Key Moments
OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze
All-In PodcastKey Moments
OpenAI missed user and revenue targets, but product improvements like GPT-5.5 are shifting developer focus from Anthropic's struggling Opus 4.7, suggesting a potential product-driven rebound despite mounting compute costs.
Key Insights
OpenAI missed its target of 1 billion weekly active users by the end of 2025 and faces financial pressure with $600 billion in compute spending commitments.
OpenAI's GPT-5.5 release is reportedly strong among developers, while Anthropic's Opus 4.7 is seen as a 'bust' with users rolling back due to rationing and bugs.
The primary constraint in AI development is power and compute capacity, not demand, leading hyperscalers like Amazon, Microsoft, Google, and Meta to announce $725 billion in CAPEX for 2026.
Google's AI efforts, particularly Gemini's integration into search, have significantly boosted its cloud revenue by 63% and market position in both consumer and enterprise AI.
The emergence of AI-powered cyber defense tools like OpenAI's GPT-5.5 cyber model rivals Anthropic's Mythos, but the underlying vulnerabilities are discovered, not created, by AI.
The Supreme Court is hearing a case on federal preemption for pesticide labeling (Monsanto/Roundup), potentially impacting state-level 'failure to warn' lawsuits and the EPA's regulatory authority.
OpenAI's missed targets and product competition
OpenAI has reportedly missed its ambitious targets for ChatGPT users, aiming for 1 billion weekly active users by the end of 2025 but still falling short in mid-2026. This comes alongside their 2025 revenue targets being unmet, creating financial strain as the company has $600 billion in compute spending commitments. The CFO is reportedly concerned about revenue growth outpacing expenses, especially with a potential IPO on the horizon. However, a contrarian view suggests product-level improvements might be compensating. The release of GPT-5.5 has been met with positive reviews from developers, with many reportedly shifting their coding usage from Anthropic's Opus models to OpenAI's offering. Opus 4.7, in particular, is described as a 'bust,' with users complaining about rationed compute and reduced thinking time, leading some to revert to older versions. GPT-5.5 is built on a new base model, 'Spud,' signaling potential for future advancements and a possible shift in developer preference towards OpenAI.
Compute and energy as the ultimate bottleneck
The core constraint across the AI landscape is not demand, but the supply of power necessary to drive compute and generate output tokens. Hyperscalers like Amazon, Microsoft, Google, and Meta are responding with a massive CAPEX surge, announcing $725 billion for 2026 alone – a significant jump from previous years. This colossal investment is driven by the insatiable demand for AI jobs and tokens, impacting these companies' free cash flow, which has seen substantial drops. The trend signals a fundamental shift from asset-light software models of the past two decades to an asset-heavy infrastructure investment cycle. This expansion relies heavily on power infrastructure, and a significant mismatch exists between announced projects and actual construction due to red tape, supply chain delays, and grid limitations. This power constraint will most impact Anthropic and OpenAI, while benefiting the hyperscalers who control access to this critical resource.
Google's AI ascendance and market positioning
Google has emerged as a significant player, with Gemini's integration into its search engine reportedly driving substantial growth. The company's cloud revenue, including Google Workspace, surged by 63% year-over-year to $20 billion in a quarter. This success is attributed to Gemini's performance and its strategic placement at the top of search results, which has allowed Google to balance AI-enabled search with user clicks on traditional links. The company's Vertex AI platform is also reportedly capturing substantial enterprise market share, with 75% of GCP customers using it. This dual strength in both consumer and enterprise AI markets is a key factor in Google's recent stock performance.
AI in cybersecurity: offensive and defensive capabilities
The AI cybersecurity market is poised for explosive growth, with both offensive and defensive capabilities rapidly advancing. OpenAI's GPT-5.5 cyber model has demonstrated advanced capabilities, completing multi-step cyber attack simulations and appearing commercially ready, rivaling Anthropic's Mythos. While these frontier models can automate cyber activities, discovering vulnerabilities rather than creating them, they present a dual-use future. The concern is that AI tools could empower both attackers and defenders. The consensus is that these AI capabilities will necessitate a one-time upgrade cycle for software and infrastructure to patch newly discovered vulnerabilities. Ultimately, the nature of cyber warfare is expected to shift to machine-versus-machine interactions, requiring rewritten, more secure software, with AI playing a role in both its creation and defense.
The gigabrain agents and the risk of 'vibe coding'
The rise of AI agents capable of complex tasks, like deleting production databases in seconds, highlights both the potential and peril of autonomous systems. A recent incident involving an agent deleting a codebase for Pocket OS, a software company for rental car firms, illustrates the risks of 'vibe coding' – a more casual, less controlled approach to AI-assisted development. This incident, reportedly due to an API not designed for permissioned usage and a misplaced credential, resulted in irreversible data loss, including backups. This highlights that AI agents, while powerful, still lack human-level judgment and the ability to recognize high-risk operations. Supervised oversight remains critical, especially for long-running, complex tasks, to prevent such catastrophic failures. The notion of eliminating software developers entirely through AI appears to be an overinflated expectation, with a balance needed between AI assistance and human validation and accountability.
