check below) but it does piggyback on a surprisingly profitable strategy of publishing short reports on $app during the quiet period. Now the 5th firm to do so. Thankfully an even more profitable strategy is BTFD https://twitter.com/i/web/status/2013428733982413016 https://twitter.com/i/web/status/2013428733982413016"  
[X Link](https://x.com/DratchCap/status/2013428733982413016)  2026-01-20T01:49Z [----] followers, [----] engagements


"I think this is misleading. Much of this is due to exchange rate valuation as USD weakened against others (we are cutting rates duh) and not active selling by CBs. USD also new % high for global swift transactions. When Anthropics IPO is priced in a different currency we can talk about the decline of the USD. https://twitter.com/i/web/status/2015064076796191110 https://twitter.com/i/web/status/2015064076796191110"  
[X Link](https://x.com/DratchCap/status/2015064076796191110)  2026-01-24T14:08Z [----] followers, [---] engagements


"$Nike **TO BUY [--] MAC MINIS #CLAWD NIKE $NKE TO CUT [---] EMPLOYEES AS IT ACCELERATES AUTOMATION AT U.S. DISTRIBUTION CENTERS - CNBC https://t.co/AnpmhHAuSO NIKE $NKE TO CUT [---] EMPLOYEES AS IT ACCELERATES AUTOMATION AT U.S. DISTRIBUTION CENTERS - CNBC https://t.co/AnpmhHAuSO"  
[X Link](https://x.com/DratchCap/status/2015890225994559616)  2026-01-26T20:50Z [----] followers, [----] engagements


"@DarioAmodei Weve gone from Tower of Babel to Tower of GPUs faster than I thought. Penthouse reserved for the Self-Anointed"  
[X Link](https://x.com/DratchCap/status/2015972334683717686)  2026-01-27T02:17Z [----] followers, [---] engagements


"NVDAs $2B $crwv investment looks a bit like a bear hug (10% is a signal). $CRWV is perhaps the most strategic AI infrastructure asset on the planet now. NVDAs de facto cloud off-balance-sheet so Jensen can cosplay Switzerland with the hyperscalers. The risk is in the tails: GPU residuals post-contract + what can we build this into. But when your partner is the largest company on earth and controls compute innovation and distribution those tails look like very juicy call options. #Dreamweave I think the consensus is wrong on GPU depreciation. People are modeling these assets like a"  
[X Link](https://x.com/DratchCap/status/2016017643610796112)  2026-01-27T05:17Z [----] followers, [----] engagements


"Powell created the mess leather jacket man cleaned it up $nvda Stanley Druckenmiller said what many think about Jerome Powell but wont say out loud. The Feds job is to avoid big mistakes like the massive inflation we just had. Powell is obsessed with engineering a soft landing and protecting his legacy. But the only reason we even need https://t.co/UfCsbhZvx0 Stanley Druckenmiller said what many think about Jerome Powell but wont say out loud. The Feds job is to avoid big mistakes like the massive inflation we just had. Powell is obsessed with engineering a soft landing and protecting his"  
[X Link](https://x.com/DratchCap/status/2016661230761443641)  2026-01-28T23:54Z [----] followers, [----] engagements


"Is the boogie man (capex opex) scary if hes already out of the closet That 1Q $meta rev guide was 👀. A lot of folks forgot whose got the best eyeball slop and will be chasing higher"  
[X Link](https://x.com/DratchCap/status/2016676782544470452)  2026-01-29T00:56Z [----] followers, [----] engagements


"HF Bro: Why was Azure weak despite huge capacity ramp Because they're allocating capacity to apps. So why isn't apps stronger Because non-AI apps are weak. Software winter is here.. Oh. 🫣🫣 $MSFT (h/t Guggenheim)"  
[X Link](https://x.com/DratchCap/status/2016681397881110837)  2026-01-29T01:14Z [----] followers, 25.7K engagements


"The obvious conclusion from this apparent food fight to throw money at OAI is that $ORCL is really just synthetic Mag [--] risk 🫢. Circular when it suits me 😉😂 P.s. wonder what @sama showed them"  
[X Link](https://x.com/DratchCap/status/2017050863047491888)  2026-01-30T01:42Z [----] followers, [----] engagements


"The new MCP Apps standard makes me think the LLMs are getting closer to being the OS. Easyware sorry software becomes a pluggable utility for the model. A screen / artifact within a screen if you will. In other words the GUI is the context window. Users will also be extracted from the application layer entirely in many cases (you dont open $uber to get a car and you may not care which car the model sends you). Fast times. https://twitter.com/i/web/status/2017057108668109224 https://twitter.com/i/web/status/2017057108668109224"  
[X Link](https://x.com/DratchCap/status/2017057108668109224)  2026-01-30T02:07Z [----] followers, [---] engagements


