E293 · All-In Alpha

conviction thresholds crossed by this episode — 65 / 45 / 15 · ideas born here show where they landed. why 65 isn't always green

▲ +17.3 📈 DRAM/HBM: the bottleneck that matters ember → watch 31.7 → 49.0

▲ +47.2 ⚡ Power is the AI moat: get long energy and the owners of powered land born at watch 47.2

▲ +46.9 🤖 Anthropic IPO (~Oct 2026) is attractively priced at $1.5-2T dormant → watch 0.0 → 46.9

▲ +20.4 📈 2026 is the year of mega-IPOs dormant → ember 6.7 → 27.1

▲ +15.0 🤖 Frontier AI labs keep durable premium pricing (not commoditized) dormant → ember 0.0 → 15.0

each voice's push on conviction this episode, in conviction points — supports and opposes · share = % of this episode's movement

Chamath

9 captures · 33% of movement · 1 idea born

+75.1 / -12.0 → net +63.1

BE

Ben Horowitz guest ×0.5

10 captures · 24% of movement

+57.5 / -6.0 → net +51.5

Jason

8 captures · 24% of movement

+41.8 / -20.2 → net +21.5

Sacks

9 captures · 19% of movement

+38.9 / -12.0 → net +26.9

NEW ▲ VST ⚡ Power is the AI moat: get long energy and the owners of powered land watch 69 ▲ +47.2 0.0 → 47.2

The US is in an electron deficit that the planned build cannot close, so power, not models or software, becomes the binding moat of the AI economy: hyperscalers will turn customers away for lack of capacity ('suck it, we have no capacity for you'), and whoever owns the land, the power and the shell holds the constrained asset. Chamath's call is 'you got to get long energy'; bullish independent power producers that sell into data-center load.

plays VST ·primary CEG TLN evals 2027-10-17

BE

Ben Horowitz support ×1 sentiment guest ×0.5 ▶ 39:50

Yeah, such a disaster. I very much agree with that last one. Yeah, I think that's right.

Chamath support ×3 explicit_prediction ▶ 45:27

but I'm telling you guys, you got to get long energy. I think startups need to find a way of like cobbling together. If you don't have access to electricity, you cannot go through the hyperscalers, is my prognostication.

▲ GOOGL 🤖 Anthropic IPO (~Oct 2026) is attractively priced at $1.5-2T watch 61 POD DISAGREES ▲ +46.9 0.0 → 46.9

Jason support ×2 sentiment ▶ 28:12

So you have to own those because they threaten every single company in the world because they're selling intelligence broadly

BE

Ben Horowitz support ×3 explicit_prediction guest ×0.5 ▶ 50:35

Yeah, so I think the IPO will be fine, like whether you want to be in at 2 trillion or 1.9 trillion or 2.3 trillion.

Sacks support ×2 explicit_prediction POLICY-ADJACENT ▶ 51:54

But overall, I think it is set up to have a strong IPO.

Chamath support ×2 explicit_prediction ▶ 53:58

Can they raise $80 billion in an IPO at $1.7, $8, or $9 trillion? I think so, handily.

▲ IPO 📈 2026 is the year of mega-IPOs ember 43 POD DISAGREES ▲ +20.4 6.7 → 27.1

Jason support ×1 sentiment ▶ 27:43

It's a big number, but it's not an unprecedented number now because SpaceX just did it, right?

BE

Ben Horowitz support ×2 explicit_prediction guest ×0.5 ▶ 28:08

And so I'm not worried at all about the demand for those IPOs.

▲ MU 📈 DRAM/HBM: the bottleneck that matters ember 42 ▲ +17.3 31.7 → 49.0

BE

Ben Horowitz support ×1 sentiment guest ×0.5 ▶ 13:43

And then there's he's got to buy memory like and there's memory shortage and in that.

Chamath support ×1 sentiment ▶ 44:56

Even Google tin cups from memory right now, because NVIDIA has clogged it all up.

▲ MSFT 🤖 Frontier AI labs keep durable premium pricing (not commoditized) ember 38 POD DISAGREES ▲ +15.0 0.0 → 15.0

BE

Ben Horowitz support ×2 sentiment guest ×0.5 ▶ 27:53

Whatever you're holding in the equity markets, whatever you're holding, OpenAI and Anthropic are threats to that. There's almost nothing that you can hold in a portfolio that those two companies don't threaten.

Jason oppose ×2 sentiment ▶ 51:15

Thirty times price to sales ratio, Sacks, and this headwind of open source and enterprises moving to that. Most businesses don't need to solve the last seven math problems in the world.

