OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze

with David Sacks, Jason Calacanis, and David Friedberg

2 May 2026 28 min read 1h 52m

TL;DR

OpenAI missed its 1 billion weekly active user target and revenue goals, but the real constraint isn't demand—it's compute and power. Sacks argues the consumer miss masks a strong enterprise and coding business where GPT 5.5 is now competitive with Claude, while the hyperscalers (Google, Amazon, Meta, Microsoft, Oracle) are best positioned to win because they control the power infrastructure.

Key Moments

David Sacks

“I think that when he made these big compute commitments, it was based on those estimates of hitting the billion users on the consumer side and hitting those revenue targets. The consumer business ended up being weak. So they missed those targets. But in the meantime, coding has become the all-important sector of AI.”

Sacks explains how OpenAI's compute commitments may still pay off despite missing consumer targets, because coding is now the key market.

▶ 6:47

Chamath Palihapitiya

“Everything in this market is power constrained. The 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 identifies power as the critical bottleneck limiting AI model deployment and revenue growth.

▶ 9:31

David Friedberg

“if you can get that 80% of searches or chat interfaces or coding requests reduced down through pruning techniques to smaller models and then you have a whole set of smaller models that can be called dynamically and you reduce inference cost by 90%. you can make much more use, call it 10 times the use on data center and energy capacity than we can today.”

Friedberg discusses MIT research on neural network pruning that could dramatically reduce energy consumption while maintaining accuracy.

▶ 16:15

David Sacks

“if you train a model to do what they do, obviously, like you said, it can operate with a scale and speed that a human hacker can't. So obviously, you know, what you need to do is get these tools in the hands of the white hats, let them do the cyber attacks themselves to then find the vulnerabilities and patch them before the black hats get a hold of these capabilities.”

Sacks explains how AI-powered cyber capabilities should be weaponized by defenders first to harden systems before attackers exploit them.

▶ 24:19

Elon Musk (quoted in lawsuit)

“If we make it okay to loot a charity, the entire foundation of charitable giving in America will be destroyed. That's my concern.”

Elon's core argument in the lawsuit against OpenAI, claiming the company improperly converted a nonprofit into a for-profit entity.

▶ 31:37

About the show

›

All-In is a daily news podcast hosted by four prominent tech investors and entrepreneurs who discuss the latest developments in AI, business, and tech. The hosts dive deep into OpenAI's missed targets, the competitive landscape between Claude and GPT models, and the Elon vs. Sam Altman lawsuit.

Takeaways

Power, not product, is the real constraint OpenAI and Anthropic's ability to scale is limited by access to power infrastructure, not AI capability or consumer demand. With 40% of announced data center projects stalled in red tape and supply chain delays, hyperscalers (Google, Amazon, Meta, Microsoft, Oracle) who control existing power grids have a structural advantage. Smaller AI companies must negotiate for compute access, which directly impacts their growth forecasts and equity dilution.

GPT 5.5 wins on coding, Claude loses compute OpenAI's latest release (GPT 5.5 based on new 'Spud' base model) is receiving strong developer feedback and gaining share in coding tasks, while Anthropic's Opus 4.7 is compute-constrained and underperforming. Developers are actively switching from Claude to GPT for code generation. This validates Sacks' thesis that even though OpenAI missed consumer targets, the enterprise and coding markets are where the real value is being created.

AI-powered cyber defense accelerates software rewrite cycle Models like GPT 5.5 Cyber and Anthropic's Mythos can now automate security research at scale, finding dormant vulnerabilities 10x faster than human teams. This will trigger a one-time upgrade cycle across all software infrastructure as organizations patch pre-AI vulnerabilities. Once complete, cyber defense and offense will reach a new equilibrium, similar to how murder rates stabilized after crime-prevention innovations.