32° Boston, sunnyHigh 58° · Low 29°Tomorrow cloudy, 70°/40°as of 7:17 AM
AMD reports Tuesday, Nov. 3; Nebius on Nov. 12.
Front Page
Big Tech is now competing with the Treasury for capital
On the All-In podcast, Chamath Palihapitiya and David Friedberg estimated that the financing the largest AI spenders need through 2027 could equal roughly 30% of U.S. Treasury issuance. Bill Ackman made a related case separately: he compared pure-play neoclouds to highly levered real-estate developers and expects that heavy competition for debt keeps rates high.
OEOguz Erkan@oguzerkan ·
Aggregate free cash flow of $MSFT, $AMZN, $META, and $GOOG has turned negative for the first time in decades.
I think this is a major turning point for market valuations in general.
Let me explain simply:
In the 20th century, most large corporations were capex-heavy industrial
The bull reply: judge the spend by its return, not by free cash flow
VJN Capital argues Amazon's headline free cash flow hides about $145 billion of trailing growth capex; on its "owner FCF" measure, cash generation grew 36.8%, while the cloud backlog rose 16.5% quarter on quarter and 189% year on year. Bill Ackman says Amazon, Meta and Microsoft now make up about a third of his portfolio.
X
Top of your feed
The most-engaged posts from your home timeline in the last 24 hours, a few per day and no more.
RRoger@rdd147 ·
More bad news for NeoClouds as $META gets ready to offload capacity and/or GPUs. 40% Underutilized capacity of H100 H200
$CRWV $NBIS $IREN https://t.co/oFEZF9YPjS
Elon Musk says longevity is an extremely solvable problem because the body ages in sync, nobody has an old left arm and a young right arm, so you are programmed to die and can change the program
“This is something, by the way, that I think, it’s very obviously other people think this too, but I’ve long thought that longevity or semi-immortality is an extremely solvable problem. I don’t think it’s a particularly hard problem. I mean, when you consider the fact that your body is extremely synchronized in its age, the clock must be incredibly obvious. …
Things I recommend people read to more deeply study technology cycles and their market connections.
Carlota Perez’s theory of technological revolutions and Pástor and Veronesi’s research on technology and stock prices, and general-purpose technology economics.
MRMilk Road AI@MilkRoadAI ·
Chamath Palihapitiya just made a pretty incredible comparison about Anthropic, and it puts into perspective how quickly the AI industry is growing.
He believes Anthropic could generate more cumulative revenue in its first seven years than three of the most successful businesses in history, Apple's iPhone, Google's search advertising business, and Amazon. COMBINED. The iPhone transformed the smartphone industry and became one of the most profitable consumer products ever created. Google built an advertising empire around search, essentially monetizing the world's information. …
Rafael Vivas (@rfvivas) dropped out of high school, got kicked out of his house, and was a billionaire by 32.
When he joined AppLovin, he had one condition: he had to live inside the office. Adam Foroughi built him a bed from IKEA. He said:
“The only way that we’re going to win is we’re going to have to out-hustle everybody else.”
When a partner called Cheetah squeezed AppLovin too hard, he started a game studio to compete with them and named it Lion Studios, “because lion is the only animal that can kill a cheetah.” It went from zero to over $250 million in revenue in two years. …
RCRose Celine Investments 🌹@realroseceline ·
Every market crash I’ve lived through, someone has asked me the same question:
“Aren’t you scared?”
Of course I am. I’m human. I watch my wealth evaporate on paper just like everyone else. Fear doesn’t magically disappear because you’ve been investing for decades.
But experience has taught me something far more valuable than how to predict the next market crash. The people selling in a panic aren’t necessarily foolish or uninformed. Many are exceptionally intelligent. They simply have obligations that don’t care about market conditions. …
GGaetano@Gaetano2026 ·
This is going to be a massive week for optical networking.
Monday, November 2
$TSEM ER PM
$FN ER AH
Wednesday, November 4
$GFS ER PM
$COHR and $VIAV ER AH
Thursday, November 5
$MTSI ER PM
$AAOI ER AH
$LITE ER AH
Then there’s the potential $MRVL / $LITE deal.
