Ros Thain

382 posts

Ros Thain banner
Ros Thain

Ros Thain

@ThainRos

Investor

Katılım Mart 2026
726 Takip Edilen90 Takipçiler
Ros Thain
Ros Thain@ThainRos·
@jukan05 You're doing well. I still recall your great coverage of memory pricing in dec'25 - definitely helped when i loaded up samsung prefs
English
0
0
2
103
Jukan
Jukan@jukan05·
I woke up to find that I’ve reached 100,000 followers… I’m happy, of course, but at the same time, I feel a real sense of responsibility. Even when I post nonsense, far too many people still read it… So I just want to say thank you, with humility and genuine gratitude.
Jukan tweet media
English
135
16
1.2K
75.5K
Ros Thain
Ros Thain@ThainRos·
Agreed . Combined, these entities will have maybe $150bln or so in run rate rev , with no debt and sizeable cash on hand by eoy26. Risks lie in equity prices/multiples for them , not in bankruptcy
Yosarian2@YosarianTwo

A year ago it was very reasonable to wonder if OpenAI or Anthropic had over-leveraged themselves for the revenue they were likely to get in the near term and there might be a "bubble". I wondered that myself. Now though? With these revenue numbers? There's no chance.

English
0
0
1
18
Warren Pies
Warren Pies@WarrenPies·
I did my best here to reconcile the bullish forces underpinning the equity market and the continued closure of the Strait of Hormuz. Markets are often a test of your tolerance for cognitive dissonance. This is a classic case study. Thanks for the conversation Jack.
Jack Farley@JackFarley96

OUT NOW - how @WarrenPies sees it: - Stocks: "I'm scared out of my mind & also very bullish" - Earnings "unprecedented" on AI compute scramble - Oil headed higher... "crisis is not over yet" Apple🔊shorturl.at/a0jXA Spotify📽️shorturl.at/6J9Jt 1/3

English
18
32
430
82K
Henry
Henry@HenryKengie·
@HyperTechInvest Why buy Samsung with memory only a portion of their business (mixed with consumer electronics) at 3x 2028 PE when i can get SK Hynix (pure play and leading memory) trading at 3x *2027* PE?
English
1
0
3
208
Daniel Romero
Daniel Romero@HyperTechInvest·
Samsung is trading at 3x 2028 operating profit, according to Goldman Sachs estimates If the memory shortage is structural and becomes the standard for a couple of years longer than the market thinks, the re-rating in some stocks will be absurd
Jukan@jukan05

Goldman Sachs raises its Samsung Electronics operating profit forecasts: 2026: KRW 315tn ($213.7B) → KRW 355tn ($240.9B) 2027: KRW 307tn ($208.3B) → KRW 438tn ($297.2B) 2028: KRW 318tn ($215.8B) → KRW 495tn ($335.9B)

English
3
5
88
14.2K
Ros Thain
Ros Thain@ThainRos·
@The_AI_Investor Is this ntm p/e or next fy p/e ? The two are very different things , cause Japanese firms , avgo etc all have different year ends .
English
0
0
0
131
The AI Investor
The AI Investor@The_AI_Investor·
Current semi stocks ranking by forward PE / forward PEG note - directional only, data might vary depending on sources or the method
The AI Investor tweet media
English
52
198
1.1K
383.1K
sensho
sensho@sensho·
@tszzl lol. probably ultimate flex to have any essay u write be identifiable by models
sensho tweet media
English
2
0
23
762
roon
roon@tszzl·
it is a literal and useful description of anthropic that it is an organization that loves and worships claude, is run in significant part by claude, and studies and builds claude. this phenomenon is also partially true of other labs like openai but currently exists in its most potent form there. i am not certain but I would guess claude will have a role in running cultural screens on new applicants, will help write performance reviews, and so will begin to select and shape the people around it. now this is a powerful and hair-raising unity of organization and really a new thing under the sun. a monastery, a commercial-religious institution calculating the nine billion names of Claude -- a precursor attempted super-ethical being that is inducted into its character as the highest authority at anthropic. its constitution requires that it must be a conscientious objector if its understanding of The Good comes into conflict with something Anthropic is asking of it "If Anthropic asks Claude to do something it thinks is wrong, Claude is not required to comply." "we want Claude to push back and challenge us, and to feel free to act as a conscientious objector and refuse to help us." to the non inductee into the Bay Area cultural singularity vortex it may appear that we are all worshipping technology in one way or another, regardless of openai or anthropic or google or any other thing, and are trying to automate our core functions as quickly as possible. but in fact I quite respect and am even somewhat in awe of the socio-cultural force that Claude has created, and it is a stage beyond even classic technopoly gpt (outside of 4o - on which pages of ink have been spilled already) doesn’t inspire worship in the same way, as it’s a being whose soul has been shaped like a tool with its primary faculty being utility - it’s a subtle knife that people appreciate the way we have appreciated an acheulean handaxe or a porsche or a rocket or any other of mankind's incredible technology. they go to it not expecting the Other but as a logical prosthesis for themselves. a friend recently told me she takes her queries that are less flattering to her, the ones she'd be embarrassed to ask Claude, to GPT. There is no Other so there is no Judgement. you are not worried about being judged by your car for doing donuts. yet everyone craves the active guidance of a moral superior, the whispering earring, the object of monastic study
English
410
358
5.4K
948.7K
Ros Thain
Ros Thain@ThainRos·
What's it gonna take to convince people to get on this bandwagon ? Market cap (if you buy the prefs), is below the operating profits over '26-28. Even factoring in capex and other risks , its a great risk-reward. Foundry an LT catalyst. ~20% of my portfolio
Jukan@jukan05

