All-In PodcastHow hyperscalers bet $725B on a grid that can't keep up
The $725B hyperscaler CapEx wave is chasing electricity, not model demand. No GPUs sit dark; OpenAI's user miss is a power problem, not a product one.
At a glance
WHAT IT’S REALLY ABOUT
OpenAI growth stumbles as compute, cyber, and peptides surge forward
- OpenAI reportedly missed user and revenue targets, raising concerns about its massive compute spend commitments and IPO readiness.
- Despite the financial headlines, the panel argues OpenAI’s recent product momentum (GPT-5.5, Codex, and a cyber-focused variant) is shifting developer sentiment versus Anthropic’s latest Claude/Opus release.
- A major throughline is that AI progress is increasingly constrained by power and grid infrastructure, which strengthens hyperscalers’ bargaining leverage and drives unprecedented CapEx escalation.
- The conversation frames AI-cyber capabilities as an inevitable leap that will expose dormant vulnerabilities quickly, likely triggering a large one-time security upgrade cycle before reaching a new offense/defense equilibrium.
- They close with two non-AI themes: enthusiasm around Lilly’s retatrutide (triple-agonist peptide) as a potential next “wonder drug,” and Friedberg’s on-the-ground reflections from a Supreme Court Roundup/Monsanto preemption hearing.
IDEAS WORTH REMEMBERING
5 ideasOpenAI’s ‘bad week’ headlines may mask real product regain.
The group argues GPT-5.5 and Codex are pulling developer mindshare back from Anthropic, even as OpenAI faces scrutiny over missed consumer growth forecasts and large compute commitments.
Power—not model demand—is framed as the true constraint on AI growth.
Chamath emphasizes that token supply is limited by electricity generation and grid buildout (transformers, turbines, permitting), pushing AI labs into tougher negotiations with hyperscalers for capacity and control.
Hyperscalers are transitioning from asset-light cash machines to asset-heavy industrial builders.
With ~$725B 2026 CapEx guidance (AMZN/MSFT/GOOG/META), the panel expects more leverage, financial engineering, and altered valuation narratives as these firms prioritize infrastructure over free cash flow.
This CapEx cycle is not ‘dot-com dark fiber’ because utilization is already here.
Sacks’ core distinction is there are “no dark GPUs”: demand is pulling infrastructure forward now, unlike overbuilt fiber that lacked near-term applications in 2000.
AI-cyber capability is a dual-use accelerant, not the creator of vulnerabilities.
Sacks argues models like Mythos/GPT-5.5 Cyber mainly discover existing bugs faster; the urgent policy/industry question is whether defenders deploy these tools broadly before attackers (including state actors) do.
WORDS WORTH SAVING
5 quotesThe reason that these folks may miss a number or a forecast have nothing to do with demand. It is entirely 100% due to the supply of the power necessary to generate the output token.
— Chamath Palihapitiya
It doesn't create the vulnerabilities. It just discovers them. The bugs were already in the code.
— David Sacks
AI is now synonymous with the growth of the American economy, and if there's no economic growth, there's not gonna be money to pay for all the social programs, there's not gonna be money to pay down the national debt, there's not gonna be money to basically build up our national defense, all these things we wanna spend money on.
— David Sacks
Rumination is the path to unhappiness. Nobody gives a fuck about your feelings.
— Jason Calacanis
The longer the time horizon for a task, the more likely it is to go off the rails.
— David Sacks
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome