Feb 9, 2024 · 1h 28m · allin
E165: Vision Pro: use or lose? Meta vs Snap, SaaS recovery, AI investing, rolling real estate crisis
Chamath Palihapitiya · 24m spoken David Friedberg · 19m spoken David Sacks · 18m spoken Jason Calacanis · 15m spoken
0:00 / 0:00
▶ Watch on YouTube →
gold bands on the timeline = statements, start to end. Hover to read, click to jump. CC turns on captions
In Episode 165 of the All-In Podcast, hosts Jason Calacanis, Chamath Palihapitiya, David Sachs, and David Friedberg discuss the practical enterprise applications and societal implications of Apple's Vision Pro spatial headset. The panel also analyzes financial performance dynamics between Meta and Snap, signals of recovery in the SaaS sector, competitive moats in AI investing, and the impending commercial real estate debt crisis.
How this conversation actually went
Every chapter scored 0–10 on four independent dynamics. Hover any point for the reasoning behind the score. The hosts hold 99.9% of the talking time here. How this is scored →
The hosts as informed peer 7.5 Guest teaching 4.5 Guest disagreement 3.9 The hosts pushing back 4.8
speaking balance: gold is the hosts, purple is the guest (3 minute bins)
Sharpest disagreement ▶ 6:54 Chamath Rejects Goggle Tech Optimism
Chamath directly attacks the premise that Vision Pro spatial computing represents human progress, linking constant screen immersion to mental health crises among young people.
Hardest push from the hosts ▶ 53:38 Chamath Challenges Sacks on Enterprise AI SLAsChamath explicitly rejects Sacks' platform argument for OpenAI, explaining that high latency and hardware limits prevent custom GPTs from working in production applications.
Biggest teaching moment ▶ 1:04:50 Friedberg's YouTube Data Scale MathFriedberg educates the group on model moats by contrasting Common Crawl's 10 petabytes of static web data with YouTube's 1 to 2 petabytes of daily fresh multi-modal data ingest.
The host holds their own ▶ 27:05 Friedberg's SBC Financial DissectionFriedberg demonstrates deep financial analysis by detailing how Snap paid employees 40 times its free cash flow in stock compensation while Meta generated $71B in operating cash.
the scores for every segment, with the reasoning behind each
| Chapter | Topic | The hosts as informed peer | Guest teaching | Guest disagreement | The hosts pushing back | Why |
|---|---|---|---|---|---|---|
| Apple Vision Pro Enterprise Applications & Workforce Productivity | 5 | 6 | 3 | 4 | Friedberg shares detailed observations from testing the Apple Vision Pro in his company's greenhouse facility. He argues enterprise productivity and streamlined workflows are the device's true killer app. | |
| Human-Centric Technology, Robotics, and Societal Wellbeing | 7 | 5 | 5 | 6 | Chamath rejects tech optimism, connecting immersive tech usage to rising youth depression, isolation, and declining societal outcomes. He argues physical human connection must take priority over virtual efficiency. | |
| Vision Pro Market Adoption, Ergonomics, and Consumer Ecosystem | 6 | 5 | 4 | 5 | Sacks outlines historical VR adoption hurdles and bulky form factors. Friedberg counters with praise for Apple's ergonomic cushioning and superior optical pass-through. | |
| Real-World Vision Pro Usage and Public Spatial Computing | 5 | 4 | 4 | 5 | Jason recounts seeing a developer working in public wearing the headset. Chamath reiterates his concern over degraded interpersonal communication habits among younger generations. | |
| Meta vs. Snap Performance Divergence and Governance Breakdown | 8 | 4 | 3 | 4 | Chamath critiques Snap's complete lack of voting power feedback loops compared to Meta, where Zuckerberg responded aggressively to shareholder pressure by cutting overhead. | |
| Financial Efficiency, Stock-Based Compensation, and Meta's AI Recovery | 9 | 3 | 3 | 4 | Friedberg delivers a forensic breakdown comparing Snap's $1.3B stock-based compensation against Meta's massive free cash flow generation and aggressive share buyback programs. | |
| The SaaS Recovery, Cloud Metrics, and Market Re-Baselining | 8 | 4 | 3 | 5 | Sacks presents quarterly cloud growth data from AWS, Azure, Google Cloud, and Atlassian to show a SaaS rebound. Chamath interrogates whether this reflects true economic growth or a baseline reset. | |
| SaaS Business Model & Build vs. Buy Dynamic | 7 | 5 | 3 | 4 | Friedberg highlights how internal engineering teams armed with AI tools can build custom replacements for expensive SaaS vendors, creating severe pricing compression across the industry. | |
| Chamath's AI Thesis: Open Source, Hardware & The Groq Interview | 9 | 3 | 4 | 4 | Chamath articulates his AI investment thesis, arguing open-source models will drive foundation model value to zero while economic returns accrue to specialized hardware providers like Groq. | |
| David Sachs on AI Categorization & OpenAI's Platform Advantage | 8 | 5 | 5 | 6 | Sacks presents a counter-argument to Chamath's commoditization thesis, asserting that OpenAI's consumer lead and custom GPT developer ecosystem create a classic network effect flywheel. | |
| Latency, Benchmarking & Speed in AI Applications | 8 | 4 | 5 | 6 | Chamath pushes back on Sacks' platform thesis, explaining that sub-150ms latency SLAs and tokens-per-second hardware constraints remain unresolved barriers for production enterprise apps. | |
| Model Quality Convergence, Proprietary Data & YouTube's Moat | 9 | 6 | 4 | 5 | Friedberg quantifies YouTube's daily upload scale of 1-2 petabytes versus Common Crawl's total 10 petabyte dataset, identifying YouTube as an unbeatable proprietary data moat. | |
| Commercial Real Estate Crisis & The 'Pretend and Extend' Problem | 9 | 4 | 4 | 5 | Sacks and Friedberg analyze the commercial real estate debt wall, examining bank balance sheet risk and how lenders use 'pretend and extend' tactics to defer recognizing massive equity write-downs. |
Statements from this episode (28)
Friedberg: Apple Vision Pro will boost greenhouse worker productivity by 10x
“And I was like, man, literally every aspect of this job will be massively improved and productivity will go up by 10 X. With these goggles.”