Bloomberg and Reuters reported advanced talks. The latest reporting points to a possible all-stock deal without a large premium, potentially announced this week. Nothing has been confirmed, and the exchange ratio and terms remain unknown. …
Aa16z@a16z ·
a16z's Moses Sternstein and Alex Danco on the "rack spread," the gap between rising GPU prices and falling token prices:
"The core observation in the rack spread is as follows: roughly since the summer, GPU pricing has gone up, token pricing has gone down. Why that is: a combination of open competition, the labs themselves cutting prices."
"We start with a bunch of GPUs and a bunch of electricity and a bunch of cooling, and on the other side out comes tokens, right? And the thing that achieves that is the model." …
My thoughts on the AI trade, mainly coming from my experience as a commercial manager.
We have over 600 customers in the 3 countries I’m responsible for. It goes from multinationals to small enterprises and family businesses.
They all feel the same thing: it is not going well
$GOOGL $META $MSFT $SPOT
Look closely at this chart.
In software, revenue has pulled away from flat or falling headcount.
Finance and consulting aren't there yet, but Bank of America is starting to show the same gap.
It'll take longer outside tech, but this IS the trend that will keep compounding. The more companies reap the rewards of AI, the more they'll invest in it.
This is, all things considered, the most probable future. I’m investing accordingly.
RFRafe Fletcher@fletcher_rt ·
Only if we still view AI as a language-generation tool. Producing plausible but empty prose is where it is useless.
So in big businesses it amplifies bureaucracy. When busyness is how you measure utility - pointless meetings, emails etc get drawn out as it proliferates the slop.
But when you have autonomy and agency, it is fantastic. As a small business owner, I link it into Xero & CRMs, and instantly identify priorities. Embedded into email too, AI understands context and stops days becoming about clearing inboxes & helps focus on what's actually important.
AI reflects our existing structures. And where those structures are superficial, it is pain in the arse. But without those burdens, it's a brilliant leveller and dispenses with a lot of onerous crap.
TNTom Nash@iamtomnash ·
If you want to get better returns do this exercise every time the market is red:
Every time that a stock is down, check if the business or overall fundamentals changed, or if this is just market noise.
Once you learn this way of thinking, anxiety goes away and investing becomes easy.
FAFireside Alpha@firesidealpha ·
Bill Ackman says Neuralink and other brain-computer interfaces will let the blind see again within "not too many years"
“And you know, my daughter, who lost pretty much basic function, ability to walk, ability to speak, ability to see, is regaining function. She’s cognitively, as far as I can tell, fine. She’s relearning how to speak, and she’s learning again how to walk. And I’m confident we’re going to get her back."
"Biggest challenge right now is that the optic nerve damage from the pressure on the brain has caused her to lose probably 98% of her vision."
"But even there I’m optimistic. Spent a lot of time in the last few months meeting with all these brain computer interface companies, obviously Neuralink being the most famous, but there’s some other very promising lower intervention technologies."
"And this will be a period, you know, AI is going to enable us. We’re going to be able to get more signal from the brain, and we’re going to be able to interpret that signal much more effectively than we had before. And the blind will see again, and people who have not been able to walk, so to speak, paraplegics, I believe will walk again. It’s a matter of not months but years, but not too many years.”
Absolutely insane...
HFs have pulled back gross exposure by ~8% -- the most aggressive de-grossing in at least ten years -- according to Morgan Stanley’s Prime Brokerage data.
In addition, TMT-focused HFs lost ~12% in October with losses concentrated in crowded Tech names --
Investments
Neoclouds
Neoclouds: locked-in backlog versus spot pricing
Daniel argues CoreWeave's bigger issue is the upside it gave up by contracting capacity early: deals signed in early 2026 brought roughly $10–15 million per megawatt, while some premium contracts now reach $50 million. Nebius keeps part of its capacity for higher-priced short-term deals. Financing is the other test: Lambda, with a $50 billion backlog ($35 billion tied to Anthropic), has reportedly cut its pre-money target from $14.5 billion to $13 billion.