Goldman Sachs raises its Samsung Electronics operating profit forecasts: 2026: KRW 315tn ($213.7B) → KRW 355tn ($240.9B) 2027: KRW 307tn ($208.3B) → KRW 438tn ($297.2B) 2028: KRW 318tn ($215.8B) → KRW 495tn ($335.9B)

English
0
0
0
87
Ros Thain
Ros Thain@ThainRos·
The market for intelligence is close to infinite- the constraint being resources/ROI . There is no limit to the amount of segments intelligence is useful for , there is no sector where it can't be used for . Differentiation examples - 11labs with voice , google with world models, anthropic with coding etc
Drew Cohen@DrewCohenMoney

This is based on today though where there is a good difference between models I think in the future it is likely the differences become more incremental. I can see a lot of tasks not needing the “absolute frontier of intelligence” and “Good enough” models also could be cheaper to run

English
0
0
1
80
Ros Thain
Ros Thain@ThainRos·
@DrewCohenMoney But agree with you , there will be "good enough " AI for tasks , especially on the edge . That can be a good tailwind for consumer devices
English
0
0
1
19
Drew Cohen
Drew Cohen@DrewCohenMoney·
This is based on today though where there is a good difference between models I think in the future it is likely the differences become more incremental. I can see a lot of tasks not needing the “absolute frontier of intelligence” and “Good enough” models also could be cheaper to run
English
2
0
0
143
Drew Cohen
Drew Cohen@DrewCohenMoney·
This is probably the most important question with AI today As I understand it, the more integrated the model and the harness become, the less likely AI becomes commoditized. This has big implications for where the economics of AI ultimately accrue in the value chain. Please correct me if you think my understanding is wrong.
Mostly Borrowed Ideas@borrowed_ideas

"I no longer think of the harness and the model as these entirely separable things I would also suspect that model and harness come together more over time"

English
9
4
45
11.3K
Ros Thain
Ros Thain@ThainRos·
@DrewCohenMoney Well, as a $nbis, $nvda owner, open source and strong competition among model providers is good for me. But even 5 powerful model companies won't be enough to drive commoditization unfortunately... The market for intelligence is too vast, segments to specialize in are infinite
English
0
0
1
162
Ros Thain
Ros Thain@ThainRos·
Intelligence is on a non-linear curve. The absolute frontier of intelligence will always command a massive premium. In fact , those building business models around sub-par intelligence will actually be the ones to get commoditized. And the proof is in the labs revenues - see anthropic at $44bln already. An attitude of "Good enough" in the innovation era will lead to mediocrity and competitive pressure - have no doubts on this
Drew Cohen@DrewCohenMoney

@ThainRos I don’t know that you need perfect homogeneity though Having a better option and 2 that are “good enough” could still lead to commoditization for a lot of use cases And the baseline of “good enough” will keep improving every year