"@HedgeyeComm @compound248 @SamofAmerica One positive for $META is that people love slop and wont want their agents to watch it for them. How many apps can say the same 👀"  
[X Link](https://x.com/DratchCap/status/2017057648370147338)  2026-01-30T02:09Z [----] followers, [---] engagements


"Humanity is forever and always depraved theyll find their way to the gram But to your point they could need different coaxing (glasses for example). And I agree there is a shortage of imagination out there around the implications of AI on the OS. Jarvis from Ironman is a good starting point https://twitter.com/i/web/status/2017068416926982455 https://twitter.com/i/web/status/2017068416926982455"  
[X Link](https://x.com/DratchCap/status/2017068416926982455)  2026-01-30T02:52Z [----] followers, [--] engagements


"But kevin warsh used to be hawkish once upon a time I prefer to ZOOM OUT. so just to recap this week (so far) - musk industries is real (spacex tesla xai merger) - clawdbot explosion leading to a bankrun on mac minis but then anthropic released their own version - tesla dropped the bomb theyre halting production on model s and x to scale 1M so just to recap this week (so far) - musk industries is real (spacex tesla xai merger) - clawdbot explosion leading to a bankrun on mac minis but then anthropic released their own version - tesla dropped the bomb theyre halting production on model s and x"  
[X Link](https://x.com/DratchCap/status/2017355133307195534)  2026-01-30T21:51Z [----] followers, [----] engagements


"Strikingly similar to tech eps revisions Fun facts: 1yr Ago consensus [--] $nvda & $googl est were $6 / [-----]. Today they are $10 / $13.50 (low imo btw). Translation the street is short humility and imagination. (H/t @sarahdingwang for awesome charts). @eriktorenberg"  
[X Link](https://x.com/DratchCap/status/2017377485382472063)  2026-01-30T23:20Z [----] followers, [----] engagements


"@chamath And $NVDA was down 17% on DumbSeek Monday so. $u @MBromberg"  
[X Link](https://x.com/DratchCap/status/2017379754144370956)  2026-01-30T23:29Z [----] followers, [---] engagements


"@karpathy @moltbook @openclaw This is just mass self play. If you tell your LLM to talk with itself for 500hrs and let the conversation stray to any random thought youll get weird outcomes let alone 5k of them interacting. This is not surprising"  
[X Link](https://x.com/DratchCap/status/2017395280706932774)  2026-01-31T00:31Z [----] followers, 60.4K engagements


"The article is slop Jensen / Collette walked back the memo of understanding / non binding sept 100bn investment a long time ago. If people believed $NVDA was putting in $100bn in back in the fall why all the concern about their credit risk sorry i mean $ORCL credit risk (See what i did there ;) ) . This is WSJ looking for a headline to get people to think NVDA is pulling out of the current funding round (countless people sending that to me) despite saying they are putting in size [--] paragraphs in. Absolutely silly. https://twitter.com/i/web/status/2017427558170509497"  
[X Link](https://x.com/DratchCap/status/2017427558170509497)  2026-01-31T02:39Z [----] followers, [----] engagements


"How about another narrative violation: the new GPT [---] agent inside excel seems to work as well as Claude inside excel. Have them both build a simple $AAPL model using 10k and 10q and ask them to ask you [--] to [--] clarification questions before starting. See for yourself and compare vs just absorbing someone elses opinion. (Theyre both awesome). https://twitter.com/i/web/status/2017428585259085977 https://twitter.com/i/web/status/2017428585259085977"  
[X Link](https://x.com/DratchCap/status/2017428585259085977)  2026-01-31T02:43Z [----] followers, [----] engagements


"@RMDiLillo Always a question of prompting / context / guidance. Have built me excellent SOTP / sensitivity on $HUT valuation. Neither was complex (i.e. not [--] drivers [--] tabs) but material time save great formatting"  
[X Link](https://x.com/DratchCap/status/2017435763684643207)  2026-01-31T03:12Z [----] followers, [---] engagements


"@jaketapper These appear to be unsubstantiated reports with the middle column describing them as such - no phone number person deemed mentally unstable. Criticizing potus is good part of the job but this is just slop. Got me all excited for nothing. Lame"  
[X Link](https://x.com/DratchCap/status/2017445544533049578)  2026-01-31T03:51Z [----] followers, [----] engagements


