At Davos, industry leaders discuss the convergence of crypto regulation, AI hardware scaling, and industrial robotics under a shifting political landscape. Key themes include the integration of AI agents with crypto wallets, the transition of data center metrics from space to megawatts, and the necessity of robots to digitize the physical world for AI training.
Overview
Broadcasting from the World Economic Forum in Davos, the All-In interview series captures a distinct shift in global business sentiment from ESG-focused idealism to pragmatic 'brass tacks' deal-making. Brian Armstrong (Coinbase) highlights a regulatory pivot in the US, predicting that the new administration and the 'Genius Act' will legitimize stablecoins and enable AI agents to utilize crypto wallets for autonomous payments. Andrew Feldman (Cerebras) argues that the AI revolution is now constrained by power availability rather than silicon, detailing a massive 750MW infrastructure deal with OpenAI and the critical role of latency reduction in unlocking new AI behaviors. Finally, Jake Loosararian (Gecko Robotics) emphasizes the gap in physical world data, explaining how purpose-built robots are essential for digitizing critical infrastructure to create 'world models' for industry. Together, these interviews paint a picture of deep tech maturing into the real economy, driven by regulatory clarity and physical-digital convergence.
Key Points
Regulatory Pivot in Crypto: Brian Armstrong describes a stark contrast between the Biden administration's perceived hostility toward crypto and the incoming administration's embrace of the sector. He notes that major US banks are now integrating crypto infrastructure, and legislation like the 'Genius Act' provides a pathway for regulated stablecoins backed by US Treasuries. Why it matters: Clear regulation allows institutional capital and traditional banks (JP Morgan, PNC) to formally enter the crypto market, stabilizing the industry. Evidence: Donald J. Trump, you got to give him credit. I mean, he campaigned on this idea of making the United States the crypto capital of the world. He's kept his promises.
AI Agents as Economic Actors: A convergence between AI and crypto is emerging where AI agents, requiring the ability to transact, will utilize stablecoins and crypto wallets. This bypasses the friction of traditional 'Know Your Customer' (KYC) banking systems that are designed for humans, creating a new B2B economy run by software. Why it matters: This represents a massive new user base for blockchain technology, moving beyond human speculation to automated, high-frequency utility. Evidence: AI agents need to get work done and they have to do payments... So AI agents I think are going to use stable coins and wallet crypto wallets.
Latency Changes Function, Not Just Speed: Andrew Feldman argues that reducing AI inference latency isn't just about waiting less; it fundamentally alters user behavior and application viability. He draws a parallel to the shift from dial-up to broadband, which didn't just make downloading faster but enabled entirely new business models like streaming. Why it matters: Real-time AI (coding assistants, complex reasoning) requires near-zero latency to maintain user 'flow,' necessitating new hardware architectures like Wafer Scale Engines. Evidence: When the internet got fast, Netflix didn't get better at delivering DVDs. Netflix became a movie studio.
Power is the New Data Center Metric: The limiting constraint for AI infrastructure has shifted from square footage to power availability. Deals are now negotiated in megawatts (e.g., Cerebras' 750MW deal with OpenAI) rather than rack units, pushing data centers toward energy-rich locations like Texas and the Middle East. Why it matters: This shift effectively ties the future of AI progress to energy policy and grid modernization, making energy abundance a prerequisite for tech dominance. Evidence: We used to talk about data centers in terms of square footage... And now nobody cares how many square feet they you have. They care about how much power you have.
The Industrial Data Gap: While LLMs have scraped the internet for text, the physical industrial world (manufacturing, defense, infrastructure) lacks digital data. Gecko Robotics focuses on using robots to inspect the 'built world' to create the datasets necessary to train AI models for physical industries. Why it matters: AI cannot optimize the physical economy (manufacturing, repair) without first digitizing it via robotics, making data collection the primary value driver for industrial robotics. Evidence: There is not a corpus of information in data set and so I'm focused on that... collecting it by solving for you know important problems on critical infrastructure.
Sections
Strategic Insights
Meta-level observations on the convergence of technology and policy.
The convergence of AI and Crypto is inevitable because traditional banking requires human identity (KYC), whereas crypto wallets can be programmed for autonomous agents. This creates a 'machine economy' layer.
A cultural shift at Davos from ESG/DEI performative discussions to 'brass tacks' deal-making suggests a global return to pragmatic capitalism and industrial policy, likely driven by geopolitical competition.
The primary bottleneck in technology has shifted from 'Bits' (software/chips) to 'Atoms' (power generation/grid capacity), making energy policy the most critical component of AI strategy.
Technical Specifications & Data
Specific metrics and hardware details mentioned in the interviews.
Cerebras Wafer Scale Engine: A single chip 56x larger than an Nvidia B200, containing 4 trillion transistors, designed to eliminate interconnect latency.
OpenAI Infrastructure Deal: A 750 megawatt power agreement, signaling the scale of compute required for the next generation of models.
'Genius Act' (Stablecoin Regulation): Requires US regulated stablecoins to hold 100% reserves in short-term (max 30-day) US Treasuries.
Gecko Robotics: Approximately 30% of business is now defense/Department of War; utilizes sensor fusion on robots to inspect welds and infrastructure health.
Memorable Quotes
Verbatim excerpts capturing the essence of the conversations.
Donald J. Trump, you got to give him credit. I mean, he campaigned on this idea of making the United States the crypto capital of the world. He's kept his promises.
When the internet got fast, Netflix didn't get better at delivering DVDs. Netflix became a movie studio.
We don't build the kind of technologies that started in Silicon Valley. And so this world of energy of metal manufacturing of of mining all these like sectors defense like they're they're just not... in the bubble.
The role of middle management, which was frequently to move information... That job has shrunk in value.