Sacks support ×2 explicit_prediction POLICY-ADJACENT ▶ 1:00:12

So they definitely have a duopoly when it comes to revenue and market share. I would argue they have a duopoly when it comes to frontier capability.

▼ NVDA 🤖 Nvidia risk: special-purpose silicon + Huawei dormant 2 POD DISAGREES ▼ -12.0 14.1 → 2.1

Chamath oppose ×2 sentiment ▶ 44:56

You have to think, NVIDIA has done an incredible job. Because if you look one layer beyond NVIDIA, all the packaging capacity is them, all the HBM capacity is mostly them.

▲ QQQ 📈 It's 1997-98, not 1999 — the melt-up has years left green 76 POD DISAGREES ▲ +11.0 67.3 → 78.3

BE

Ben Horowitz support ×2 sentiment guest ×0.5 ▶ 14:35

He's got supply in a great shortage, which is extremely different than the Internet bubble.

Sacks support ×2 explicit_prediction POLICY-ADJACENT ▶ 20:11

I would make the argument that this is just volatility, which you're going to get a lot of whenever you're underwriting to multi-trillion-dollar outcomes in the future

Jason support ×1 sentiment ▶ 25:31

So maybe this is just the market overall saying, hey, let's test this, you know, and maybe there's some indigestion there

▲ NBIS 🤖 Enterprise AI sovereignty: workloads leave frontier APIs for private stacks green 94 ▲ +6.1 79.9 → 86.0

Sacks support ×3 explicit_prediction POLICY-ADJACENT ▶ 1:15:12

And so what you're seeing in the enterprise right now is that for a lot of enterprise workloads, you don't need a general superintelligence. You don't need a frontier model. You need a small specialist model that's been trained on your own data.

Jason support ×2 sentiment ▶ 1:15:46

That you own and that you have AI sovereignty, you're not educating a frontier model company.

▲ SPCX 📈 SpaceX post-IPO: multiple ways to win from $1.4T green 91 ▲ +4.6 81.7 → 86.3

Sacks support ×2 sentiment ▶ 40:25

SpaceX, they pay back in roughly one year. That makes sense because I think that SpaceX is monetizing at about 50 billion per gigawatt, and CoreWeave at about 20 billion per gigawatt.

Chamath support ×1 sentiment ▶ 59:26

He's not dumb. He's the smartest of the smart. He's doing spot because he's like, I may cancel on you.

▲ NVDA 🤖 AI compute capex supercycle has 12-24 months of runway green 86 POD DISAGREES ▲ +3.0 92.5 → 95.5

BE

Ben Horowitz support ×3 explicit_prediction guest ×0.5 26mo horizon ▶ 14:35

I said, yeah, like, we're out of everything, you know, till 2028 He's like, oh, it's beyond 2028 Like, I don't think we get stuff back. Like, end of 2028, maybe. So everything else is sold out.

Sacks support ×3 explicit_prediction POLICY-ADJACENT ▶ 21:36

So we know the demand is off the charts right now. We're seeing other indicators. So, for example, old GPUs, like even the H100s or the A100s, which are several generations old, they're still renting at premium prices right now.

Chamath support ×2 explicit_prediction ▶ 44:56

I think we're capacity constrained enough through the entire pipeline.

▲ NVDA 🏛️ The AI doomer campaign fails — no FDAI, open source survives ember 31 POD DISAGREES ▲ +2.6 94.6 → 97.3

Sacks support ×3 sentiment POLICY-ADJACENT ▶ 1:05:38

They're trying to create a panic. This is their latest panic against open source.

Jason support ×2 sentiment ▶ 1:11:02

They're making up scenarios that get a lot of attention. These are fantastical.

Chamath support ×2 sentiment ▶ 1:24:22

And he's the doomer. He's the flag-carrying doomer, trying to essentially have what he had the first time around, which is command and control from a global hierarchy.

BE

Ben Horowitz support ×3 sentiment guest ×0.5 ▶ 1:26:50

Again, zero of those four have been met. We're not on track for any of those four to be met unless something is going on in the big labs that we don't know about.

▼ TLT 🏦 US fiscal spiral keeps long-end yields grinding higher watch 85 POD DISAGREES ▲ +2.1 86.1 → 88.2

Chamath support ×2 sentiment ▶ 6:07

we have a problem in the United States, which is that the US is moving the risk-free rate higher and higher. Who is doing that? These are the central bankers. These are the bond vigilantes.