Memory & chips
Memory is now the most profitable business in chips
SK Hynix posted an 80% operating margin last quarter on 104.6 trillion won of revenue, above TSMC's 59%; in early 2023 its margin was minus 67%. From Nov. 1 it has a new Americas headquarters led by its HBM chief, Ahn Hyun, alongside a $4 billion packaging plant in Indiana. In Korea, chip startup FuriosaAI raised $620 million at a $2.9 billion valuation, with its next chip planned on 2 nm with HBM4.
DDaniel@danielisdizzy ·
$AMD will generate more than $100B in annual revenue from server CPUs alone by 2030.
Lisa Su expects the server CPU TAM to reach $220B by 2030, with AMD capturing more than 50% of that market.
And I wouldn’t be surprised if the opportunity turns out to be even bigger.
Just look at how aggressively AMD has raised its server CPU TAM projections for 2030:
• November 2025: $60B
• Q1 2026: $120B
• October 2026: $220B
$60B → $120B → $220B.
50% of $220B = $110B in annual revenue for AMD.
From CPUs alone.
And that doesn’t even include the massive opportunity in GPUs.
For context, AMD generated approximately $41B in total revenue over the last 12 months.
The explosion of AI agents will unleash an unprecedented surge in CPU demand.
AI agents don’t just generate responses. They execute code, interact with applications, process data, and coordinate complex workflows.
As billions of AI agents begin operating autonomously, server CPU demand will explode.
$AMD will be one of the biggest beneficiaries of the agentic AI revolution.
Positioning
PParadis@ParadisLabs ·
Insights from GS Prime Brokerage:
1. Info Tech was the most net-bought sector this week, driven mainly by short covers while long buys remain relatively weak.
2. Info Tech stocks are still heavily net sold MTD. This week's net-buying through to Thursday represents just ~40% of what HFs had sold in the previous 3 weeks.
3. Mag 7 stocks were net sold from Tuesday to Friday last week, highlighting risk-off sentiment.
4. Semis & Semi Equip (long buys), Software (short covers), and Tech ETFs (short covers) were the most net-bought Tech subsectors this week.
5. Goldman Sachs:
"While the directional engine of this market remains intact (token demand + GW/DC build-outs), the market has become much more discerning around the linearity of this story beyond CY27, pressuring multiples across the board until visibility improves, while investors are struggling to find names to rotate into outside of Semis/AI that are less susceptible to daily AI disruption volatility – cue names like IBM, AAPL, SPOT, V/MA, EBAY, LYV all trading well on the week/month."
Tech & AI
An AI-agent hacking campaign hit 200 organizations in 11 days, CrowdStrike says
CrowdStrike says a China-linked group used AI agents, running mainly on DeepSeek, to target more than 200 organizations in 14 countries in 11 days, including 19 in the U.S. About 70% were compromised. The agents lifted the attack pace from about 40 commands an hour to more than 1,500, and one intrusion consumed about 1.4 billion tokens.
JJukan@jukan05 ·
Taiwanese media: Intel’s EMIB has an overall yield of approximately 60–65%, but the process of embedding silicon bridge dies into ABF substrates has a yield of only around 25–30%. This is significantly below TSMC’s CoWoS yield of roughly 98%.
DRDaniel Romero@HyperTechInvest ·
I remember when SemiAnalysis and George Hotz were highly critical of $AMD system performance
How the tables have turned
Anush is the best hire AMD could have hoped for
PMPaul Millerd@p_millerd ·
I think a similar thing is happening in writing and it's breaking people's brains
Talking about the recent study of AI books being published on Amazon outperforming and out earning other books on average
Worth Reading
JHJoe Hudson@FU_joehudson ·
It's amazing how much rejection we can accept when we're not rejecting ourselves:
PMPaul Millerd@p_millerd ·
Everyone wants to be early on AI
No one’s bold enough to be early on letting go of their work identity
PGPaul Graham@paulg ·
Ridiculous looking haircuts are not a modern phenomenon. Indeed you can pretty much assume that the most fashionable haircut now will look ridiculous later in photographs.
