As of November 30, 2025 (about 10 months after the January 25, 2025 podcast), the direction of events is consistent with Friedberg’s claim—there is a large-scale build‑out of AI data centers and advanced chip fabs in the U.S.—but his prediction explicitly concerned what would happen “over the next several years,” which has not yet elapsed.

On the facts: the U.S. is seeing huge chip and data‑center expansion. TSMC has announced that total U.S. investments will reach about $165 billion, including multiple additional fabs, advanced packaging plants, and an R&D center in Arizona, with a long‑term build‑out schedule through at least 2030. (azcommerce.com) Arizona and surrounding states are being positioned as a major semiconductor hub, with dozens of semiconductor-related projects and TSMC’s Arizona fabs already producing advanced chips (including leading AI processors) while more capacity is planned. (economy.ac) In parallel, hyperscale cloud providers are sharply increasing capex for AI servers and data centers: TrendForce estimates combined capex of the eight largest CSPs will jump to about $420 billion in 2025 and exceed $520 billion in 2026, driven heavily by GPU-based AI infrastructure build‑outs. (finance.yahoo.com) Major individual projects like Amazon’s recently announced $50 billion in new U.S. data centers for AI and high‑performance computing for federal customers reinforce the scale of ongoing expansion. (wsj.com)

However, Friedberg’s testable claim was that this “data center and chip infrastructure effort” would “obviously” go forward with or without government involvement over the next several years. To judge that, we would need to see whether the build‑out continues (or would have continued) regardless of evolving U.S. federal policy through a multi‑year window—something we cannot yet fully observe only 10 months later. In addition, many of the marquee chip investments explicitly intertwine with U.S. federal policy and incentives (e.g., CHIPS Act–related support for TSMC’s Arizona expansion), which further complicates any present attempt to separate what’s happening because of government from what would happen without it. (techpowerup.com)

Because the timeframe of “several years” has not passed and the counterfactual of “without government involvement” cannot yet be meaningfully evaluated, the prediction is best classified as inconclusive (too early to tell) rather than clearly right or wrong.