▼ IGV 🤖 The $3-4T software industrial complex contracts in 2026 ember 34 POD DISAGREES ▼ -1.3 40.1 → 38.8

Chamath support ×1 sentiment ▶ 28:27

If you own a bunch of workflows that have been rented to people as SaaS, theoretically, open AI and Ant are a threat to that.

Jason support ×3 sentiment ▶ 31:14

It feels to me like the deflationary nature of this software, AI, super intelligence, is unique and different than anything we've ever seen.

Sacks oppose ×2 sentiment POLICY-ADJACENT ▶ 33:33

So again, people just tend to think in terms of these zero-sum games and who's going to be wiped out or eliminated.

BE

Ben Horowitz oppose ×2 sentiment guest ×0.5 ▶ 35:53

look, I don't think so. Like, this has been the ongoing prediction starting with the SaaS apocalypse.

▼ NBIS 🤖 Neoclouds are a transitory GPU arbitrage, not a durable business ember 34 POD DISAGREES ▼ +0.0 0.0 → 0.0

BE

Ben Horowitz support ×1 sentiment guest ×0.5 ▶ 14:35

It's why, by the way, we're not invested in any of these neoclouds either, for that reason, because that whole thing scares me, having been through it.

Jason oppose ×2 sentiment ▶ 18:30

15% chance at the end of 27 next year, 23% at the end of 2028 seems a little far-fetched. Given the fact that they have a massive demand for their equipment, as Ben pointed out.

Sacks oppose ×3 explicit_prediction POLICY-ADJACENT ▶ 40:25

So, in any event, payback period is really critical and as long as the payback periods are really short, I don't see how these companies can lose

written during extraction and stored in data/extractions/ep293.json — the auditable source of truth, including everything market-adjacent that did not earn a capture

