The AI War: OpenAI Ads & Sora 2, Grok Partners With US Government & Google’s Ad Business is at Risk w/ Dave Blundin, Salim Ismail, & Alex Wissner-Gross | EP #198
Timestamps are as accurate as they can be but may be slightly off. We encourage you to listen to the full context.
Podcast Summary
This episode of Moonshots explores the explosive growth of AI-generated content across video, audio, and code creation. The hosts discuss major AI developments including Meta's Vibes app, OpenAI's Sora 2 video generation, and Anthropic's Claude Sonnet 4.5 for coding. (11:10) The conversation covers massive infrastructure investments, with Sam Altman requesting unprecedented compute capacity to avoid choosing between applications like curing cancer or providing education. (70:40) The episode also examines energy challenges, robotics expansion, and longevity breakthroughs, while addressing the cognitive biases that prevent experts from recognizing exponential growth patterns in technology adoption.
- Core themes include the transition from algorithmic content selection to algorithmic content generation, the emergence of sovereign AI strategies by governments, and the acceleration toward what the hosts call "singularity" or intelligence explosion
Speakers
Peter Diamandis
Founder of XPRIZE and Singularity University, bestselling author of multiple books including "The Future is Faster Than You Think." Currently running the Abundance Longevity Summit focused on extending human lifespan and has raised $157 million for longevity research through XPRIZE HealthSpan competition.
Alex Guzey
AI researcher and accelerationist who has published research on AI alignment and superintelligence. Known for his technical analysis of frontier AI models and his belief in extremely short timelines for artificial general intelligence, often referred to as the "accelerationista in the room."
Dave Blundin
Partner at Link Ventures, operating out of MIT and Harvard. Focuses on early-stage AI startups and requires founding teams to be "bona fide best friends." Previously involved with taking D-Wave quantum computing public and has significant experience in the Cambridge startup ecosystem.
Salim Ismail
Co-founder of Singularity University and expert in exponential organizations. Focuses on helping organizations achieve 10-100x improvements through exponential thinking and frequently speaks about cognitive biases that prevent recognition of exponential growth patterns.
Key Takeaways
Embrace AI-Native Development Over Feature Addition
Microsoft and Apple are making the critical error of adding AI as a feature to existing products rather than building AI-native solutions from the ground up. (50:52) As Salim noted, this approach fails because "It's not a feature, it's a brand new everything." Companies should view AI as a complete transformation of their business model, not an incremental enhancement. The most successful AI implementations come from startups built with AI-first architecture, while legacy companies struggle when they try to retrofit AI into existing systems.
Recognize and Combat Exponential Blindness
Expert predictions consistently underestimate exponential growth by orders of magnitude, as demonstrated by mobile phone adoption and solar energy deployment. (99:00) Industry experts predicted 16% growth for mobile phones while actual growth was 100%, and energy experts said solar could never get below $1/watt when it's now at 2 cents/watt. This cognitive bias affects decision-making across industries. To overcome this, focus on actual exponential curves rather than expert linear projections, and prepare for "orthogonal effects" where breakthroughs in one domain unexpectedly impact seemingly unrelated areas.
Invest in Sovereign Technology Capabilities
The U.S. government's strategic investment in Intel demonstrates the critical importance of maintaining domestic capabilities in essential technologies. (82:15) Dave revealed how spy chips embedded in Chinese-manufactured PC boards infiltrated U.S. data centers for years, highlighting national security risks. Countries cannot afford to outsource manufacturing of critical technologies like semiconductors that go into weapons systems and infrastructure. This extends beyond chips to robotics, where China is establishing dominance in emerging markets by being first to deploy cost-effective automation solutions.
Prepare for Compute-Constrained Future
Access to computational resources will determine competitive advantage as AI capabilities expand exponentially. (27:05) Business leaders who haven't reserved compute capacity will find themselves unable to access real-time AI capabilities when demand explodes. Sam Altman's request for 10 gigawatts of compute to avoid choosing between curing cancer and providing education illustrates the massive scale required. Organizations should secure compute resources now, as the demand for AI-powered applications will far exceed supply, creating a fundamental scarcity that will stratify access to advanced capabilities.
Leverage AI for Economic Transformation
AI's ability to pass advanced professional exams like the CFA Level 3 signals the beginning of widespread automation of knowledge work. (61:05) This creates an opportunity to eliminate the circular complexity that characterizes much of the service economy - complex regulations that require armies of lawyers and accountants to navigate. Rather than viewing this as job destruction, it represents a chance to redirect human talent from managing artificial complexity toward creating genuine value for humanity. The key is recognizing that AI will level the playing field, giving everyone access to expert-level advice and analysis.
Statistics & Facts
- OpenAI's Stargate project aims for a $500 billion investment to achieve 10 gigawatts of AI compute capacity, dwarfing Microsoft's $40 billion, Amazon's $16 billion, Google's $29 billion, and Meta's $23 billion investments in 2024. (73:24)
- Global solar capacity increased from 40 gigawatts in 2010 to nearly 3 terawatts in 2025 - a 75x increase over 15 years, while industry experts consistently predicted linear growth. (99:00)
- 32 gigabytes of DDR5 RAM costs $101 today versus an estimated $150 million in the 1980s, representing a price decrease of over 99.9%. Similarly, a 4-terabyte hard drive costs $84 today versus an estimated $1.26 billion in the 1980s. (91:55)