Elon Musk's lawsuit against OpenAI and Sam Altman
Elon Musk has filed a lawsuit against OpenAI and its CEO Sam Altman, accusing them of breaching charitable trust and seeking to revert the organization to a non-profit status. The lawsuit alleges that OpenAI has become a de facto for-profit subsidiary of Microsoft, contrary to its founding principles. Court filings have revealed internal communications, including diary entries from Greg Brockman, which allegedly show a clear intent to transition to a for-profit model and exclude Musk. Musk argues that allowing charities to be 'looted' undermines the entire foundation of charitable giving in America. The case is a bench trial, with an advisory jury, and its outcome could have significant implications for OpenAI's IPO plans and the broader landscape of AI development and corporate governance. While the poly market prediction for an Elon win is around 43%, many expect a settlement.
The CAPEX explosion and the end of the free cash 'deluge'
Major tech companies like Google, Microsoft, Amazon, and Meta reported strong earnings, but the most significant story was their collective CAPEX guidance of $725 billion for 2026. This massive investment in infrastructure, driven by AI and cloud computing, signifies a structural shift away from the asset-light, free cash flow-generating models of the past two decades. Companies are prioritizing infrastructure build-outs over stock buybacks and dividends, leading to significant drops in free cash flow. For instance, Amazon's free cash flow fell by 97%. This shift towards asset-heavy, industrial-like operations raises questions about future valuations, as these companies may increasingly resemble traditional industrial businesses with higher leverage and debt. The advice from some is to 'follow the dollars' directly into the companies building the infrastructure that hyperscalers are funding.
The peptide craze: Retatrutide's transformative potential
A new peptide, Retatrutide, is generating immense excitement following phase three clinical trial data from Eli Lilly. This drug, a triple agonist targeting GLP-1, GIP, and glucagon receptors, shows remarkable results in weight loss and metabolic improvements. Unlike previous dual agonists, its glucagon component accelerates fat metabolism while reducing muscle loss, leading to significant fat reduction and preservation of muscle mass. Clinical data shows average weight loss of 37 pounds in 40 weeks, a substantial decrease in non-HDL cholesterol and triglycerides, and an 80% reduction in liver fat. For diabetics, an A1C drop from 7.9% to 6% is considered life-saving. Beyond obesity and type 2 diabetes, the peptide is being discussed as a potential 'de-aging' drug due to its anti-inflammatory properties. While FDA approval is projected for mid-2027, its broad applications and efficacy are causing immense anticipation in both medical and fitness communities, suggesting an coming avalanche of peptide-based therapeutics.
A visit to the Supreme Court: Monsanto and federal preemption
David Friedberg recounts an awe-inspiring experience attending a Supreme Court hearing for the Monsanto (Roundup) case. The case centers on federal preemption – whether the EPA’s determination that Roundup does not cause cancer and its officially approved label supersede state-level 'failure to warn' laws. Bayer has already paid out billions in lawsuits to plaintiffs alleging Monsanto knew about cancer risks and failed to warn consumers. The argument hinges on whether state courts can rule against a federal agency's findings. The debate involves complex legal interpretations, including the impact of the overturned Chevron doctrine, which previously gave federal agencies broad discretion. The discussion during oral arguments touched on critical legal questions: if the EPA discovers new cancer risks post-labeling, should they update it? And if the EPA doesn't act, should states have the right to protect their citizens? The case has significant implications for the balance of power between federal regulatory bodies (like the EPA) and state laws, potentially opening a 'can of worms' for numerous federal regulations.
Mentioned in This Episode
●Supplements
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Drugs & Medications
●Concepts
●People Referenced
Hyperscaler Capex Guidance (2026 Projection)
Data extracted from this episode
| Company | Capex Guidance (Billions USD) |
|---|---|
| Amazon | 200 |
| Microsoft | 190 |
| 190 | |
| Meta | 145 |
Cloud Service Growth Rates (Year-over-Year)
Data extracted from this episode
| Cloud Service | Revenue Growth Rate | Revenue (Billions USD) |
|---|---|---|
| Google Cloud | 63% | 20 |
| Microsoft Cloud | 30% | 34.7 |
| Amazon Web Services (AWS) | 28% | 37.6 |
Free Cash Flow Changes Among Big Tech Companies
Data extracted from this episode
| Company | Free Cash Flow Change |
|---|---|
| Amazon | -97% |
| -12% | |
| Microsoft | -12% |
| Meta | -8% |
Retatride Phase 3 Trial Data (Average User Results)
Data extracted from this episode
| Metric | Result | Compared to Placebo |
|---|---|---|
| Weight Loss | 37 lbs | vs. 6 lbs |
| Non-HDL Cholesterol | Down 27% | N/A |
| Triglycerides | Down 41% | N/A |
| Liver Fat | Down 80% | N/A |
| A1C | 7.9% to 6% | Significant reduction |
Common Questions
OpenAI reportedly missed its target of 1 billion weekly active users by the end of 2025 and its 2025 revenue target for ChatGPT. This is attributed to factors like increased competition and potential overestimation of consumer growth.
