The All-In podcast team broadcasts from the World Economic Forum in Davos, conducting interviews with three technology leaders shaping the AI and crypto landscape. Brian Armstrong, CEO of Coinbase, discusses regulatory wins under the Trump administration and the explosive growth of stablecoin payments. Andrew Feldman, CEO of Cerebrus Systems, explains how his wafer-scale AI chips are revolutionizing inference speed and securing major partnerships with OpenAI. Jake Lucerarian, CEO of Gecko Robotics, details how purpose-built inspection robots are accelerating defense manufacturing and energy infrastructure maintenance.

The conversations reveal a shift at Davos from ESG and DEI focus toward practical business applications and ROI-driven technology deployment. Armstrong emphasizes the importance of clear crypto regulations and partnerships with traditional banks, while Feldman addresses the critical infrastructure challenges of powering AI at scale. Lucerarian advocates for robotics applications in high-value industrial sectors rather than consumer applications like laundry folding.

Crypto Regulation Breakthrough Under Trump Administration

The Genius Act passed into law requiring U.S. regulated stablecoins to hold 100% reserves in short-term U.S. treasuries, with maximum 30-day duration providing unprecedented safety standards.

"The Biden administration really tried to unlawfully kill this industry in America from my point of view. Donald J. Trump campaigned on making the United States the crypto capital of the world" - Brian Armstrong on the regulatory transformation.

Bank trade groups are attempting to undo the Genius Act despite its recent passage, which Armstrong calls "a red line" for the crypto industry.

Five of the top 20 global banks now use Coinbase infrastructure, with JPMorgan and PNC Bank publicly disclosed as partners building crypto products.

Stablecoin Payments Emerge as Crypto's Killer App

Cross-border B2B payments became the fastest-growing crypto application, with Coinbase Business serving small and medium companies for international invoicing and accounting.

Stablecoin rewards programs differ legally from interest payments by requiring customer activity like trading or Coinbase One subscriptions, not just balance holdings.

USDC remains the largest regulated stablecoin under Genius Act compliance, while Tether operates a separate compliance framework and may split into U.S. and international versions.

"Something like 500 million people have used crypto globally. Bitcoin was the best performing asset class of the last decade" - Armstrong on crypto's mainstream adoption.

AI Chip Wars: Wafer-Scale Engines vs Traditional GPUs

Cerebrus's wafer-scale engine contains 4 trillion transistors and measures 56 times larger than NVIDIA's B200, solving a 75-year-old problem of building chips this large.

OpenAI partnership involves 750 megawatts of computing power delivered over several years, marking the industry's shift from measuring chip sales by units to power consumption.

"GPUs have a lot of memory capacity, but it's really slow. That's a huge bottleneck in inference. It's why NVIDIA spent $20 billion buying Grok" - Feldman on GPU limitations.

Deep research queries create cascading computational demands with 20 queries spawning 20 sub-queries each, requiring 10-15 seconds per query versus Cerebrus's 4-10 second total completion.

Industrial Robotics Drives Real-World AI Applications

Gecko Robotics achieved 90% manufacturing speed improvements for submarine and destroyer production using robotic inspection technology, according to Admiral Houston.

"Folding laundry and cleaning dishes is not a high ROI use case. You're not going to pay $40,000 for $20 an hour work" - Lucerarian on consumer robotics versus industrial applications.

Defense applications represent 30% of Gecko's business, focusing on weld quality inspection and manufacturing acceleration for 100-year-old forge infrastructure.

Energy sector growth driven by hyperscaler infrastructure demands, with robots collecting proprietary datasets unavailable on the internet for training specialized AI models.

AI Infrastructure Bottlenecks and Power Challenges

Memory shortage expected to persist 18 months due to supply chain confusion, with customers placing 18-month orders instead of typical 6-month demand forecasts.

Data center power requirements now measured in megawatts rather than square footage, with cheapest sources being hydro power followed by natural gas including former flare-off gas.

Microsoft announced guarantees that data center energy usage won't increase local utility costs, addressing community concerns about power grid strain.

"We chose as a nation not to invest in our grid for 40 or 50 years. Our grid is decrepit compared to other advanced nations, particularly China" - Feldman on infrastructure challenges.

Geopolitical AI Competition and Trade Policy

China leads in open-source AI models while U.S. maintains chip manufacturing advantage with six of the world's top 10 chip teams concentrated in Santa Clara.

Previous administration's export controls kept chips from allies like UAE and Saudi Arabia, while new administration focuses on empowering allies and encouraging investment.

"The way you get good at building high-speed chips is to build high-speed chips. You play the game, you get better at the game" - Feldman on maintaining technological leadership.

Canada's recent alliance with China highlights risks of inconsistent U.S. trade policy pushing allies toward alternative partnerships for raw material exports.