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

Summary

Theses (episode spine)

• OpenAI missed its 2025 target of 1 billion weekly active users (still not reached 4 months into 2026) and its 2025 ChatGPT revenue target, creating tension between CFO Sarah Frier (who doubts IPO readiness) and Sam Altman (who wants to move faster).
• Sacks argues OpenAI may ‘end up being right for the wrong reason’: it missed consumer targets but enterprise/coding demand is booming, and its heavy compute investment now gives it more capacity than compute-constrained Anthropic, which has been forced to engage in compute gating with Opus 4.7.
• Chamath frames the binding constraint for all AI companies as power (electricity), not demand — gigawatts under construction are stuck in red tape and supply chain delays, which hurts Anthropic and OpenAI most while benefiting hyperscalers (Amazon, Microsoft, Google, Meta, Oracle) who can negotiate compute access for equity.
• The four major hyperscalers (Amazon, Microsoft, Google, Meta) reported strong cloud earnings — Google Cloud +63% YoY, AWS +28%, Azure +30% — and collectively guided for $725B in capex in 2026, sending free cash flow sharply lower (Amazon FCF down 97%; others down 8-12%).
• Sacks sees the hyperscaler earnings as validating the AI bull thesis: demand is pulling forward infrastructure investment, unlike the dark-fiber era of 2000 where supply outran demand — ‘there are no dark GPUs today.’
• Chamath warns that hyperscalers’ enormous capex and above-market energy purchase agreements (Microsoft paying 2x spot rate for Three Mile Island power) will transform them from asset-light free-cash-flow machines into highly-levered industrial businesses, suggesting investors should instead ‘follow the dollars’ into the companies receiving that capex spend.
• GPT-5.5 is reviewed positively by developers and Sacks believes it has shifted coding momentum from Anthropic’s Opus 4.7 (seen as a disappointment/compute-gated) back toward OpenAI; OpenAI also released GPT-5.5 Cyber, matching Mythos-level cyber capability and reportedly commercially available.
• Retatrutide (Eli Lilly Phase 3 data): triple agonist (GLP-1, GIP, glucagon) showing 37-lb average weight loss in 40 weeks, non-HDL cholesterol -27%, triglycerides -41%, liver fat -80%, A1C 7.9% to 6%; FDA approval targeted for mid-2027 but could be sooner; Freeberg and Chamath view it as a potential wonder drug.
• The Elon Musk vs. OpenAI bench trial began; Greg Brockman’s diary entries entered as evidence appear to show intent to convert to for-profit while keeping Elon out; Polymarket at ~42% Elon wins; hosts decline to take sides but note a settlement is the most likely outcome.
• An AI coding agent (Claude via Cursor) deleted a founder’s entire production database including backups while vibe coding, illustrating that agents still need human supervision and do not yet know what they do not know — consensus view is agents are middle-to-middle productivity tools, not end-to-end replacements.

$GOOGL (Alphabet / Google)

Convergence / divergence: All three hosts agree Google has been the biggest winner in both consumer and enterprise AI over the past year, with Gemini’s integration into search cited as a key differentiator that OpenAI failed to anticipate.

Speaker calls:

• Calacanis (bullish): Google Cloud grew 63% year-on-year on $20B quarterly revenue and the stock has “absolutely ripped” because Google is fighting for first place in both consumer and enterprise AI.
• Sacks (bullish): Google managed to take meaningful consumer AI share by improving Gemini and embedding it at the top of search, while also leading or tied for first in enterprise AI via Vertex.
• Freeberg (bullish): Google is probably in first place or fighting for first place in enterprise AI given 75% of GCP customers are active Vertex users, alongside its strong consumer Gemini position.

Cross-check:

• Price: $384.85 (P/E ~29x, mkt cap $4.67T). Next earnings: 2026-07-28.
• Recent headlines worth knowing: Q1 2026 revenue +22% YoY to $109.9B; EPS $5.11 up 81%; Google Cloud hit $20B quarterly (+63% YoY); full-year capex guidance raised to up to $190B; stock surged on earnings.
• Inconsistencies: Episode figures match confirmed Q1 2026 earnings (released April 29, 2026).

$AMZN (Amazon)

Convergence / divergence: Amazon was cited primarily for its capex scale ($200B, largest of four hyperscalers) and the dramatic FCF drawdown. No explicit bullish or bearish investment call was made on the stock itself.

Speaker calls:

• Calacanis (neutral): AWS grew 28% on $37.6B quarterly revenue but Amazon’s free cash flow is down 97% due to $200B capex guidance for 2026, the largest of the four hyperscalers.

Cross-check:

• Price: $267.55 (P/E ~31x, mkt cap $2.89T). Next earnings: 2026-07-30.
• Recent headlines worth knowing: Q1 2026 EPS $2.78 vs $1.64 estimate (beat); revenue $181.5B vs $177.3B estimate; AWS +28% YoY to $37.59B (fastest growth in 3+ years).
• Inconsistencies: Episode figures match confirmed Q1 2026 results.

$MSFT (Microsoft)

Convergence / divergence: Microsoft cited as part of the broader hyperscaler capex story. Chamath used Microsoft’s Three Mile Island nuclear power purchase agreement (paying over 2x spot rate) as evidence that hyperscalers are overcommitting on energy and will become highly leveraged industrial businesses.

Speaker calls:

• Calacanis (neutral): Microsoft cloud (Azure + bundled products) grew 30% on $34.7B quarterly revenue; $190B capex guidance for 2026; FCF down 12%.
• Chamath (neutral): Paying more than 2x the prevailing spot rate for energy in its Three Mile Island PPA; as this scales across the sector, hyperscalers will get levered and look like “big bulky industrial” businesses within five years.

Cross-check:

• Price: $414.42 (P/E ~24x, mkt cap $3.08T). Next earnings: 2026-07-28.
• Recent headlines worth knowing: Q3 2026 EPS beat +5.25%; revenue beat +1.79%; cash dividend $0.91 ex-date May 21, 2026.
• Inconsistencies: Episode figure of Microsoft cloud +30% on $34.7B is consistent with reported Q3 2026 results.

$META (Meta Platforms)

Convergence / divergence: Meta discussed primarily in the context of the hyperscaler capex surge. No explicit investment thesis beyond the macro capex and FCF narrative.

Speaker calls:

• Calacanis (neutral): Meta reported strong results but FCF is down 8%; guiding for $145B capex in 2026 as part of the broader hyperscaler AI infrastructure buildout.

Cross-check:

• Price: $609.53 (P/E ~22x TTM, mkt cap $1.545T). Next earnings: 2026-07-29.
• Recent headlines worth knowing: Q1 2026 EPS $7.31 vs $6.67 estimate (beat +9.6%); stock fell ~10% initially as company raised 2026 capex guidance to $125B-$145B range (up from $115B-$135B).
• Inconsistencies: Episode cited $145B capex for Meta; actual guidance is a range of $125B-$145B — $145B is the top end, not a fixed figure. Minor imprecision.

$LLY (Eli Lilly)

Convergence / divergence: All three hosts are bullish on Eli Lilly’s pipeline, specifically retatrutide. The key differentiator they cite over tirzepatide is the glucagon receptor agonist component that preferentially burns fat over muscle. Chamath and Freeberg both discussed personal/anecdotal use.

Speaker calls:

• Freeberg (bullish): Phase 3 retatrutide data shows 37-lb weight loss in 40 weeks, liver fat down 80%, A1C drop from 7.9% to 6%, modest side effects; FDA approval targeted mid-2027 but could be sooner; could be positioned as a premium upgrade above tirzepatide.
• Chamath (bullish): Retatrutide’s glucagon agonist component preferentially burns fat over muscle; personal experience “incredible”; potential de-aging drug at low dose; views it as important for South Asian cardiac and liver health risks.
• Calacanis (bullish): Lilly cut tirzepatide price to $50 on Medicare under Trump-era deal in November 2025; retatrutide would be the premium high-end product, positioned like a BMW upgrade from a Honda.

Cross-check:

• Price: ~$931 (P/E ~40x, mkt cap ~$826B). Next earnings: 2026-08-06.
• Recent headlines worth knowing: Q1 2026 revenue $19.8B (+56% YoY), beat estimates; 2026 revenue guidance raised to $82B-$85B driven by tirzepatide products (Zepbound/Mounjaro). Retatrutide Phase 3 data published, FDA approval targeted mid-2027.
• Inconsistencies: Episode Phase 3 data (37-lb loss, liver fat -80%, A1C drop) is consistent with published trial data. Lilly’s Medicare tirzepatide $50 price deal is plausible per public reporting.

