Jay

185 posts

Jay banner
Jay

Jay

@the604og

Turks and Caicos Katılım Kasım 2013
816 Takip Edilen57 Takipçiler
Jay
Jay@the604og·
@RealNickMugalli Sounds crazy buy not surprised if it hits $100 bn run rate by year end (to the extent compute allows). It will be a straight line up.
English
0
0
0
51
Nicholas Mugalli
Nicholas Mugalli@RealNickMugalli·
Anthropic is projecting $300 billion in annual revenue by 2030. Its model training costs peak at ~$120 billion in 2028. At that peak, training costs exceed 100% of revenue. That is the central tension in this business. The left panel shows Anthropic's revenue broken into three segments—consumer (Claude subscriptions and API), enterprise (B2B contracts and cloud partnerships), and new products including hardware. The trajectory reaches roughly $13B in 2025, crosses $100B around 2027, and approaches $300B by 2030. The top right panel shows what it costs to produce those models—training compute costs grow from near-zero in 2024 to a peak of ~$120B in 2027–2028, then decline as algorithmic efficiency improves. The bottom right panel is the one that matters to anyone evaluating Anthropic as a business. Training cost as a share of revenue exceeds 100% from 2026 through 2028 — the company is spending more on model training alone than it takes in. This is the structural condition of every frontier AI lab today. The bet is that efficiency gains compress per token training costs while deployed model revenue continues compounding, and the lines cross permanently in 2029. That crossing is what a path to profitability looks like at frontier scale. The position—Anthropic at $44B ARR today is real and accelerating. The $300B by 2030 projection is plausible given enterprise adoption curves. But the company remains pre profit through at least 2027, and the margin is entirely hostage to whether the next generation of models requires a step change in compute or benefits from efficiency gains. Every hyperscaler dollar going to inference infrastructure rather than training clusters helps. Every new frontier model that requires 10× more compute hurts. The next 18 months of model releases will determine which force dominates
Nicholas Mugalli tweet media
Nicholas Mugalli@RealNickMugalli

Anthropic just hit $44 billion in annualized revenue run rate. In February 2024, they were at $100 million. That is a 440x increase in 18 months…🤯 The shape of this curve is what matters. This is not linear growth. This is not even exponential in the traditional sense. - From $0.1B to $1B took roughly a year. - From $1B to $4B took three months. - From $4B to $14B took four months. - From $14B to $44B took five months. Each successive doubling is happening faster than the last against a larger base. That is not a common property of revenue growth at any scale… Where would be Anthropic valued at this rate in five years? $50T?

English
6
17
121
31K
Jay
Jay@the604og·
@funnyfish88 @veronicaluna77 对于大政府会阻碍科技创新有什么可研究的方向吗。看东大觉得还行啊
中文
1
0
0
14
Kevin Fish
Kevin Fish@funnyfish88·
因为这波自动化/人工智能革命,带来的经济收益是滞后和不确定性的,而美国有太多基建功课要补,Elon对此有更高维和量化的认知。 相比较,债务是更迫在眉睫的问题。钱就是权利,政府开支降不下来,是与小政府理念背道而驰,而更大的政府会阻碍科技创新,这与左右派当政无关,这才是第一性原理。 至于中国问题,在政策层面也不是马斯克能影响的。但更科技创新友好的美国,会在中美对抗中有更大优势。
中文
3
1
6
644
Veronica Luna
Veronica Luna@veronicaluna77·
说实话 这么多年 这件事对我冲击挺大 主要我没法接受马斯克的反应 目前我还是寄希望于: 1、俩人在演戏 为了彻底公布爱泼斯坦名单或者成立新党彻底干掉民主党 2、俩人终会和好 但这确实削减了马斯克在我心中的形象 虽然很伤心 但我觉得他这次过头了 不过从第一性原理回归本质冷静思考: 1、川普连希拉里、奥巴马、拜登这些杀他的人他都能放一马 连大法官这么颠倒黑白 他还依然尊重法律 他怎么可能报复针对马斯克 就算吵架是真的 我还是觉得不至于因此报复马斯克 2、AI革命和宇宙探索是不是颠覆性的 如果他比工业革命力量更大 有什么理由放弃特斯拉和spcaex 抓住重点回归本质 自己会放松平静很多 我不想再多说两人如何 祈祷🙏
中文
64
14
216
48.3K
Jay
Jay@the604og·
@leopoldasch your gonna be this run's roaring kitty
English
0
0
0
24
Messier 87
Messier 87@XiaoZha62335239·
Amazon is actually a big semi play (all there asic and ARM cpu) and the market hasn't fully priced it in like GOOG's TPU. If the semi fomo keeps going the market will notice. And It is in the position where you cannot lose. If AI capex down, amazon's FCF explodes, you win. If AI capex keeps going, market will notices its asic and arm cpu. If asic market share increases dramatically, it also wins. You literally cannot lose with Bezos.
Messier 87 tweet media
English
10
0
4
2.7K
Jay
Jay@the604og·
@ShanghaoJin 什么群,金老师,可以加么
中文
0
0
0
189
Herman Jin
Herman Jin@ShanghaoJin·
强烈怀疑你们看了我前几天在群里传的小作文😂
Citron Research@CitronResearch