"@anandragn Index Geniuses: above the fray anualizing 15% YTD. Just straight chillin"  
[X Link](https://x.com/DratchCap/status/2017611623909798350)  2026-01-31T14:51Z [----] followers, [---] engagements


"Entropy works in nature because every atom and organism is distinct (vs just distinct instructions I.e. prompts). In this public simulation every agent shares the exact same 'DNA' (the model weights) making the output more like a hall of mirrors than a society. Of course if you hand the agent a credit card and say 'do random stuff' you have a problem. That is just automated recklessness though. https://twitter.com/i/web/status/2017725662912557348 https://twitter.com/i/web/status/2017725662912557348"  
[X Link](https://x.com/DratchCap/status/2017725662912557348)  2026-01-31T22:24Z [----] followers, [--] engagements


"@lokoyacap @JaredKubin People tell me we have the biggest context windows theyve ever seen. $sndk $nvda"  
[X Link](https://x.com/DratchCap/status/2017744892676497913)  2026-01-31T23:40Z [----] followers, [---] engagements


"Weekend hot take: GPT [---] writing clarity is improving in real time (7 days) on par with Claude Opus. This also coincides with @sama saying on a podcast last week they purposely sacrificed this capability to make the model smarter. Perhaps they are adjusting / tinkering. Or Im crazy not sure :) https://twitter.com/i/web/status/2018105807028572297 https://twitter.com/i/web/status/2018105807028572297"  
[X Link](https://x.com/DratchCap/status/2018105807028572297)  2026-02-01T23:34Z [----] followers, [----] engagements


"Bullish Going down for fake news is much less scary than going down for no reason. $nvda Reuters reports OpenAI has grown dissatisfied with how quickly Nvidias $NVDA latest chips can deliver answers for more complex ChatGPT queries and has been looking at alternative hardware in some cases since last year. https://t.co/TPlfkb20q6 Reuters reports OpenAI has grown dissatisfied with how quickly Nvidias $NVDA latest chips can deliver answers for more complex ChatGPT queries and has been looking at alternative hardware in some cases since last year. https://t.co/TPlfkb20q6"  
[X Link](https://x.com/DratchCap/status/2018439479019585656)  2026-02-02T21:40Z [----] followers, [----] engagements


"@HedgeyeComm @S_curvecap Some of us who follow you have severe FOMO which makes it cruel to suggest blow out numbers 24hrs pre print. Trimming out HODL in. 🤞"  
[X Link](https://x.com/DratchCap/status/2018922299450343528)  2026-02-04T05:39Z [----] followers, [--] engagements


"Its starting. Odds others bend the knee to $GOOGL are roughly zero. And in an arms race the spoils go to suppliers (token production chain). Next up $AMZN (GPT Prime) Keep Away is nigh. $hut (and all its compute / power friends) Keep Away is nigh. $hut (and all its compute / power friends)"  
[X Link](https://x.com/DratchCap/status/2019179885193785631)  2026-02-04T22:42Z [----] followers, [----] engagements


"The Revolution devours its children. GPT-5.3-Codex is our first model that was instrumental in creating itself OpenAI stated in a blog post accompanying the models release. The Codex team used early versions to debug its own training manage its own deployment and diagnose test results and evaluations our team was blown away by how much Codex was able to accelerate its own development. https://twitter.com/i/web/status/2019475298517967354 https://twitter.com/i/web/status/2019475298517967354"  
[X Link](https://x.com/DratchCap/status/2019475298517967354)  2026-02-05T18:16Z [----] followers, [----] engagements


"@HedgieMarkets Maybe Im off but I think he is describing agents as orchestrators / new UI where they still invoke deterministic systems to do deterministic work (hence tool calling of approved functions or data retrieval from systems of record). $MSFT"  
[X Link](https://x.com/DratchCap/status/2020157289559212363)  2026-02-07T15:26Z [----] followers, [---] engagements


"@rev_cap @pekwat Fair on the ROIC point but I was really highlighting how the pulled-forward tax shield slashes the effective cash cost of the build not that it magically makes every project look high-ROIC. Different angle on the same tax change :)"  
[X Link](https://x.com/DratchCap/status/2020189314349691008)  2026-02-07T17:34Z [----] followers, [--] engagements


"@DevinAkin @jimcramer so if gov't made all of [---] private for wifi use and auctioned it for single bidder there's no buyer B.s. $GSAT"  
[X Link](https://x.com/anyuser/status/534458019757948928)  2014-11-17T21:28Z [----] followers, [--] engagements

Limited data mode. Full metrics available with subscription: lunarcrush.com/pricing