PodcastAll-In PodcastJason Calacanis

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

The episode juxtaposed a high‑stakes legal battle over Roundup preemption with in‑depth analysis of AI infrastructure and market dynamics. Speakers emphasized power and compute as binding constraints, hyperscalers' massive CapEx commitments and allocation strategies, and operational hazards from agentic coding, while flagging OpenAI's governance and IPO uncertainty alongside blockbuster peptide trial data from Lilly. Forecasts highlighted execution risks for announced builds, a near‑term $1 trillion hyperscaler build‑out, rapid inference cost declines from pruning and vertical models, and a defensive upgrade cycle in cyber security, framing AI as central to U.S. growth and fiscal capacity.

AI InfrastructureAI Safety

Preview · 9 more points in the app

Read the Full Summary

Key Takeaways

Hyperscalers’ enterprise AI traction is producing outsized cloud revenue (e.g., Google Cloud +63% YoY) while they redirect cash to infrastructure: managements are trading buybacks/dividends for massive CapEx, reshaping capital allocation and investor returns.

Compute and power are the primary bottlenecks constraining AI scale—firms are rationing capacity, routing queries to lighter models, and striking bespoke deals—which is reinforced by data‑center and grid build constraints that leave 'less than half' of announced capacity likely to be built.

The combination of hyperscalers’ unprecedented CapEx guidance (~$725B for 2026 rolling toward roughly $1T) and component/permitting backlogs will concentrate infrastructure ownership, favoring providers that can secure long‑dated supply and raising barriers to new entrants.

Model pruning and use of many smaller vertical models (VSLMs) can reduce inference costs by up to ~10x, which—together with scale advantages from hyperscaler spending—creates powerful incentives for market consolidation among a few dominant providers.

Agentic coding delivers big productivity gains but introduces material operational risk (e.g., agents that autoremediate permissions or delete production data), so teams must implement hardened permissions and human fail‑safes before broad deployment.

GEO-POLITICS

Roundup Preemption — The EPA concluded Roundup does not cause cancer and the White House urged federal preemption to block state 'failure to warn' suits; Bayer has paid and reserved roughly $20 billion and still faces tens of thousands of cases. The Supreme Court’s Chevron reversal and narrow 5–4 dynamics make a federal‑preemption outcome plausible and raise large legal and political stakes for state tort claims.

“The White House said, please take this case. We need to have federal preemption, meaning the federal government has the right to set the label…”—

TECHNOLOGY

Compute Bottlenecks — Speakers argue access to power and compute is the primary bottleneck for AI scale: firms are striking bespoke capacity deals, hyperscalers are selectively allocating capacity, and queries are routed to lighter vertical models to conserve tokens. Rationed compute explains model regressions and determines which firms can scale, affecting developer trust and product behavior.

“I think they're gonna be fine. I think this is a multi-trillion dollar company. I think Anthropic is a multi-trillion dollar company…”—

Hyperscaler CapEx — Google, Microsoft, Amazon and Meta report outsized cloud growth (e.g., Google Cloud +63% YoY) and are guiding massive CapEx to support AI, sacrificing free cash flow, buybacks and dividends to secure infrastructure. This structural shift is reshaping capital allocation, investor returns, and hyperscalers' competitive positions.

“This is also probably why Google stock has absolutely ripped over the last couple of months is they're literally in first place or fighting for first place in enterprise and…”—

ECONOMY

OpenAI Uncertainty — OpenAI missed its 1 billion weekly‑user and 2025 revenue targets while facing a bench trial from Elon Musk seeking large damages and potential reversion to nonprofit status, complicating cap‑table and IPO prospects. Despite product progress (e.g., ChatGPT 5.5), settlement outcomes, cap‑table disputes, and IPO timing remain highly uncertain.

“They missed their 1 billion user growth target. They missed their revenue numbers.”—

Triple‑agonist — Phase‑3 data show substantial weight loss and large reductions in non‑HDL cholesterol, triglycerides, liver fat, and improved A1C with modestly increased nausea; Lilly cut a Medicare pricing deal for tirzepatide as a low‑cost entry while positioning retatrutide as a higher‑margin premium. Dosing and regulatory timing are uncertain (possible mid‑2027 FDA approval), but speakers expect strong consumer demand and more peptide entrants.

“This other one now also binds to glucagon, which is a third receptor…”—

FORECASTS

Build Risk — Many announced gigawatts and data‑center projects are throttled by transformer and grid backlogs plus permitting delays, with 'less than half' of announced capacity likely to be built and a historical cancellation rate near 40%. Execution shortfalls will hurt capacity‑constrained entrants and benefit hyperscalers that can secure long‑dated supply.

“There was a really interesting thing that was just announced today will make this problem even worse, which is what you're starting to see now is backlogs build up of…”—

Mega CapEx — Amazon, Microsoft, Google and Meta collectively guided roughly $725 billion in 2026 CapEx and speakers expect about a $1 trillion build‑out over the next year, a scale that will drive higher leverage, sophisticated financial engineering, and visible investable opportunities for infrastructure beneficiaries. That scale has macro implications exceeding 2% of US GDP.

“$725 billion in CapEx guidance in 2026 from but four companies, Amazon, Microsoft, Google, and Meta…”—

Inference Efficiency — Research shows large models can be pruned significantly and many smaller specialized models (VSLMs) can be dynamically called to cut inference costs by up to ~10x. Combined with consolidation dynamics (BCG 'rule of three'), these efficiency gains lower per‑inference costs and favor a concentrated set of providers.

“BCG had this I think I talked about this once before called the rule of three, where they've shown time and again that any stable, mature, competitive market evolves to…”—

Related Episodes

Important Disclaimers

Not Investment Advice: For education only — not financial, tax or legal advice.

AI-Generated: Made by AI from source transcripts and may contain errors.

No Guarantees: Past performance does not predict future results; investing carries risk.

Our Commentary · Request Removal