E165: Vision Pro: use or lose? Meta vs Snap, SaaS recovery, AI investing, rolling real estate crisis

Apple Vision Pro: Usage and Prospects

• David Freeberg reports ordering Apple Vision Pros online due to long queues and "lunatics" at physical Apple Stores, noting a one-month delivery delay before pickup.
• Freeberg utilized the device in a North Carolina greenhouse facility, observing that the headset could streamline complex tasks like pollen collection and data entry, potentially increasing workforce productivity by 10X.
• Freeberg identifies spatial video recording and real-time data ingestion as key enterprise use cases, allowing for immersive training and "Minority Report"-style workflow management.
• Freeberg remains optimistic about enterprise adoption despite the device being a "v1" prototype, predicting it will follow the trajectory of the iPad, which initially faced skepticism before becoming ubiquitous in agricultural sales.
• Chamath and Freeberg express concern that immersive technology could exacerbate societal issues among youth, citing correlations between digital immersion and rising depression, suicide rates, and isolation.
• David Sacks acknowledges the device is comfortable and superior to previous VR headsets like Oculus regarding motion sickness but argues the form factor must eventually shrink to the size of regular sunglasses for mass daily adoption.
• Sacks estimates Apple could reach $100 billion in Vision Pro revenue within five years, surpassing the initial adoption curve of the iPad, driven by the existing app ecosystem and high-value use cases.
• Freeberg notes the device creates a hyper-personal experience requiring eye scans for login, preventing easy sharing of the device between users.

Meta vs. Snap: Financial and Governance Divergence

• Meta stock hit an all-time high of $470/share with a $1.2 trillion market cap, driven by record quarterly profits of $14 billion, despite a 30% drop in its ad spend environment.
• Meta reduced headcount by 22% (from 86,000 to 67,000) and used excess cash to buy back $20 billion in stock, reducing share count by 0.5%.
• In contrast, Snap's stock has fallen 60% from its 2017 IPO price, with operating losses widening from $700 million in 2021 to $1.4 billion in 2023.
• Snap increased operating expenses by $1 billion annually between 2021 and 2023 while revenue growth remained flat (4.1B to 4.5B).
• Chamath Palihapitiya highlights Snap's governance issues, where founder Evan Spiegel holds 99% voting control, creating a "zero ability" feedback loop for shareholders and preventing activist intervention.
• Snap's stock-based compensation (SBC) for 2023 totaled $1.3 billion, representing 40 times its $35 million annual free cash flow, whereas Meta's SBC was only 20% of its $71 billion operating cash flow.
• Chamath notes Snap's 10% headcount cut was a reaction to market pressure, whereas Meta's 22% cut was a proactive restructuring to align with profitability and AI infrastructure.

SaaS Sector and AI Investment Landscape

• SaaS growth appears to be re-accelerating after 6–7 quarters of deceleration or negative net New ARR, with Q4 2023 showing a rebound in cloud platform revenue for AWS, Azure, and Google Cloud.
• Companies like Atlassian report positive net New ARR growth in Q4, suggesting the sector has bottomed out and is transitioning from a "SaaS depression" to a recovery phase.
• Freeberg observes a "build vs. buy" dynamic where enterprises are replacing expensive SaaS tools with in-house software, leading to price compression in the SaaS market.
• David Sacks identifies three VC camps in AI: those betting on incumbents (Microsoft, Google, Amazon), those waiting for open source to win, and those making direct bets on startups.
• Sacks predicts foundational models will become economically worthless commodities due to open-source competition (e.g., Llama, Mistral), with value shifting to "picks and shovels" like token-per-second infrastructure providers.
• Sacks identifies Amazon, Cerebras, Google (TPU), and NVIDIA as the primary players capable of owning the infrastructure required for AI inference and training due to proprietary hardware.
• Sacks argues OpenAI maintains a lead through a network effect: a massive consumer base attracts developers to build custom GPTs, which further enhances the platform's utility compared to open-source alternatives.
• David Freeberg emphasizes that speed (latency) is a critical, often overlooked bottleneck; if AI inference exceeds 30 seconds, production adoption fails regardless of model quality.
• Freeberg warns that current cloud costs and orchestration for AI development are prohibitive, with credits often insufficient to cover power costs, making production deployment difficult for startups.
• Freeberg predicts data advantages will shift toward platforms with proprietary data pools, specifically identifying YouTube as a repository 300x larger than Common Crawl, creating an insurmountable moat for Google.
• Freeberg suggests that while open-source models will converge on quality parity within 6–12 months, proprietary data from sources like Google Workspace or YouTube will determine long-term economic winners.

Commercial Real Estate Crisis

• The US commercial real estate market is valued at $20 trillion, with the office sector representing $3 trillion, a value Freeberg estimates is overstated and closer to $1.8 trillion.
• If office values drop to $1.8 trillion, equity holders could lose two-thirds of their investment, with the debt exposure (roughly $1.2 trillion) held largely by regional banks and thrifts.
• David Sacks notes that typical office deals are leveraged 2/3 debt to 1/3 equity, meaning a value write-down wipes out equity entirely and leaves debt holders with significant losses.
• Banks are employing a "pretend and extend" strategy, allowing landlords to roll interest into principal and extend loan terms to avoid foreclosing on properties that are now underwater.
• Multifamily real estate faces a different crisis driven by refinancing risk rather than vacancy; developers locked into low rates (3–4%) must refinance at 8–10% rates, creating negative leverage.
• Sacks contrasts the two markets: residential can adapt quickly to market conditions via annual lease resets, whereas office vacancies are locked in long-term leases, creating a delayed but inevitable "cliff."
• Freeberg predicts government intervention will likely be required to prevent pension funds from absorbing the massive write-downs in office real estate equity.
• The crisis is exacerbated by a structural shift to remote work, where younger employees resist returning to offices, leading to sustained high vacancy rates and declining lease renewals.
• Unlike the residential market, where defaults are often manageable, the office market's insolvency is structural, with many buildings holding debt-to-equity ratios of 70:1 that no longer support the loan.