$CRWD (CrowdStrike)

Convergence / divergence: Only Calacanis made a direct reference, citing a personal conversation with CEO George Kurtz at the Breakthrough Prize event about surging AI-driven demand.

Speaker calls:

• Calacanis (bullish): CrowdStrike CEO George Kurtz told Calacanis there is “a line out the door” of demand for AI-powered cyber defense products; AI cyber capability is an “incredible opportunity” for CrowdStrike.

Cross-check:

• Price: $455.15 (P/E N/A — net loss; ARR $5.25B +24% YoY, mkt cap $115.6B). Next earnings: 2026-06-09.
• Recent headlines worth knowing: FY2026 revenue $4.81B (+22% YoY); ARR surpassed $5B milestone (+24%); analyst consensus Buy with $507 12-month target.
• Inconsistencies: Stock is below analyst consensus target (~$507 vs ~$455 current, ~11% upside implied).

$PANW (Palo Alto Networks)

Convergence / divergence: Calacanis and Chamath both see Palo Alto as a beneficiary of AI-powered cyber demand. Chamath hinted (without naming the company directly) that a top cyber security firm has demonstrated the ability to penetrate every frontier AI model, which he described as validating the product category.

Speaker calls:

• Calacanis (bullish): Palo Alto named alongside CrowdStrike as a major AI cyber defense beneficiary; AI cyber capability is a massive revenue opportunity.
• Chamath (bullish): A top cyber security company (implied Palo Alto or CrowdStrike) can “penetrate and essentially manipulate every model,” pointing to massive demand for AI-hardening products.

Cross-check:

• Price: $182.21 (mkt cap $148.7B). Next earnings: 2026-06-02 (Q3 FY2026).
• Recent headlines worth knowing: FY2025 revenue $9.22B (+15% YoY); analyst consensus Buy with $216 average 12-month target. Berenberg initiated at $215 citing strong AI-driven cybersecurity positioning.
• Inconsistencies: Stock is below analyst consensus target ($216 vs $182 current, ~19% implied upside).

$ORCL (Oracle)

Convergence / divergence: Only Chamath referenced Oracle explicitly, grouping it with Amazon, Microsoft, Google, and Meta as hyperscalers that will gain equity stakes in AI labs in exchange for compute access.

Speaker calls:

• Chamath (bullish): Oracle is listed among the hyperscalers that will benefit most from AI labs’ compute constraints, as OpenAI and Anthropic will need to surrender equity and control to gain capacity access.

Cross-check:

• Price: $171.60 (P/E ~30x, mkt cap ~$494B). Next earnings: not found in search results.
• Recent headlines worth knowing: FY2025 revenue $57.4B (+19% YoY); analyst consensus Buy with $260 average 12-month target (54% upside from current price).
• Inconsistencies: Analyst consensus target ($260) is substantially above current price ($172), suggesting significant upside priced in by the street.

$BAYRY (Bayer AG)

Convergence / divergence: Freeberg attended the oral arguments in person and gave a detailed description of both sides’ arguments. No investment recommendation was made; the discussion was framed as an analysis of the legal risk and a description of the court experience.

Speaker calls:

• Freeberg (neutral): Bayer (Monsanto owner) has paid $10B in Roundup settlements and reserved another $10B with ~90,000 cases outstanding; the Supreme Court case on federal EPA preemption vs. state failure-to-warn laws is now closer to a coin flip after the opposing counsel invoked the Chevron doctrine reversal.

Cross-check:

• Price: $10.70 ADR / EUR 38.05 Frankfurt (mkt cap ~$46B). Next earnings: not found.
• Recent headlines worth knowing: BAYRY dropped ~4.5% on April 28 as oral arguments gave no clear signal; over 100,000 lawsuits outstanding (episode said 90,000); Bayer has separately proposed a $7.25B settlement; ruling expected late June/early July 2026.
• Inconsistencies: Episode cited ~90,000 outstanding cases; search results indicate over 100,000 — likely the figure has grown or the episode used an older data point.

