Jason Calacanis hosts this episode of the All-In Podcast with co-hosts Chamath Palihapitiya (Social Capital founder), David Friedberg (Ohalo CEO), and David Sacks (Craft Ventures founder). The discussion covers major developments in AI infrastructure spending, OpenAI's missed targets, and the ongoing legal battle between Elon Musk and Sam Altman.

The conversation explores the massive capital expenditure commitments from Big Tech companies totaling $725 billion for 2026, driven by insatiable demand for AI compute capacity. The hosts analyze whether this represents a sustainable investment cycle or a potential repeat of the dot-com infrastructure bubble.

Additional topics include the emergence of AI-powered cybersecurity capabilities, new weight-loss drug developments, and Friedberg's firsthand experience attending a Supreme Court hearing on federal regulatory preemption in pesticide labeling cases.

OpenAI's Growth Targets Miss Despite Massive Compute Commitments

OpenAI failed to hit 1 billion weekly active users by end of 2025 and missed 2025 revenue targets, despite operating at a $20-30 billion run rate according to Wall Street Journal reporting.

The company has $600 billion in spending commitments for compute infrastructure, roughly equal to their entire enterprise valuation on secondary markets.

CFO Sarah Fryer reportedly worries revenue isn't growing fast enough to support expenses, creating tension with Altman over IPO readiness and public reporting standards.

Sacks argues OpenAI may succeed 'for the wrong reasons' - missing consumer targets but benefiting from enterprise coding demand and superior compute capacity versus competitors like Anthropic.

Big Tech's $725 Billion AI Infrastructure Spending Spree

Amazon leads with $200 billion CapEx guidance for 2026, followed by Microsoft and Google at $190 billion each, and Meta at $145 billion.

Google Cloud revenue grew 63% year-over-year to $20 billion quarterly, while Microsoft Cloud grew 30% to $34.7 billion and AWS grew 28% to $37.6 billion.

Free cash flow declined dramatically: Amazon down 97%, while Google, Microsoft, and Meta each dropped 12%, 12%, and 8% respectively as companies prioritize infrastructure investment.

Chamath warns hyperscalers are signing energy contracts at 2x prevailing spot rates, potentially transforming them into 'big, bulky industrial businesses' with questionable valuations.

Power Constraints Emerge as AI's Primary Bottleneck

Chamath identifies power access as the critical constraint: 'Everything in this market is power constrained. The reason these folks may miss numbers has nothing to do with demand.'

Less than half of announced gigawatt projects are actually under construction due to supply chain delays and regulatory red tape affecting transformers and grid infrastructure.

Power constraints benefit hyperscalers (Oracle, Amazon, Meta, Microsoft, Google) over independent AI companies, forcing equity and control negotiations for compute access.

This creates opportunities for Elon's xAI/Grok, which has excess capacity, potentially enabling deals with companies like Anthropic beyond the existing Cursor partnership.

Musk vs. OpenAI Trial Reveals Damaging Internal Communications

Elon seeks $150 billion in damages, claiming breach of charitable trust and unjust enrichment as OpenAI transitioned from nonprofit to for-profit structure.

Greg Brockman's diary entries provide smoking gun evidence: 'We truly want the B Corp. The true answer is that we want Elon out' and 'Can't see us turning this into a for-profit without a nasty fight.'

Brockman documented: 'In the end, it's still about wanting a for-profit just without him' and worried 'this story will correctly be that we weren't honest with him.'

The bench trial before Obama-appointed Judge Rogers in Oakland could delay OpenAI's IPO plans and force corporate restructuring if Elon prevails.

AI Cyber Capabilities Reach Frontier-Level Performance

OpenAI's GPT-5.5 Cyber became the second model to complete multi-step cyber attack simulations end-to-end, matching Anthropic's Mythos capabilities according to AI Security Institute testing.

Unlike Mythos, GPT-5.5 Cyber appears commercially ready with sufficient compute capacity to serve enterprise customers, while Anthropic remains compute-constrained.

Sacks emphasizes these models discover existing vulnerabilities rather than creating them: 'The bugs were already in the code. They were sitting there waiting for some hacker to discover.'

Chamath reveals 'the best cybersecurity company in the world' has demonstrated ability to 'penetrate and essentially manipulate every model,' highlighting ongoing security challenges.

Retitrutide Shows Breakthrough Weight Loss and Health Benefits

Phase 3 trials demonstrated average 37-pound weight loss from 214 pounds baseline in 40 weeks, compared to 6 pounds on placebo.

The triple-agonist drug (GLP-1, GIP, and glucagon receptors) showed 80% liver fat reduction, 27% non-HDL cholesterol decrease, and 41% triglyceride reduction.

Glucagon receptor binding favors fat burning over muscle loss, making it attractive to fitness communities for cutting weight while maintaining muscle mass.

Expected FDA approval by mid-2027, potentially sooner, with Lily positioning it as premium upgrade to $50 Medicare-priced Tirzepatide.