All-In with Chamath, Jason, Sacks & Friedberg

Coinbase CEO Brian Armstrong Breaks Down the Three Biggest Trends in Crypto + More from Davos!

January 23, 2026

Key Takeaways Copied to clipboard!

  • Coinbase CEO Brian Armstrong highlighted that the Biden administration's regulatory approach was perceived as trying to unlawfully kill the crypto industry, contrasting sharply with Donald J. Trump's stated commitment to making the US the crypto capital through clear regulation. 
  • The recently passed 'Genius Act' mandates that U.S. regulated stablecoins must hold 100% of their assets in short-term U.S. treasuries, a rule that banking trade groups are reportedly attempting to undo. 
  • Brian Armstrong identified the three biggest trends in crypto as: all assets coming on-chain for trading, the growth of prediction markets, and the expansion of stablecoin payments, particularly in B2B cross-border transactions. 
  • Hyperscalers initially mishandled rural data center build-outs by failing to communicate effectively with communities, leading to public backlash over increased power rates, a mistake Microsoft is attempting to correct with guarantees not to raise utility costs. 
  • The current massive demand for AI compute is still in its early stages, driven by increasing model quality, usage frequency, and the need for faster memory access, which is currently causing a significant HBM (High Bandwidth Memory) shortage. 
  • Geopolitically, the US leads in high-speed chip design talent concentrated in Santa Clara, while China has advanced in the open-source model category, necessitating a strategic approach to empowering allies and maintaining technological standards globally. 