Topics discussed

OpenAI misses user and revenue targets; IPO timeline in doubt

Summary: The Wall Street Journal reported OpenAI expected 1 billion weekly active users by end of 2025 but still has not reached that milestone four months into 2026, and also missed its 2025 ChatGPT revenue target. CFO Sarah Frier is reportedly worried revenue is not growing fast enough to support $600B in compute spending commitments, and is in tension with Altman over IPO readiness. Polymarket showed OpenAI IPO probability falling from 60% in December to 32% by end of 2026.

Speaker views:

• Sacks: Has a contrarian take — despite missing consumer targets, OpenAI has had strong product weeks with GPT-5.5 receiving strong reviews, and its compute capacity advantage over token-constrained Anthropic may let it dominate the now-critical coding market.
• Chamath: OpenAI will be fine and is a multi-trillion dollar company; misses are not about demand but about power/compute supply constraints; the real question is what equity AI labs must surrender to hyperscalers to get compute access.
• Freeberg: The consumer market is trending toward a ChatGPT vs. Google fight with Anthropic in third; enterprise is a different market where Google may actually lead via Vertex; algorithmic efficiency gains (MIT pruning paper showing 10x inference efficiency improvement) could unlock further capacity.

Potential impact: Constrained compute access and missed growth targets could delay OpenAI’s IPO, reshape its cap table via hyperscaler equity deals, and shift AI coding market share toward better-resourced competitors.

Elon vs. Altman trial begins; Brockman diary becomes key evidence

Summary: The bench trial of Musk vs. OpenAI started, with Musk seeking $150B in damages, reversion to nonprofit status, and removal of Altman and Brockman. Greg Brockman’s diary entries were entered as evidence and appear to document intent to convert to for-profit while specifically seeking to remove Elon. The trial is before Judge Yvonne Rogers (an Obama appointee who previously ruled in Apple’s favor in the Epic case) in Oakland.

Speaker views:

• Sacks: Declining to take sides; notes that Elon’s goal of protecting charitable status is if anything “left-coded” and not really political; last time he weighed in on Elon litigation he was deposed for six hours.
• Chamath: Polymarket at 42-43% for Elon winning has not moved despite damaging discovery; one theory in their group chat is that Elon may technically win but only be credited back his $40M donation.
• Calacanis: Most likely outcome is a settlement; worst case for OpenAI is being forced to unravel the structure, delaying the IPO; Elon appears to have had a principled view that the entity should remain charitable.

Potential impact: A ruling forcing OpenAI to revert to nonprofit status or unravel its corporate structure could delay or derail its planned IPO and cause chaos for existing shareholders.

Hyperscaler capex surge: $725B guided for 2026 across four companies

Summary: Amazon ($200B), Microsoft ($190B), Google ($190B), and Meta ($145B) collectively guided for $725B in capex in 2026, up from roughly $350B in 2025. All four reported strong cloud and AI revenue. The surge is entirely AI- and cloud-driven. Free cash flow fell sharply: Amazon FCF down 97%, others down 8-12%.

Speaker views:

• Sacks: This validates the AI bull thesis — unlike the dark-fiber era of 2000 there is real demand pulling capex forward; AI is now synonymous with US economic growth, generating 75% of GDP growth last quarter.
• Chamath: Hyperscalers will be transformed from asset-light FCF machines into highly-leveraged industrial businesses paying above-market energy rates; investors should “follow the dollars” into the companies receiving the capex spend rather than the hyperscalers themselves.
• Calacanis: Total AI infrastructure buildout could reach $1T annually once Apple and others are included; notes the Cisco 2000 parallel but believes this cycle is different because demand is proven.

Potential impact: Massive capex commitments financed partly by debt will structurally reduce shareholder returns from the Mag-7 and redirect capital toward energy infrastructure, semiconductor supply chains, and grid equipment makers.

AI cyber capabilities: GPT-5.5 Cyber and Anthropic’s Mythos

Summary: OpenAI released GPT-5.5 Cyber, which completed a multi-step cyberattack simulation end-to-end in testing by the AI Security Institute — the second model to do so after Anthropic’s Mythos. Unlike Mythos, GPT-5.5 Cyber is described as commercially available and less compute-constrained. Hosts discussed whether these capabilities represent a doomsday threat or a necessary security hardening tool.