E293 (Jason, Chamath, Sacks; Friedberg out 'saving the world from famines'; guest Ben Horowitz of a16z) covers the CoreWeave credit scare and neocloud selloff (2:10-45:40), Anthropic investor day and IPO sizing (45:40-1:03:00), Anthropic's IoT-worm and distillation blog posts as 'PsyOp Corner' (1:03:00-1:22:00), and Bill Gates as doomer (1:22:00-end). MARKET THREADS. (1) CoreWeave: Chamath walks CDS at ~953bp (~55% five-year default implied), the DDTL loan at 15%, ~$35B more debt to raise and 75% revenue from three customers, but his net is genuinely mixed ('I don't think that they're in real risk', 'CoreWeave can't go bankrupt' vs 'scary town', ~$5B from a covenant, should cancel fixed contracts and sell spot), so NO mention filed for him on neocloud-arbitrage; logged in digest only. Ben passes on all neoclouds as a 'dangerous arbitrage business' (LoudCloud bandwidth $1,500 to $50/Mb) yet calls shortage a big CoreWeave tailwind, filed as support strength 1. Sacks argues 'just volatility' with short payback periods (oppose, 3) and Jason calls the 15-23% bankruptcy Polymarket far-fetched/free money (oppose, 2). Registry thesis mechanism is hyperscaler commoditization; today's bear case is rate-driven financing cost, filed there because it is the same bearish-neocloud thesis and Chamath's own GPU-useful-life framing in it. (2) Compute scarcity: Ben relays Andy Jassy 'out of everything... beyond 2028' (horizon hint 26 months, end-2028), Sacks old H100/A100s renting at premium, Chamath 'spot will remain highly elevated', 'the fact that Elon did that deal tells you spot's going up'. Chamath's Nvidia packaging/HBM 'complete ball control' filed as oppose on nvidia-silicon-fragmentation (the rival of capex) and support on ai-memory-supercycle. (3) NEW ai-power-scarcity-moat: Chamath doubles down on power as 'the ultimate moat' (1.4 TWh short by 2050, own 'the land, the power, and the shell', 'you got to get long energy', can't go through the hyperscalers); Ben agrees. Lineage: rhymes with the closed ai-power-demand-strains-us-generation-2024; not attached to natgas-powers-ai because Chamath names no fuel and is on record short nuclear/oil and long solar+storage. His 'I don't think you can work through the hyperscalers anymore' cuts against hyperscaler-clouds-capture-the-ai-application-dollars-2025 but reading it as bearish for the cloud stocks is inference (scarcity also sells out their capacity), so no oppose filed. (4) Rates: Chamath says the bond vigilantes are walking the risk-free rate up, a 10Y at 5.5-6% (conditional 'if') makes any financing-spread model 'roadkill', and post-midterms the fix is deficit restraint; filed on us-debt-spiral-long-rates (support, 2). (5) Anthropic IPO (Q3 rev $16.8B, NRR 900%, ARR estimates $70-100B+, Polymarket 74% IPO this year, Thanksgiving target): Ben 'IPO will be fine' at $1.9-2.3T, Chamath 'raise $80B at $1.7-1.9T handily' but argues they should raise $200-250B at ~$1-1.5T as a war chest against a correction, Sacks 'set up to have a strong IPO', Jason 'you have to own those'. Chamath's 'not much other investible stuff in AI to be long AI other than those two' names no rival, so no oppose on google-ai-winner. Duopoly: Sacks 'duopoly in revenue and frontier capability' (frontier-ai-premium support) while also arguing small specialist models are taking enterprise share (filed on enterprise-ai-sovereignty; his same-quote comparative against frontier-ai-premium not filed because his strongest frontier mention this episode is the support, one mention per speaker+idea). Sacks says Meta's Muse, Google and Elon are 'getting close' but behind; not filed as opposes on meta-open-source-ai-leader/google-ai-winner (mixed, no clear verdict). Jason 'compute duopoly, not a token duopoly' and open-source headwind (oppose). Chamath's frontier view (walking down the price curve vs open source; AI pricing moving from CPM-style per-token to CPC/CPA and then 'absolutely batsh** crazy') left in digest as mixed. Sam Altman 2-4% grant chatter not market. (6) Software deflation: Jason's 'decompiling Photoshop', '90% of the value removed' (software-industrial-complex support, 3); Sacks 'vast new category of bespoke software' and Ben 'Jevons Paradox of software' oppose; Chamath's SaaS-workflows-threatened line supports at 1. Rivals visited: enterprise-saas-moats, palantir-no-alternative, tech-melt-up-pre-peak, none named in any quote. (7) Doomers: Sacks on Anthropic's IoT-worm and distillation posts (open-source scare campaign; administration 'rejected the idea of regulating open source'), Ben's Ghodsi four-preconditions 'zero have been met', Chamath and Jason on Gates; all support ai-doomer-regulation-fails. Sacks' Claude-constitution 'Frankenstein problem' concern is safety commentary, not captured. DECLINED: Middle East war making Gulf data centers uninvestable (Jason/Chamath/Sacks; no registry fit, no instrument named); data-center local entitlement headwinds (describes backlash, does not argue for or against its reversal); Ben's Opus 4.5 SQL-injection story and Jason's XBOW/HackerOne white-hat pitch (anecdote, no security-stock call, ai-cyber-upgrade-cycle not captured); Sacks on Apple home-automation hub (not the local-model thesis); Chamath's 'one or two-year correction' and 'eventual correction' (undated hedge); Ben on private markets absorbing public-market growth (structural, past tense). Not discussed: the other 44 registry ideas. GUESTS/CONFLICTS: Ben Horowitz (a16z) says a16z holds 'a lot more stock' in OpenAI and is 'philosophically long both' labs, so his Anthropic-IPO and frontier-premium supports are book-talking; a16z signed the open-weights letter, so his doomer support is also aligned with the firm's lobbying. SACKS: flagged sacks_policy true on AI statements (former czar, err-toward-marking); SpaceX payback statement flagged false. LABEL DEFECTS: roster matches the intro (Friedberg genuinely absent, stated by Jason at 1:59); Jason top talker (99/64/55/51); no time gaps. 'Michael Intrator' at 4:17 is a played March interview clip (cued 'roll that tape'); his label at 5:56 ('What is that?') and 1:18:09 ('I feel unsafe. It's going to be okay, Ben.') are host bleeds. Ben's label also carries short host bleeds with no captures: 0:23 (answering Jason's greeting to Sacks), 35:53 opening 'Ben?', 53:31 ('We will get back to PsyOp Corner'), the mid-turn 'We manifested what you're saying, Ben' inside 1:22:38, and the opening 'I was like, we're good. We're going to have Ben Horowitz on' at 1:26:50 (Jason); the captured Ben quote in that 1:26:50 turn sits after the bleed and is his Databricks/Ghodsi argument. Content receipts: Ben = a16z not in neoclouds, LoudCloud, former CEO near-death financing; Sacks = addressed as out of the czar position immediately before 1:05:38; Chamath = 'my team's estimates'.

CoreWeave Crash, AI Trade Cracks, Anthropic IPO Red Flags, Bill Gates Doomer

Tier crossings

Who moved the board

What got argued (14 ideas)

Episode digest