Segments

Davos Context and Bank Partnerships
Copied to clipboard!
(00:00:41)
  • Key Takeaway: Coinbase is actively engaging global regulators at Davos to push for crypto market structure legislation.
  • Summary: Coinbase is meeting with world leaders to discuss updating financial systems with crypto, noting that five of the top 20 global banks are integrating their crypto infrastructure. Specific public partnerships include JPMorgan and PNC Bank, alongside powering integrations for BlackRock’s goal to tokenize all their funds.
Regulatory Climate Shift
Copied to clipboard!
(00:02:10)
  • Key Takeaway: The regulatory environment has significantly improved for crypto under the Trump administration compared to the Biden administration’s perceived hostility.
  • Summary: Brian Armstrong credits Donald J. Trump with taking crypto regulation seriously and campaigning on making the U.S. the crypto capital, keeping promises to create clear rules. This shift is seen as crucial for American companies to thrive and for global competitiveness against nations like China regarding digital currency.
Stablecoin Legislation and Bank Competition
Copied to clipboard!
(00:04:28)
  • Key Takeaway: The ‘Genius Act’ mandates that U.S. regulated stablecoins must hold 100% reserves in short-term U.S. treasuries, fundamentally differing from banks’ fractional reserve lending models.
  • Summary: Bank CEOs view crypto as existential, fearing disruption similar to how the internet challenged traditional media, though many are now leaning into integration. Bank trade groups are reportedly trying to undo the recently passed Genius Act, which Coinbase views as a red line.
Stablecoin Compliance and Competition
Copied to clipboard!
(00:10:36)
  • Key Takeaway: USDC, compliant under the Genius Act, is the largest regulated stablecoin, while Tether is currently not compliant with U.S. treasury reserve requirements.
  • Summary: Coinbase supports USDC as the largest regulated stablecoin, though they list others like PayPal’s stablecoin and support Tether in certain jurisdictions. The likely scenario is two versions of Tether: one compliant in the U.S. and one operating in the ‘Wild West’ outside the U.S.
Crypto Listing Standards and Trends
Copied to clipboard!
(00:13:17)
  • Key Takeaway: Coinbase’s platform philosophy is to be the ’everything exchange,’ listing all legal assets while relying on disclosures rather than explicit investment recommendations.
  • Summary: The three biggest crypto trends identified are: all assets moving on-chain for trading, rapid growth in prediction markets (where Coinbase is exploring partnerships like Calci), and the expansion of stablecoin payments. Coinbase uses minimum listing standards and disclosures, avoiding explicit investment ratings like AAA bonds.
Tokenization and Private Market Access
Copied to clipboard!
(00:17:16)
  • Key Takeaway: Tokenization of private assets like fund shares and real estate is inevitable, promising massive democratization of wealth creation by lowering costs and increasing access for the 4 billion unbrokered adults globally.
  • Summary: Tokenization must be done with company permission to avoid undermining employee retention mechanisms like vesting. The current system creates an unintended consequence where high demand for private shares leads to inefficient, high-fee SPV markets for secondary trading. Coinbase Tokenize helps funds and real estate projects tokenize assets to increase demand and reduce settlement risk.
California’s Fiscal Crisis
Copied to clipboard!
(00:23:53)
  • Key Takeaway: The exodus of high-net-worth individuals from California has already created a significant tax hole, exacerbated by ineffective spending on social issues like homelessness.
  • Summary: The departure of an estimated 20% of California’s billionaires has created a negative $10 billion tax hole, even accounting for those who remain. The state’s budget has increased dramatically while services have worsened, creating perverse incentives where increased spending on homelessness correlates with more homeless individuals.
Brian Armstrong’s Longevity Side Hustle
Copied to clipboard!
(00:27:54)
  • Key Takeaway: Brian Armstrong co-founded a biotech company, New Limits, focused on epigenetic reprogramming to restore cellular function, with its first drug candidate entering clinical trials next year.
  • Summary: Inspired by Elon Musk’s move into ‘atoms,’ Armstrong is investing capital from Coinbase’s IPO into this longevity venture. The fundamental science involves reprogramming cells to restore younger function, and the company has shown success in human cells within three years.
Davos Shift to Business and Economic Growth
Copied to clipboard!
(00:29:34)
  • Key Takeaway: The focus at Davos has shifted from ESG/DEI discussions to pragmatic deal-making and business growth, driven partly by the strong U.S. economic performance.
  • Summary: The current economic environment, characterized by high GDP growth (5.6%) and low unemployment, proves that growth stems from deregulation and private market activity, not government spending. This success is creating a win-win scenario under capitalism, benefiting even the poorest through increased economic freedom.
AI and Job Displacement Optimism
Copied to clipboard!
(00:32:20)
  • Key Takeaway: Brian Armstrong is a techno-optimist regarding AI job displacement, viewing it as a historical transition similar to agricultural automation that will lead to greater abundance and new types of work.
  • Summary: AI and crypto are the two most important tech trends, expected to converge as AI agents require stablecoin payments. Historically, automation shifts labor from hard manual work to cognitive tasks, and AI will likely continue this trend, leading to new jobs that focus on creativity or philosophy rather than tedious tasks.
Internal AI Implementation at Coinbase
Copied to clipboard!
(00:36:30)
  • Key Takeaway: Coinbase implemented an internal, hosted AI model connected to all company data, enabling ‘reverse prompting’ where the AI coaches the CEO on strategic blind spots.
  • Summary: The internal AI acts as an ‘Oracle of Coinbase,’ analyzing all internal communications and documents to surface disagreements or misaligned time allocation for the CEO. This contrasts with traditional prompting by asking the AI what the user should be thinking about, effectively serving as a high-level coach.
Cerebras WSE Chip Architecture
Copied to clipboard!
(00:39:02)
