Replay 2025: David Sacks on AI, Crypto, and America's Technology Future

What It Covers

David Sacks, serving as the Trump administration's AI and crypto czar, outlines the policy framework across both domains: deregulating AI to preserve permissionless innovation, establishing crypto regulatory clarity through legislation like the Genius Act and Clarity Act, and competing with China by expanding American technology exports to allied nations rather than restricting them.

Key Questions Answered

• •Crypto Regulatory Strategy: The Clarity Act, currently moving through the Senate requiring 60 votes, would provide regulatory frameworks for the 94% of crypto tokens not covered by the Genius Act's stablecoin provisions. Founders need 10-20 year certainty, not just favorable current leadership at the SEC. Negotiators are working with roughly a dozen Senate Democrats, mirroring the 18 Democrats who supported the Genius Act.
• •AI Export Policy: Restricting GPU and chip sales to allied nations like Saudi Arabia and the UAE drives those countries toward Huawei and Chinese AI infrastructure. Every excluded nation strengthens China's technology ecosystem. The strategic error is treating diffusion as a threat rather than recognizing that maximum global adoption of American technology is how the US wins the AI race long-term.
• •Regulatory Capture Risk: Anthropic's policy leadership publicly acknowledged that transparency reporting requirements like California's SB 53 are stepping stones toward mandatory pre-approval of new AI models before release. A pre-approval regime would functionally eliminate startup competition, since large companies navigate regulatory bureaucracies effectively while early-stage founders cannot, replicating the innovation-suppressing dynamics seen in pharma, banking, and defense.
• •Energy Bottleneck: Nuclear power requires 5-10 years to deploy at scale. Near-term AI data center power depends on natural gas, but gas turbine manufacturing has a 2-3 year backlog with only 2-3 global producers. A faster bridge solution: shedding 40 peak hours per year to backup diesel generators could free approximately 80 gigawatts of existing grid capacity currently held in reserve for seasonal demand spikes.
• •AI Development Reality Check: The imminent AGI narrative is losing credibility among practitioners. Andrej Karpathy now estimates AGI is at least a decade away, citing reinforcement learning's limits. Current models remain middle-to-middle processors requiring human prompting, objective-setting, and output validation. Agents perform reliably only on narrow, well-defined tasks — broad objectives produce unreliable results requiring frequent human intervention before completion.
• •State Regulation Fragmentation: Over 1,200 AI-related bills are moving through state legislatures, with more than 100 measures already passed. California, New York, Colorado, and Illinois account for 25% of activity. Colorado, Illinois, and California have each enacted algorithmic discrimination laws holding model developers liable for disparate impact outputs — a compliance standard that effectively mandates embedding DEI filtering layers directly into model outputs.