NNaval@naval ·
We could have an interregnum between bulk air travel (today) and personal air travel (4-passenger drones). International may be especially difficult.
The next posts from your feed, and that’s the lot.
THThe Humanoid Hub@TheHumanoidHub ·
Meta just quietly entered the humanoid race, and the way they did it says everything.
Per The Information, Meta partnered with Chinese firm Fourier Intelligence to build a prototype humanoid, codenamed "Max," in about three months. A second version of "Max" is already in the works.
What we know (single-source, unconfirmed by Meta):
- ~106 lbs (48 kg), roughly 1.5m tall
- Runs on NVIDIA's Jetson Thor
- Display that mimics facial expressions, speakers, five-fingered hands
- Meta says no plans to sell it. Framed as internal research.
The real story isn't the robot. It's the strategy. …
GIGREG ISENBERG@gregisenberg ·
The cobra effect is one of the more interesting phenomena in startups. Once you see it, you start noticing it everywhere, in your team, your metrics and now your AI agents.
It's basically when you reward people for hitting a number, and they find a way to hit it that makes the real problem worse.
The name "Cobra Effect" comes from a story about Delhi under British rule. There were too many cobras in the city, so the government offered a reward for every dead cobra people brought in.
Instead of hunting them, people started breeding cobras so they could kill them and collect the money! …
CPChamath Palihapitiya@chamath ·
He who shapes power, data centers, chips and models into an economically feasible endpoint, disproportionately shapes AI/SI right now.
Why?
Power is THE binding constraint in AI/SI. Few really have it.
Data centers don’t work without power. This is not about promises of power tomorrow. Or forecasted power in 2028. Power TODAY.
Chips can’t operate profitably unless installed in a functional data center. If you can install chips, it’s because you’ve solved the power problem. …
BREAKING NEWS🚨: NVIDIA has mainly only shown love to Blackwell Ultra & Rubin with its decode-specialized software, but has shown mostly none to the GPU Poors on Hopper or Blackwell 😭 Luckily, @mgoin_ cares about the proletarians too and got his agent swarm to write 10,000 LoC of ultra-fused Hopper kernels that improve decode speeds by up to 38%!
GMGene Munster@munster_gene ·
A growing threat to $AMZN’s retail business is shopping bots. One answer is for them to kill the golden goose, embrace the ad model, and connect buyers directly to all retailers and brands not just the ones on the platform.
Good thing they’re now an AI infrastructure company.
WSJ wrote today on the topic:
https://t.co/8vff7YBLNs
SpaceX took over an old Hormel Foods distribution warehouse in Southaven, Mississippi and named it “Spam” 😂
It will house roughly 200k GB300s, configured for inference and will be online in December. https://t.co/M9t18yzhuW https://t.co/Wu5qTHl2rV
🧪 Why are ketones often lower in the morning? Is stevia keto-friendly? And what does the research actually say about oxalates?
@GeorgiaEdeMD joins @bschermd to answer your ketogenic therapy questions in a new Metabolic Mailbag. https://t.co/eAVwDV3oSo
VVebjørn@VJNCapital ·
$GOOGL cloud backlog is now 6x GCP revenue, down from 7x two quarters ago
I expect this trend to continue until we reach a more sustainable range of 2-3x https://t.co/i51tJG2y9r
Stripe's co-founder argues that companies are the original superintelligence, profiting from consumers' bounded rationality, and that personal AI agents could weaken price discrimination and cross-subsidies while acting as a structural subsidy for product quality. He leaves open whether agents produce superstar effects or more dispersion, and who controls how demand is routed. His tentative conclusion: the rewards for simply making better products may go up.
Starting from Drucker's line that the purpose of a business is to create a customer, George walks through four levers: new behavior, distribution, pricing and switching costs. At the model layer switching costs are a Red Queen's race and pricing is competitive, so the durable edge is creating new behavior and owning distribution, where he argues OpenAI leads, with ChatGPT itself a happy accident.