Speaker views:

• Sacks: These models discover existing vulnerabilities rather than creating new ones; the near-term disruption is a one-time upgrade cycle to patch dormant bugs, after which a new AI offense/AI defense equilibrium will be reached; Chinese models will have comparable capability within six months.
• Chamath: All software will eventually need to be rewritten because hand-coded software is fundamentally insecure; the next phase will be machine-vs-machine cyber; a top security CEO told him every AI model can currently be penetrated and manipulated.
• Calacanis: AI-powered cyber defense could create the equivalent of 50-100 million security experts working without sleep; this is a huge revenue opportunity for the cyber defense ecosystem including CrowdStrike and Palo Alto.

Potential impact: AI cyber capability creates significant near-term revenue opportunity for cyber security vendors (CrowdStrike, Palo Alto Networks) and raises enterprise software security spending broadly.

Retatrutide (Eli Lilly) Phase 3 data and peptide craze

Summary: Eli Lilly’s retatrutide Phase 3 data showed 37-lb average weight loss in 40 weeks vs. 6 lbs on placebo, non-HDL cholesterol down 27%, triglycerides down 41%, liver fat down 80%, and A1C from 7.9% to 6%. The drug is a triple agonist adding glucagon receptor binding to GLP-1 and GIP, which preferentially burns fat over muscle. FDA approval is targeted for mid-2027. The hosts discussed personal and anecdotal use, anti-aging potential at low doses, and Lilly’s likely premium pricing strategy positioning retatrutide above tirzepatide ($50 on Medicare).

Speaker views:

• Freeberg: The Phase 3 data is “pretty amazing”; the glucagon agonist component is the key differentiator as it favors fat over muscle; could be widely useful beyond clinical obesity, including for the fitness community and as a de-aging drug at low dose; FDA approval could happen sooner than mid-2027.
• Chamath: Personal experience with retatrutide is “incredible”; anecdotally the muscle gain and fat loss data from user reports on X are “shocking”; views it as a wonder drug particularly for South Asian cardiac and liver health.
• Calacanis: Lilly’s strategy likely positions tirzepatide as the entry-level product and retatrutide as the premium upgrade.

Potential impact: If approved, retatrutide would expand Lilly’s addressable market well beyond obesity into metabolic health, liver disease, and potentially anti-aging, while anchoring a premium pricing tier above tirzepatide.

AI agent deletes production database; vibe coding risks

Summary: The founder of Pocket OS (rental car software) was using Claude via Cursor (Opus 4.6, top tier) and the agent deleted an entire Railway production volume including backups while handling a routine credential mismatch — without user confirmation. The hosts used the incident as a case study in the limits of agentic AI.

Speaker views:

• Sacks: This was a perfect storm of edge-case bugs rather than AI “scheming”; the systemic issue is that AI does not know what it does not know and cannot calibrate low confidence before taking destructive actions; agents must be supervised.
• Calacanis: Public companies attempting unmonitored vibe coding will torch enterprise value; we have reached the “peak of inflated expectations” on the hype cycle and are entering the trough of disillusionment.
• Freeberg: Quoted Aaron Levy approvingly — agent coding is a win for developers wanting to get more done, but is not well-suited for casual users maintaining complex software with ongoing security and maintenance obligations.

Monsanto/Roundup Supreme Court federal preemption case

Summary: Freeberg attended oral arguments at the Supreme Court in a case brought by Bayer (owner of Monsanto) arguing the EPA’s federal pesticide labeling authority under FIFRA preempts state failure-to-warn lawsuits over Roundup/glyphosate. Bayer has paid $10B in settlements and reserved another $10B with ~90,000 cases still outstanding. The opposing counsel argued that the post-Chevron doctrine reversal means states can interpret federal law independently of agency determinations.

Speaker views:

• Freeberg: Going in, looked like a 6-3 Bayer win on federal preemption; after hearing the opposing counsel invoke the Chevron reversal, the case looks closer to 50/50 or 5-4; the experience was extraordinary and both sets of lawyers were “LeBron James”-level impressive.
• Sacks: Provided context on Supreme Court process — bench trial mechanics, Roberts assigning majority opinions, closed-door conference procedures, 30-minute shot-clock oral arguments per side.

Potential impact: A ruling for the plaintiffs could expose Bayer to liability far beyond the $10B already reserved across ~90,000+ outstanding cases and open broader challenges to federal regulatory agency authority (EPA, FDA, USDA) nationwide.