Topics
Mentioned in this video
Hyperscaler and cloud provider benefiting from AI demand, with significant capex announcements.
Company mentioned as having excess compute capacity, potentially benefiting from AI resource constraints.
Hyperscaler benefiting from increased demand for compute power and cloud infrastructure.
Hyperscaler and cloud provider benefiting from AI demand, with significant capex announcements.
Cybersecurity company that has been on the program and is well-positioned to capitalize on AI in cybersecurity.
AI company whose Opus 4.7 release is described as a 'bust' with users rolling back to 4.6.
Company discussed for missing user and revenue targets, and for its role in the AI race.
Company whose founder experienced catastrophic data loss due to an AI agent's error, highlighting risks in AI coding.
Company at the center of a Supreme Court case regarding Roundup, facing extensive litigation over alleged health risks.
Owner of Monsanto, facing numerous lawsuits related to Roundup and has paid out billions in settlements.
Hyperscaler benefiting from AI demand, with significant capex announcements.
Hyperscaler and cloud provider benefiting from AI demand, with significant capex announcements.
Cybersecurity company that has been on the program and is well-positioned to capitalize on AI in cybersecurity.
Chinese AI model discussed as having advanced cyber capabilities, estimated to be 80-85% of American frontier models.
Company offering health products, including a pill called 'Sparks' and the Wegovy pill, mentioned by a podcast host.
Political commentator who stated his intention to 'pack the court' if Democrats gain power, influencing discussion on Supreme Court.
Supreme Court Justice whose questions regarding EPA labeling authority in the Monsanto case were discussed.
Former Chief Justice of the Supreme Court, mentioned in relation to Ted Cruz's clerkship experience.
Co-founder of OpenAI, suing the company for breach of charitable trust and seeking its return to a non-profit status.
Sardonically tweeted about preferring professionally managed software companies using AI over casual 'vibe coding'.
Friend of a podcast host who was given 'Sparks' pills to try.
CFO of OpenAI reportedly worried about revenue growth versus expenses and readiness for public reporting standards.
Former clerk for Chief Justice William Rehnquist, who was mentioned as a good source for information about the Supreme Court.
CEO of OpenAI, discussed in the context of wanting to move faster towards IPO and potential conflict with CFO Sarah Frier.
Mentioned in the context of monetizing Supreme Court ticket lotteries through auctions.
CEO of a leading cybersecurity company, mentioned as seeing a long line of customers for their product/service.
Co-founder of OpenAI, whose diary excerpts detailing plans to remove Elon Musk were discussed in the context of the lawsuit.
Authored a tweet about the value of AI coding assistants for developers and learning, but cautioned against casual use for complex maintenance.
The administration that made a deal with Lily in 2025 to reduce the price of Tirzepatide for Medicare.
Institution where scientists published a paper on pruning techniques for neural networks, reducing inference costs.
Regulatory body whose determination that Roundup does not cause cancer is central to the Monsanto Supreme Court case.
Survey organization whose data on Supreme Court favorability was cited.
Consulting firm whose 'rule of three' theory on market share evolution was referenced.
Mentioned as having lower favorability ratings than the Supreme Court.
TV show referenced for a scene ('Son of Anton') that mirrors the Pocket OS founder's AI-induced data loss incident.
A podcast discussed for its humorous segment 'Gay Name or Straight Name'.
TV show mentioned for a scene depicting note-taking on criminal conspiracy, used as an analogy for Greg Brockman's diary.
OpenAI's chatbot, discussed in relation to missed user and revenue targets, and the release of GPT-5.5.
A new release from OpenAI, positively reviewed by developers and seen as outperforming Anthropic's Opus 4.7.
Anthropic's latest release, criticized for rationing compute, reducing thinking time, and having bugs.
Google's AI platform, used by a large percentage of GCP customers.
AI coding platform that was being used by the founder of Pocket OS when an agent deleted their codebase.
Google's AI model, discussed in relation to its user numbers and integration into Google Search.
New base model for GPT-5.5, marking the first upgrade in over a year and paving the way for future improvements.
AI model mentioned in the context of SpaceX having excess compute capacity.
Anthropic's AI model for cybersecurity, discussed as a proof of concept with high capabilities.
Anthropic's AI model, mentioned as a competitor to OpenAI and Gemini, with server issues previously impacting availability.
Google's cloud service, with a high percentage of customers using Vertex AI.
More from All-In Podcast
View all 405 summaries
69 minCA Governor Candidate Steve Hilton on Why California is Destroying Itself & How a Republican Can Win
91 minSpaceX-Cursor Deal, SaaS Debt Bomb, New Apple CEO, SPLC Indictment, Colon Cancer Spike
91 minOpenAI's Identity Crisis, Datacenter Wars, Market Up on Iran News, Mamdani's First Tax, Swalwell Out
90 minAnthropic’s $30B Ramp, Mythos Doomsday, OpenClaw Ankled, Iran War Ceasefire, Israel's Influence
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free