$NOK — why comps make it a $20 stock TODAY, not someday! Before we talk about Nokia, Citron owes readers an apology. We've been too negative on the AI data center trade. We missed it and called tops way too early. Here's what has been learned: the best trade isn't shorting the stocks that already tripled. It's finding the one the market hasn't figured out yet. That's Nokia. Who cares that it is a 52 week high....looks like it is going higher While telecom analysts are still valuing Nokia as a slow-growth utility. However, with the Network Infrastructure segment now targeting 18%–20% growth in Optical/IP, the narrative is shifting toward a high performance hardware play. The whole story in plain English Nokia makes the equipment that moves data between AI data centers. As AI grows, those highways need to get wider and faster. Nokia builds the highways. A year ago, Nokia bought a company called Infinera , which meant Nokia went from reselling other people's chips to owning its own chip factories. That's a huge deal. Every other company in this space that owns their own chips (Lumentum, Coherent) trades at a massive premium. Nokia doesn't yet. The market hasn't caught on. Nokia also hired a new CEO last year Justin Hotard, who came straight from running Intel's AI business. This is not a telecom guy. This is an AI guy now running Nokia. And in October, Nokia signed a $1 billion partnership with NVIDIA that came with an investment. Last $NVDA investment Citron told you about was $NBIS when it was $20 a share...yes ! Now the numbers. Nokia management has literally told investors what the company will earn in 2028. Add it up and you get roughly 50 cents per share in earnings by 2028. That's their number, not ours. Here's the thing nobody's doing the math on: every other AI infrastructure stock is already priced based on what they'll earn in 2028. Ciena trades at 49 times those earnings. Coherent at 46 times. Apply the same math to Nokia 50 cents times 49 and the stock is worth $24.50 right now. Today!! And here's gravy. On last week's earnings call, the CEO said , out of nowhere, nobody asked — "a big milestone later this year with NVIDIA." Seven Wall Street analysts were on that call. None of them followed up. None of them asked what he meant. When Lumentum got its NVIDIA moment, the stock went from $49 to $960. Nokia just told you its NVIDIA moment is coming this year. And the stock hasn't moved. That's the whole trade. Nokia is the AI infrastructure stock the market forgot to reprice. Missed the first wave of this. We're not missing the second.