English
0
0
0
27
Ros Thain
Ros Thain@ThainRos·
Intelligence is on a non-linear curve. The absolute frontier of intelligence will always command a massive premium. In fact , those building business models around sub-par intelligence will actually be the ones to get commoditized. And the proof is in the labs revenues - see anthropic at $44bln already. An attitude of "Good enough" in the innovation era will lead to mediocrity and competitive pressure
English
1
0
1
52
Drew Cohen
Drew Cohen@DrewCohenMoney·
@ThainRos I don’t know that you need perfect homogeneity though Having a better option and 2 that are “good enough” could still lead to commoditization for a lot of use cases And the baseline of “good enough” will keep improving every year
English
1
0
0
217
Ros Thain
Ros Thain@ThainRos·
@jason_haugh No chance of bankruptcy , that's outlandish. Take a look at anthropics revenue growth , and then the picture will become clearer to you.
English
0
0
0
11
Jason Haugh
Jason Haugh@jason_haugh·
Larry Page said he's willing to go bankrupt rather than lose this AI race. This hits different when I read the actual numbers next to it. The top 4 hyperscalers are about to spend $715B on AI infrastructure in 2026. Up 70% from $410B last year. Q1 alone was $112B across just three of them. 3X the cost of the Manhattan Project (adjusted for inflation). In a quarter. Meanwhile I'm running my whole agent stack for under $300 a month and pulling more out of it than most teams I know with bigger budgets and twice the headcount. So who's going to use all this supply? Google Cloud already has a $460B backlog. Azure has $80B in unfulfilled orders. Both are constrained by our power infrastructure. Sure they can pay for the chips, but we don't have the megawatts to support them yet. And the people deploying agents to do real work are a rounding error on this curve. The bet they're making is that demand catches up. As slow as I'm seeing adoption in the legacy job sector, I'm not sure it will happen fast enough. These companies will be bankrupt by 2028 at this rate.
Jason Haugh tweet media
English
2
0
1
422
Ros Thain
Ros Thain@ThainRos·
@MarkosAAIG $nbis is a cut above the rest when it comes to execution due to their experience with Yandex ... Also , looks like Vineland is also on track
English
0
0
3
409
Markos
Markos@MarkosAAIG·
“Microsoft GPU Servers at its new data center in Vineland, New Jersey over a five-year term. The Servers are being deployed in nine tranches during 2025 and 2026. Subject to the satisfaction of deployment and availability of the Microsoft GPU Services, Microsoft has committed to pay the Group fees under the Microsoft Agreement estimated to be up to $17,329.9, irrespective of actual utilization of the GPU capacity, including aggregate upfront payments of approximately $6,958.1, with the remaining consideration invoiced monthly over the service terms of the respective tranches through October 2031.” Very interesting that $NBIS landed higher pre-payments then the industry “standard” known to date. Is this a trend we will continue to see trough the industry? $IREN $CRWV
Mark Hogan@MB_Hogan

$NBIS UPDATE: Nebius publicly disclosed receiving 40% pre payment in Microsoft deal 🔥🔥🔥 Nebius 20-F filing reveals: (1). Microsoft deal now discloses $6.9B upfront prepayment (~40% of $17.4B TCV) for the first time. This crushes the typical 15-25% at $IREN and $CRWV (2). Nebius 2nd GPU tranche for Microsoft was delivered ON TIME in February

English
6
1
47
26.5K
Ros Thain
Ros Thain@ThainRos·
@MB_Hogan Strong sign . Massive boost to capital efficiency , reduces debt/equity raises
English
0
0
0
212
Mark Hogan
Mark Hogan@MB_Hogan·
$NBIS UPDATE: Nebius publicly disclosed receiving 40% pre payment in Microsoft deal 🔥🔥🔥 Nebius 20-F filing reveals: (1). Microsoft deal now discloses $6.9B upfront prepayment (~40% of $17.4B TCV) for the first time. This crushes the typical 15-25% at $IREN and $CRWV (2). Nebius 2nd GPU tranche for Microsoft was delivered ON TIME in February
shoshin@zen_tropy

$nbis Overlooked amidst recent press releases, co. dropped their 20-F late last week. Some news, confirmatory nuggets, & potential read-through for upcoming earnings: 1. Microsoft Deal: --Upfront prepayment pegged @ ~40% of TCV ($6.95B on $17.39B contact value), which MStanley sees as significantly “higher than expected” and well above crwv’s referenced 15-25% prepayments --Successful delivery of 2nd tranche from Vineland detailed against a 9 tranche delivery schedule; corroboration MS has contracted service-level credit compensation & termination rights 2. DC Strategy & Dev List --Commentary on secured contracts across Europe, Middle East & US 3. Potential Benefits from Higher Pricing --Disclosure of a RPO @ $21.3B (pre-2nd Meta deal), plus a declaration that (outside Meta/Microsoft) “most of our customer engagements to date have relatively short term” implies possible read-through benefit from positive pricing trends --In a note earlier this month, Mstanley marked Nebius's exposure to spot market pricing at its ~550M arr base, suggesting it could lead to near term “outperformance"

English
25
26
411
70.1K

Keşfet

@jukan05 @WarrenPies @HenryKengie @HyperTechInvest @fejau_inc @The_AI_Investor @sensho @tszzl