  • Key Takeaway: Cerebras’ Wafer Scale Engine (WSE) is a single chip 56 times larger than an NVIDIA B200, designed specifically to accelerate AI training and inference by processing massive amounts of data simultaneously.
  • Summary: The WSE contains 4 trillion transistors and is designed to deliver results 20 to 50 times faster than traditional compute for AI workloads. Systems can be purchased on-premise for around $1-1.5 million or rented via cloud services by the token, month, or year.
Speed as the Key to AI Adoption
Copied to clipboard!
(00:46:02)
  • Key Takeaway: Reducing latency in AI inference, even by milliseconds, fundamentally changes the user experience from a sequential task to an integrated flow, enabling entirely new applications.
  • Summary: The shift from slow compute to fast compute mirrors the transition from Netflix mailing DVDs to streaming movies, enabling a fundamental change in kind, not just degree. OpenAI placed a major purchase order for Cerebras capacity to deliver this speed, as slow responses cause users to switch services or lose focus.
AI Data Center Power Constraints
Copied to clipboard!
(00:49:05)
  • Key Takeaway: The limiting constraint for building large AI data centers is no longer square footage but the amount of electrical power that can be delivered to the site.
  • Summary: The industry now discusses data center capacity in terms of megawatts (MW) rather than square feet, exemplified by the 750 MW deal between OpenAI and Cerebras. The cheapest power sources are hydro, followed by natural gas, with flare-off gas being utilized by Bitcoin miners and now potentially AI infrastructure.
Data Center PR and Community Relations
Copied to clipboard!
(00:52:52)
  • Key Takeaway: Poor initial communication by hyperscalers regarding data center power needs led to local rate hikes and community resentment.
  • Summary: Misperceptions about AI data centers are being fueled by dark PR, but initial community issues stemmed from hyperscalers cutting deals with power companies that amortized infrastructure costs onto local residents’ rates. A better approach involves being good citizens, paying significant taxes, and investing in local amenities like schools. Microsoft recently committed to ensuring their energy usage will not increase local utility costs.
Grid Investment and Energy Storage
Copied to clipboard!
(00:55:45)
  • Key Takeaway: The US grid is decrepit due to decades of underinvestment, making distributed energy storage solutions like home batteries crucial for managing peak demand.
  • Summary: The nation failed to invest in its electrical grid for 40 to 50 years, leaving it behind advanced nations like China. Home battery solutions, which charge during low-demand periods, can alleviate load on the grid and provide resilience during outages. Nuclear power, while highly efficient long-term, is too far out to solve immediate data center power needs over the next three to four years.
AI Compute Architecture Trade-offs
Copied to clipboard!
(01:01:34)
  • Key Takeaway: Effective computer architecture requires balancing calculation speed, memory storage, and data transport speed, with GPU memory being a current bottleneck for fast inference.
  • Summary: Computer architects must constantly balance three dimensions: calculation speed, storage (memory), and the speed of getting results to the user (transport). GPUs have high memory capacity but slow memory access, which severely bottlenecks inference performance. This memory constraint is driving demand for faster access solutions and contributing to a current memory shortage.
Geopolitics and AI Chip Race
Copied to clipboard!
(01:05:13)
  • Key Takeaway: The US maintains a significant lead in high-speed chip design due to talent density, while China excels in the open-source model space and has aggressively modernized its power grid.
  • Summary: The US advantage in chip making stems from the high density of talent in the Santa Clara area, where the ability to iterate quickly builds expertise. China is behind in chip manufacturing but has pushed ahead in open-source models and has been more effective at rapidly deploying power infrastructure. The race is winner-take-all because early advantages in iteration speed are rapidly magnified.
US Policy Support for AI Industry
Copied to clipboard!
(01:08:03)
  • Key Takeaway: The Trump administration’s approach to AI policy is praised for empowering allies by allowing chip sales and streamlining investment processes, contrasting with the previous administration’s perceived adversarial stance.
  • Summary: The previous administration made mistakes by withholding chips from allies like the UAE and Saudi Arabia, which the current administration corrected by empowering these partners. Streamlining processes within organizations like CIFIA and Treasury has made investment and collaboration much easier. Efforts to create reasonable, standardized regulations across localities and improve the national grid are also viewed as highly positive for rapid growth.
AI vs. SaaS Impact on Employment
Copied to clipboard!
(01:14:51)
  • Key Takeaway: Current tech layoffs are primarily due to the delayed impact of efficient SaaS tools flattening organizational structures, rather than immediate AI-driven job displacement.
  • Summary: While AI displacement is certain to come, it is not the cause of current layoffs, which are concentrated in middle management. The value of middle management roles focused on information flow has shrunk due to leaders’ extended scope enabled by SaaS tools. Companies are flattening organizations in anticipation of intense competition, which is a structural change preceding widespread AI job elimination.
Robotics and Industrial ROI
Copied to clipboard!
(01:19:18)
  • Key Takeaway: The highest ROI for robotics lies in solving critical infrastructure and industrial problems, not in consumer tasks like folding laundry, requiring a focus on physical world data collection.
  • Summary: Gecko Robotics focuses on high-ROI applications in defense and energy, such as inspecting nuclear plants and expediting submarine manufacturing, where improvements of up to 90% in speed have been noted. The company acts as a ’nervous system’ by collecting proprietary data sets from physical assets that do not exist on the internet, creating a significant advantage for future AI model training. The next phase involves using this data to enable robots to take direct repair and action, moving beyond mere inspection.