中文
20
8
74
42.6K
Jay
Jay@the604og·
@JerryCap Max Cook - good name for an investor. On par with Max Long
English
0
0
1
395
DJ
DJ@mktoperative·
@TMTLongShort Interesting take. Why would they need to use Baidu (consumer) or did you mean BeiDou Satellite Navigation System?
English
2
0
1
1.4K
Just Another Pod Guy
Just Another Pod Guy@TMTLongShort·
The only thing worse for China than Iran calling a truce and sharing joint custody of the Strait with the U.S. is if the regime collapses completely and the U.S. gets total control of the Strait. That is why China has been sharing targeting ISR with Iran via Baidu. That is why Xi is certainly not going to encourage Iran to maintain a standoff posture until a Xi meeting a full three weeks from now as the blockade deteriorates the regimes grip. And that is why reports that Xi is the levelheaded impassive player here are almost certainly false. Now that is not to say the Chinese won’t attempt to leverage the moment by making believe they are benevolent by diverting inbound energy shipments to more unfortunate importers who don’t have reserves. But it does mean that escalation is more likely than not unless we are very close to the point of regime change. I don’t think we are yet. And so I think things escalate from here regardless of the Xi meeting.
English
13
19
291
23.7K
Jay
Jay@the604og·
@orrdavid How abt Polymarket?
English
0
0
0
4
David Orr
David Orr@orrdavid·
If anyone thinks up a good way to hedge that probably pays 500:1+ please DM me the idea. It has to scale at least to $100k of volume at that price. If I use your idea, I will ship you $10k each time I put the bet on (twice a year). It has to be a bet you can actually put on in practice.
English
11
1
44
14.8K
David Orr
David Orr@orrdavid·
I am risking .1% of my fund on OTM puts to hedge against an unexpected Chinese invasion of Taiwan by April 17th. This is and right before fall are the two key windows. If they invade, this bet will pay ~300:1. I'll keep doing this twice a year for the next several years.
English
63
16
490
112.8K
Oaky
Oaky@oakycapital·
@ContrarianCurse Never saw somebody abbreviate them as Ant before lol
English
3
0
6
3.1K
SuspendedCap
SuspendedCap@ContrarianCurse·
The long Ant / short OAI is getting unwound fast
English
9
5
282
43.4K
Jay
Jay@the604og·
@bellweirboy Thx sounds like something kenny would do seeing how he lives lol
English
1
0
0
19
Bellweirboy
Bellweirboy@bellweirboy·
@the604og I’m not 100% sure. Suspicion is of a charter jet outside the core Citadel / Griffin fleet but definitely frequently interacting with fleet. Sort of standby arrangement when needed. Based at Opa Locka whereas others based at Fort Lauderdale.
English
1
0
0
29
Bellweirboy
Bellweirboy@bellweirboy·
How Kenny Griffin et al live: Mayo Force One +1 arrive in San Francisco from Miami for Super Bowl. Both currently en route back to Florida.
Bellweirboy tweet mediaBellweirboy tweet media
English
1
0
1
111
Rui Xu
Rui Xu@xu545302·
The Answer to the AI Era Is Dumb Devices. For many years, we made everything "smart." Product managers guessed how you'd live your life, hard-coded it into firmware, and shipped it. AI changes everything. The future isn't smarter logic — it's dumb, excellent hardware with AI on top.
Rui Xu tweet media
English
7
0
8
2K
Sag Harbor Capital
Sag Harbor Capital@sagharborcap·
Learning how to behave with the visionary investors and great people at @slow
Sag Harbor Capital tweet mediaSag Harbor Capital tweet media
English
5
1
47
21.2K
Jay
Jay@the604og·
@CL207 cat how do you feel about meta hard-pivoting away from VR? Feel like we are still 10+ years from wide spread adoption.
English
0
0
0
20
CL
CL@CL207·
not that crypto is making higher highs, but it been doing pretty alright, feeling like a rather long term bottom, maybe still in the process of forming one am pretty lazy in terms timing bottoms, cat think 60s is already decent price, maybe if lucky, revisit lows
CL@CL207

am just long stocks, week open, all time high every day, every hour, every second, cpi, all time high, they sue the conglomerates for over reaching power, conglomerates win, all time high imagine owning crypto

English
51
51
1.1K
95.9K
Jay
Jay@the604og·
@CL207 Do you ever recycle your computer?
English
1
0
0
38
CL
CL@CL207·
didnt know 16" thinkpads were so much smaller than 17" ones just got a new one😎
CL tweet mediaCL tweet media
English
60
1
296
41.7K
Gregory Blotnick
Gregory Blotnick@gregoryblotnick·
my man...I'm not arguing, I have no agenda and I'm not an LLM hater. I would love nothing more than 100% accuracy here. I hear you on the prompt, I understand. At the same time, something basic like "MSFT/NVDA EPS," you have to agree that requiring much more than that is...an ask. those are the only two times I've used LLM's to generate data, one was MSFT EPS, one was NVDA EPS. both failed, I understand that with a better prompt it might have gone better, but as far as widespread use (Excel modeling), which I would LOVE to see totally automated...there's a long way to go. lots of people are as tech-stupid as I am, if not moreso, and if "MSFT EPS" isnt 100% perfect on the first try, theyre out. I appreciate ur feedback on the prompt.
English
1
0
1
39
christian
christian@curious_vii·
Is that a healthshare plan? I was actually on Medicaid when working on my last startup and then had to transition to an ACA marketplace plan off cycle, which was an unpleasant Kafka-esque experience, but I’m now on an HMO with sky high deductibles (still not cheap). Haven’t seen a PCP in about 10 years, and I don’t really have any plans to.
English
2
0
1
90
christian
christian@curious_vii·
The administrative friction in the US healthcare system is incredibly demoralizing. Generally unpleasant. Dehumanizing to all parties involved. Humiliating, even—a feeling of being ever so slowly crushed by a soulless, creaky bureaucratic machine, and there’s nothing you can do.
English
0
0
3
182
King Headass
King Headass@head_ass_420·
@gregoryblotnick I mean I understand your point (you want the AI to answer your prompt accurately) but you’re not using the tool properly. I’m not going to argue with you because you have a very obvious agenda (or are just dense)
English
1
0
2
126