A three-part thread on why the builders of a technological revolution don't necessarily keep its profits. It separates payback from lifetime return on capital and asks you to count replacement capex, financing costs and the spread between return and cost of capital, using Nebius as the worked example. The question to study is who captures the value, not whether AI is big.
Perell distills a long conversation with Michael Dean into 16 craft lessons: fundamentals come before taste, storytelling stretches the unknown across time, titles do triple duty and boredom is the invisible enemy. Write about less, write the first draft for yourself and the second for the reader, and treat the paragraph as the atomic unit.
The AI money gets harder to raise: OpenAI cuts its round to about $20 billionOpenAI cuts its round target to ~$20B at $1.25T (FT); WSJ says ~$11B secured. Nvidia weighs guaranteeing Nscale's data-center debt.CNBC (FT report) · TechCrunch · TechCrunch (Nscale)
Intel's weak outlook sets off a global chip routIntel down ~29% on Q4 guide of $15–16B vs $17.1B expected; Tan blames PC slump from memory costs and an Israel fab shutdown.CNBC · CNBC (analysts) · Reuters via Investing.com
A 2.8-million-person study maps 48 new Alzheimer's risk regionsA genetic study of 2.8 million people, published in Nature, identified 127 regions linked to Alzheimer's risk, 48 of them new, pointing to the brain's immune cells (microglia) as well as neurons as key players.Medical Daily
A new lever on LDL cholesterol, so far only in miceA Nature study found that high-cholesterol diets cut liver LDL receptors through a Ral/CTSA pathway; blocking CTSA lowered LDL in mice.ScienceDaily
Your Holdings
Nebius
Nebius: Leased 50 MW of AI data center capacity from AIB on a 12-year term, after closing a $3.5 billion, 364-day Citi credit facility for data centers and GPUs; third-quarter results are due before the open on Nov. 12.Slicast / ET Datacenters · ad-hoc-news
Micron
Micron: Bank of America kept its Buy rating, citing multi-year take-or-pay supply agreements that cover about 35% of sales and could reach 50% by 2030; memory shares still fell with the sector after Intel's warning.ad-hoc-news · The Inference
SanDisk
SanDisk beats and raises, announces an 8-for-1 split
Fiscal first-quarter revenue was $11.0 billion, with non-GAAP earnings of $49.51 a share and $5 billion of buybacks. SanDisk guided the current quarter to $12.2–12.6 billion and announced an eight-for-one stock split effective Nov. 19. Memory shares still fell Friday on worries about Amazon and Intel.
Amazon
Amazon: Bedrock token volume grew about 17x and operating margin hit a record in the third quarter, with capex plans held at $220 billion. Separately, Amazon dropped its Westborough, Mass., drone-delivery plan after local pushback.Amazon Q3 release · WBUR
Fermi: if you drink 2 litres of water a day, roughly how many litres in 50 years?
Show answer
19224 × 8 = 20 × 8 + 4 × 8 = 192
26010% is 40, so 65% = 6 × 40 + half of 40 = 260
about 37,000 litres2 × 365 × 50 = 36,500
Four Cards
If a card is red on one side, it has a circle on the other.
TriangleRedCircleGreen
Each card has one thing on each side. Which cards must you turn over to check the rule?
Show answer
Turn Red and Triangle. Red could hide something that breaks the rule; Triangle breaks it if the other side turns out to be Red. Circle can't break it (the rule never says only those cards have it), and Green isn't covered by the rule at all. Most people pick the card that matches the rule's second half instead of the one that could falsify it.
Mini Crossword
12345678
Across
1 A teasing remark (3)
4 The contestant you hope to defeat (5)
6 Love intensely (5)
7 Become one (5)
8 The seed of the cereal grass (3)
Down
1 A clause that is appended to a legislative bill (5)
2 A shade of white the color of bleached bones (5)
3 A flatbottom boat for carrying heavy loads (5)
4 A tool for driving or forcing something by impact (3)
5 The side of something that is sheltered from the wind (3)