Calacanis predicts Apple will sell 500,000 Vision Pros in 2024
“They've sold already 200,000 units. Which doesn't seem like a lot, but for a V one, that is a lot. And they're gonna sell a half million this year.”
Friedberg: Apple will reach $100B in Vision Pro sales within five years
“I think they're going to sell a hundred billion dollars of Apple vision pros, not this version, but this version plus the next version probably over the next. I would guess for them to get to a hundred billion in sales, it'll take them less than five years.”
Friedberg: Apple Vision Pro hardware setup prevents easy device sharing
“In order to log in, you know, it does like a eye scan. Or you have to have like a lock in like login like you do with your phone, but then you got to reset the eye because it automatically sets the eye position. So when you put on someone else's headset, you g…”
Chamath requires his children to text back immediately in perfect English
“I told one of my children that when I send a text message, I expect an immediate response. Otherwise I am going to cancel that child's phone and take it away. And then separately, when they respond, it has to be in structured, well thought out, perfectly forma…”
Calacanis: Meta hit $14B Q4 profit after 22% headcount reduction
“Reduction in headcount, 22% year over year from 86,000 to 67,000 the last quarter for Meta. And their quarterly profits have increased to an all-time high of fourteen billion dollars. That's profits folks in Q four for Meta. All-time high for the stock price, …”
Chamath refuses to analyze Snap due to lack of voting rights
“I've never taken more than one second to look at that company. And the reason is there's just zero ability for me to have any useful say. So I've never honestly looked at its performance. I've never studied a single characteristic. I've never trended it.”
Friedberg: Snap paid employees 40x its 2023 free cash flow in stock comp
“They paid employees 40 times the free cash flow that was generated for shareholders during the year, which is also equivalent to 10% of the enterprise market value of this company.”
Friedberg: Meta reduced headcount by 22% in 2023 while Snap cut only 3%
“Meta cut employee count by 22% and Snap cut employee headcount by three percent during the year.”
Sacks: SaaS markets drop like an elevator but recover like an escalator
“The market is generally an escalator on the way up and elevator on the way down. So the recovery is going to take a long time. But at least we've bottomed out and we're in recovery as opposed to continuing declines.”
Sacks: Struggling startups can no longer blame macroeconomic conditions
“By the same token, if you're a startup and you're not seeing improvement in your Q for sales, then you no longer have a macro excuse for why you're not doing well.”
Friedberg: Software adoption won't slow down, but SaaS price compression will rise
“And that's what's going to cause perhaps price compression and more competition than has been the case. But I don't think that the adoption of software is going to slow down.”
Calacanis: Founders Fund invested in OpenAI at $29B valuation, avoids other AI deals
“Founders fund bought into open AI. At a twenty nine billion dollar valuation, but aside from that investment, they're generally avoiding the AI deals.”
Chamath predicts AI foundational models will become free with zero economic value
“I think foundational models will have no economic value. I think that they will be an incredibly powerful part of the substrate and they will be broadly available and entirely free.”
Sachs: OpenAI is six months ahead of open-source models
“Most of the tests show that open AI is still ahead of the open source models, and I think even people in the open source movement will tell you that open AI is, call it six months ahead.”
Sachs: OpenAI could win vast majority of consumer GPT market
“If open AI just maintains a little bit of a lead over open source, then it could compound. Yeah, it could basically win the vast, vast majority of the Call it consumer search or consumer GPT market.”
Sachs: OpenAI benefits from a classic developer network effect
“So you have a classic developer network effect where you've got open AI aggregating hundreds of millions of consumers because they perceive that chat GPT is the best. Then you've got developers wanting to reach that audience. So they build custom GPTs on the o…”
Chamath: AWS makes building production AI apps unfeasibly expensive
“For example, like AWS, if you look inside of how expensive it is to build an app there, I've tried, even when they give you credits, the credits they give you aren't sufficient enough to even pay for half the power. And then the way that they schedule and the …”
Chamath: AI apps with 35-40 second response times are dead on arrival
“And the reason is because you could build the best experience in the world that runs on localhost, but if all of a sudden you actually try to launch it as an app and the thing takes 35 and 40 seconds to generate something, it's DOA.”
Chamath predicts public internet AI models will hit parity within nine months
“So if you are using the open internet, Lama, Mistral, OpenAI, they're all getting to the same Quality code point, and they will be there within the next six to nine months.”
Friedberg: YouTube's data repo is 300x larger than Common Crawl
“In YouTube growing by one to two petabytes per day, which makes YouTube's data repository 300 times larger than common crawl, which makes it bigger than anything else that anyone else has.”
Friedberg: YouTube is the world's most valuable asset for AI
“So I think it's the most valuable asset in the world today, based on this thesis, that AI value is going to accrue to the data owner.”
Friedberg: Government will intervene to stop pension commercial real estate losses
“You're not going to see governments let that happen. There's going to be some action at some point.”
Calacanis: China has 50 million excess homes capable of housing 150 million
“China has fifty million homes ahead of schedule. They have fifty million additional supply that can house a hundred fifty million people.”
Sacks: Refinancing low-interest loans will drive commercial real estate insolvencies
“And as the old loans role that were at a much lower interest rate, you have to get financing. Even if you get it at a much higher interest rate, that's when all of a sudden these buildings go from being basically solvent to insolvent.”