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

TL;DR

OpenAI missed user and revenue targets but GPT-5.5 and Codex are gaining on Claude while hyperscalers posted massive cloud beats with $725B in capex planned.

Key Points

• 1.OpenAI missed its 1 billion weekly active user target and 2025 revenue goals. The Wall Street Journal reported they still haven't hit the milestone 4 months into 2026, creating tension between CFO Sarah Frier and Sam Altman over IPO readiness.
• 2.OpenAI carries $600 billion in compute spending commitments roughly equal to its secondary market valuation. CFO Frier reportedly believes revenue isn't growing fast enough to cover expenses, while Altman wants to push toward a 2026 IPO.
• 3.GPT-5.5, built on a new base model called Spud, is winning back developer mindshare from Anthropic. Sacks notes Opus 4.7 has been widely criticized for compute rationing and reduced thinking time, with many developers rolling back to 4.6.
• 4.Codex is taking meaningful share in coding tokens right now according to Sacks. He argues Sam Altman may prove right for the wrong reason — missing consumer targets but winning in the enterprise and coding markets, which are now the most critical AI sectors.
• 5.Anthropic is compute-constrained due to power limitations, not weak demand. Chamath argues every miss in AI forecasts is 100% a supply problem — insufficient power to generate output tokens — not a demand problem, and this hurts Anthropic and OpenAI most while benefiting hyperscalers.
• 6.The Elon Musk vs. Sam Altman trial began with Elon seeking $150 billion in damages and demanding OpenAI revert to nonprofit status. Greg Brockman's personal diary, entered into discovery, contains damning entries including 'we truly want the BCorp' and 'it's still about wanting a for-profit just without him.'
• 7.The trial is a bench trial with Judge Yvonne Rogers, an Obama appointee who oversaw Epic vs. Apple. Polymarket gives Elon a 42–43% chance of winning; panelists speculate the most likely outcome is a settlement crediting Elon back roughly $40 million.
• 8.The four hyperscalers — Amazon, Microsoft, Google, and Meta — reported spectacular earnings with $725 billion in combined 2026 capex guidance. Google Cloud grew 63% year-over-year on $20 billion in quarterly revenue; AWS grew 28% on $37.6 billion; Microsoft Cloud grew 30% on $34.7 billion.
• 9.Free cash flow is collapsing at the hyperscalers: Amazon's fell 97%, while Google, Microsoft, and Meta fell 12%, 12%, and 8% respectively. Chamath warns these companies will become highly levered industrial businesses, paying over 2x spot energy rates on long-term power contracts like Microsoft's Three Mile Island deal.
• 10.GPT-5.5 Cyber has become the second model to complete an AI Security Institute multi-step cyberattack simulation end-to-end, matching Anthropic's Mythos. Sacks argues this is not a doomsday weapon but an automated vulnerability discovery tool — the same capability that powers coding now powers cyber offense and defense.
• 11.Freeberg highlighted an MIT paper showing neural network pruning can reduce model size by 90% with no accuracy loss, cutting inference costs 10x. This technique — dynamically calling smaller pruned models for common queries — could deliver 10x more output per unit of energy from existing data center infrastructure.