M Mohan (@mukund) header
M Mohan (@mukund) profile picture

Long term investor - https://xitter.gq/t.co/9CsbGgAktB Day job: PE fund @IncVangal with investments in @yirlaagency, @peridotdev @fundamentell & 18 other startups

Joined March 2007
My threads, data, tools I use, investment philosophy and frequently asked questions. Please review these links. 1. Overview docs.google.com/document/d/e… 2. Investing Process docs.google.com/document/d/e… 3. Tools I use and like docs.google.com/document/d/e…
84
226
10
1,141
Once you @grok you wont go back to @AnthropicAI or @OpenAI The model matters, but not that much
3
3
1,597
Are we seeing exhaustion yet? Feeling the blue wave
1
3
2,354
18 IPO lockups end between now and Jan 11. Quantinuum, Liftoff, INNIO, Bending Spoons, Csquare... The day insiders can sell isn't when you call. By then the plan is set. Advisors: how far ahead of a lockup do you reach out?
1
1,587
Turns out they revoked harness in less than a week. $HUBS 😂
Only technical people will even understand what a “Customer Relationship Harness” is. Apparently that is the product. More jargon. $HUBS
1
4
3,216
Loops. I finally figured it out. The success is in loops. I am slow but the AI bros taught me. Loops.
9
38
10,007
$C Citi changed its terms so customers are responsible for what third-party AI tools do with their accounts, including after hacks more banks are pushing AI-agent liability onto users
3
3
2,675
From July
Play this out a bit. @OpenAI and @AnthropicAI know that training frontier models is not where money is going to be. So they build their own inference cloud. Or just acquire $CRWV or $NBIS in a reverse merger and now they are public. Then it becomes all-in one AI model plus cloud vs another. Inference is where margins will be. Not as good as software or ads but cloud margins.
1
4
4,117
Replying to @PositiveInvst
I think the big issue with Neo clouds is their ability to fund build out. Not demand. I think of this like Evergrande. In china. No issues in selling homes. Big issues finding money to build them. Then what happens ? Bankruptcy
5
2
11
4,904
Most of $CRWV $NBIS $IREN and 90 others will die.
There are 96 neoclouds. Of them 11 have raised more than a billion. All of them have raised more than $100M $NBIS $CRWV $IREN @CrusoeAI etc. Of them I think 5 survive the AI winter from 2027-2029. The rest will sell for 1-2 cents of their assets on the dollar to real estate companies.
13
1
1
21
10,525
They will keep going down
High interest rates means AI infrastructure and neoclouds will not recover anytime soon. The first casualty of higher inflation and interest rates will be their stock. $CRWV $NBIS
2
1
10
5,922
$CRWV I mentioned this.
$CRWV == Qwest communications That is all.
3
7
3,972
$NVDA Nvidia has an interim CEO. The board is still choosing a search firm. And Nvidia is in advanced talks to buy @_inception_ai, a two-year-old startup, for more than $500 million plus earnouts. Inception builds diffusion models that write text in parallel, not one token at a time. Its Mercury model runs over 1,000 tokens a second on Nvidia GPUs. The deal would feed Nvidia's own Nemotron model, its answer to Anthropic. $ANT I keep coming back to this: Inception raised a $50 million seed round a year ago, and Nvidia's own venture arm was in it. So Nvidia is now buying, at ten times the cash it raised, a company it already partly owns.
11
22
13,095
$AMD fell more than 6% after Hunterbrook disclosed a short position and published "AMD Has Its Own China Problem." The allegation: a blacklisted Chinese server maker kept shipping BIOS updates with newer AMD firmware after the March 2025 ban, at least six documented. Hunterbrook also tallied 320 million shares of obligations worth $161 billion, plus more than $33 billion in guarantees and leases, and called AMD's deals circular, in the Nvidia mold. It says AMD has made way more promises than Nvidia.
3
2
3,849
Capex is going to hit a Trillion this year alone.
The $300-500B figure is low. The big four hyperscalers alone are guiding ~$720-740B in 2026 capex (mostly AI data centers, chips, power). Add Oracle, CoreWeave and others and it approaches $800B+. US GDP is ~$33T, so this is ~2.2%+ of GDP in a single year. Historically that already matches or exceeds peak annual shares of the interstate highways or telecom buildout. The multi-year AI wave (projected $10T+ through 2032) is on track to average ~3.6% of GDP, larger than the railroad boom of the late 1800s.
1
2,333
$TGT Target’s back: comps +5.6%, strongest growth in 5 years. But half the margin gain was tariff refunds.
1
1,885
Everpure $P (ex–Pure Storage):beat and raise revenue $1.46B, +51% but the FY28 guide was left unchanged, so implied growth drops to ~37%
2,335
$ANT Anthropic's IPO left more than half of the 1,200 institutions that bid with zero shares. Allocations went heavily to US buyers, and foreign orders mostly got nothing. The banks had a reason for the China and Hong Kong orders. They cited ITAR, the arms export rules, to deny allocations in a chatbot company. The filing never mentions it. I have seen allocation games before. I have never seen one justified with arms export rules. The deal still raised $107.3 billion, the largest first-time share sale on record, passing SpaceX. Demand ran about six times the stock on offer. .
1
1
4,049
UnitedHealth −10.2%, Humana −13% after CMS proposed expanding Medicare Advantage audits to reclaim ~$90B in overpayments over 10 years $UNH $HUM
5
5
5,215
Was AI not supposed to be deflationary?
Fed minutes: nearly all officials expect more rate hikes. Gov. Waller floated a 50bp December hike, saying the Fed may need to “shock people.” Minutes call AI demand a material source of inflation risk
4
2,034