OpenAI Misses Targets, Codex vs Claude, Elon vs Sam Trial, Big Hyperscaler Beats, Peptide Craze
Competition drives excellence in AI development. While OpenAI missed consumer growth targets (aiming for 1 billion weekly users but stalling at 900 million), their GPT 5.5 release and enterprise focus show strategic adaptation. The real constraint isn't demand—it's power and compute capacity. Compan
1h 20mKey Takeaway
Competition drives excellence in AI development. While OpenAI missed consumer growth targets (aiming for 1 billion weekly users but stalling at 900 million), their GPT 5.5 release and enterprise focus show strategic adaptation. The real constraint isn't demand—it's power and compute capacity. Companies are negotiating equity for access to data centers, creating opportunities for those with excess capacity like SpaceX's Grock. Focus on solving bottlenecks, not just chasing headlines.
Episode Overview
The All-In podcast hosts discuss OpenAI's missed revenue and user targets, the competitive AI landscape between OpenAI, Anthropic, and Google, and the critical role of compute capacity and energy in AI development. They also explore the implications of AI-powered cybersecurity tools and preview the Elon Musk vs. Sam Altman legal battle over OpenAI's nonprofit-to-profit transformation.
Key Insights
Power Constraints Are the Real Bottleneck in AI
The limiting factor in AI isn't demand—it's access to power and compute capacity. OpenAI and Anthropic are power-constrained, forcing them to ration compute and potentially miss growth targets. Meanwhile, hyperscalers like Oracle, Amazon, Meta, Microsoft, and Google have advantages. This creates opportunities for companies with excess capacity, like SpaceX with Grock, to capture market share through strategic partnerships.
AI Cybersecurity Will Create a One-Time Upgrade Cycle
AI-powered cyber tools like OpenAI's GPT 5.5 Cyber and Anthropic's Mythos can automate vulnerability discovery at unprecedented scale. While this creates short-term risk as both attackers and defenders gain access to these capabilities, it will ultimately lead to hardened infrastructure. The key is ensuring white hats (defenders) use these tools first to patch vulnerabilities before black hats exploit them. A new equilibrium between AI-powered offense and defense will emerge after this transition.
Model Pruning Can Reduce AI Inference Costs by 90%
MIT research shows neural networks can be pruned by 90% while maintaining accuracy, reducing inference costs by 10x. Instead of firing up massive models for simple queries, smaller pruned models can handle common requests (weather, stock prices, basic coding). This approach—dynamically selecting appropriately-sized models—can deliver 10x more output per energy unit, addressing the power constraint challenge.
The AI Market Is Evolving Toward a 4:2:1 Market Share Structure
Following BCG's 'rule of three,' mature competitive markets typically stabilize with three players holding a 4:2:1 market share ratio. In consumer AI, OpenAI (900M weekly users) and Google Gemini (700M-1B users) are fighting for first and second place, with Anthropic's Claude in third. The enterprise market is separate, with Google Vertex claiming 75% of GCP customers, positioning Google strongly in both segments.
OpenAI's Compute Advantage May Compensate for Consumer Weakness
OpenAI's massive data center commitments ($600B in spending) were based on consumer growth projections that didn't materialize. However, their compute capacity advantage over token-constrained competitors like Anthropic positions them well for the exploding enterprise and coding markets. Sam Altman may end up being 'right for the wrong reason'—missing on consumer but winning on enterprise where demand is surging.
Notable Quotes
"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."
"If we make it okay to loot a charity, the entire foundation of charitable giving in America will be destroyed. That's my concern."
"The bugs were already in the code. They were sitting there waiting for some hacker to discover. If we can now use AI to find these bugs in advance, these vulnerabilities and patch them, then you actually harden our infrastructure and and you harden our security."
"I think Sam may end up being right here for the wrong reason, which is he missed on consumer, but enterprise is going gang busters and is giving him the ability now, I think, to catch up on code."
Action Items
-
1
Prioritize Compute Access Over Perfect Product-Market Fit
If you're building AI products, securing reliable compute capacity is more critical than fine-tuning features. Partner with providers who have excess capacity or negotiate equity-for-compute deals. The bottleneck is infrastructure, not customer demand.
-
2
Implement AI-Powered Cyber Defense Immediately
Organizations should deploy AI cybersecurity tools from vendors like CrowdStrike or Palo Alto Networks now to discover and patch vulnerabilities before attackers gain access to similar capabilities. This one-time upgrade cycle is urgent—waiting puts you at risk.
-
3
Explore Model Pruning for Cost Efficiency
If you're running AI inference at scale, investigate pruning techniques to reduce model size by up to 90% without sacrificing accuracy. Create a suite of smaller models for common queries to reduce costs by 10x and extend your compute budget.
-
4
Position for the Enterprise AI Opportunity
The enterprise AI market is growing faster than consumer. If you're a developer or founder, focus on vertical solutions for businesses rather than general consumer tools. Companies have urgent needs and budgets to match—demand far exceeds supply.