What are the key takeaways from this Practical AI episode?

Key insights include: **Computing Power as the Strategic Variable:** A 2020 OpenAI paper called "Neural Scaling Laws for Neural Networks" established that AI capability scales primarily with compute, not data. This physicality transforms AI from an abstract software race into a supply chain competition centered on semiconductor chips—making export controls and chip manufacturing geography the most consequential policy levers available to governments.; **Taiwan Semiconductor Concentration Risk:** TSMC manufactures approximately 97% of the world's advanced AI chips using equipment from ASML (Netherlands), plus US and Japanese suppliers—all democracies. Analysts estimate a disruption to Taiwan's chip output would cost trillions in global GDP losses. The US Chips and Science Act began domestic production in Arizona, but Taiwan retains a multi-year manufacturing lead.; **DeepSeek Confirms, Not Refutes, Chip Dominance:** DeepSeek's V3 paper explicitly acknowledges compute scarcity as its primary constraint, and its CEO stated publicly that talent and capital are not limiting factors—chips are. DeepSeek's models rely on stockpiled or smuggled US chips. This means algorithmic efficiency gains by Chinese researchers do not undermine the strategic value of US export controls.

What did Ben Buchanan discuss on Practical AI?

Ben Buchanan, former White House Special Adviser on AI under Biden, examines how computing power—not data—drives AI geopolitical competition, why Taiwan's TSMC produces 97% of advanced chips, and how democracies can maintain AI leadership through export controls, international coordination, and values-aligned deployment frameworks. Key topics include: **Computing Power as the Strategic Variable:** A 2020 OpenAI paper called "Neural Scaling Laws for Neural Networks" established that AI capability scales primarily with compute, not data. This physicality transforms AI from an abstract software race into a supply chain competition centered on semiconductor chips—making export controls and chip manufacturing geography the most consequential policy levers available to governments.; **Taiwan Semiconductor Concentration Risk:** TSMC manufactures approximately 97% of the world's advanced AI chips using equipment from ASML (Netherlands), plus US and Japanese suppliers—all democracies. Analysts estimate a disruption to Taiwan's chip output would cost trillions in global GDP losses. The US Chips and Science Act began domestic production in Arizona, but Taiwan retains a multi-year manufacturing lead..

How long is this episode of Practical AI?

This episode is 48 minutes long. SignalCast provides an AI-generated summary so you can get the key insights in about 3 minutes.

Practical AI

AI policy and the battle for computing power

March 9, 2026

48 min episode · 2 min read

Ben Buchanan

Episode

48 min

Read time

2 min

Topics

Productivity, Startups, Leadership

AI-Generated Summary

Published Mar 10, 2026

Key Takeaways

✓Computing Power as the Strategic Variable: A 2020 OpenAI paper called "Neural Scaling Laws for Neural Networks" established that AI capability scales primarily with compute, not data. This physicality transforms AI from an abstract software race into a supply chain competition centered on semiconductor chips—making export controls and chip manufacturing geography the most consequential policy levers available to governments.
✓Taiwan Semiconductor Concentration Risk: TSMC manufactures approximately 97% of the world's advanced AI chips using equipment from ASML (Netherlands), plus US and Japanese suppliers—all democracies. Analysts estimate a disruption to Taiwan's chip output would cost trillions in global GDP losses. The US Chips and Science Act began domestic production in Arizona, but Taiwan retains a multi-year manufacturing lead.
✓DeepSeek Confirms, Not Refutes, Chip Dominance: DeepSeek's V3 paper explicitly acknowledges compute scarcity as its primary constraint, and its CEO stated publicly that talent and capital are not limiting factors—chips are. DeepSeek's models rely on stockpiled or smuggled US chips. This means algorithmic efficiency gains by Chinese researchers do not undermine the strategic value of US export controls.
✓Safety and Speed Are Complementary, Not Opposing: The railroad analogy illustrates this: early US rail had no standardized track gauges, no air brakes, and no time zones, causing thousands of deaths. Government-private sector coordination over decades produced both safer and faster trains. Applying this to AI, safety frameworks like Biden's Executive Order and National Security Memorandum were designed to enable adoption, not restrict it.
✓Three-Part Democracy Scorecard for AI: Evaluate democratic AI leadership across invention (are democracies building frontier models?), adoption (are governments and economies deploying AI effectively for security and productivity?), and values alignment (does deployment guard against job displacement, power centralization, surveillance, and disinformation?). Buchanan rates invention as the strongest current advantage and values alignment as the least resolved.

What It Covers

Ben Buchanan, former White House Special Adviser on AI under Biden, examines how computing power—not data—drives AI geopolitical competition, why Taiwan's TSMC produces 97% of advanced chips, and how democracies can maintain AI leadership through export controls, international coordination, and values-aligned deployment frameworks.

Key Questions Answered

•Computing Power as the Strategic Variable: A 2020 OpenAI paper called "Neural Scaling Laws for Neural Networks" established that AI capability scales primarily with compute, not data. This physicality transforms AI from an abstract software race into a supply chain competition centered on semiconductor chips—making export controls and chip manufacturing geography the most consequential policy levers available to governments.
•Taiwan Semiconductor Concentration Risk: TSMC manufactures approximately 97% of the world's advanced AI chips using equipment from ASML (Netherlands), plus US and Japanese suppliers—all democracies. Analysts estimate a disruption to Taiwan's chip output would cost trillions in global GDP losses. The US Chips and Science Act began domestic production in Arizona, but Taiwan retains a multi-year manufacturing lead.
•DeepSeek Confirms, Not Refutes, Chip Dominance: DeepSeek's V3 paper explicitly acknowledges compute scarcity as its primary constraint, and its CEO stated publicly that talent and capital are not limiting factors—chips are. DeepSeek's models rely on stockpiled or smuggled US chips. This means algorithmic efficiency gains by Chinese researchers do not undermine the strategic value of US export controls.
•Safety and Speed Are Complementary, Not Opposing: The railroad analogy illustrates this: early US rail had no standardized track gauges, no air brakes, and no time zones, causing thousands of deaths. Government-private sector coordination over decades produced both safer and faster trains. Applying this to AI, safety frameworks like Biden's Executive Order and National Security Memorandum were designed to enable adoption, not restrict it.
•Three-Part Democracy Scorecard for AI: Evaluate democratic AI leadership across invention (are democracies building frontier models?), adoption (are governments and economies deploying AI effectively for security and productivity?), and values alignment (does deployment guard against job displacement, power centralization, surveillance, and disinformation?). Buchanan rates invention as the strongest current advantage and values alignment as the least resolved.

Notable Moment

Buchanan reveals that China's own published roadmaps acknowledge they cannot independently produce chips matching current US-export-level performance until 2028—meaning the chip export controls enacted between 2022 and 2024 created a concrete, time-bounded technological gap that current policy reversals are actively eroding.

Know someone who'd find this useful?

Episode Transcript

Welcome to the Practical AI podcast, where we break down the real world applications of artificial intelligence and how it's shaping the way we live, work, and create. Our goal is to help make AI technology practical, productive, and accessible to everyone. Whether you're a developer, business leader, or just curious about the tech behind the buzz, you're in the right place. Be sure to connect with us on LinkedIn, x, or Blue Sky to stay up to date with episode drops, behind the scenes content, and AI insights. You can learn more at practicalai.fm. Now onto the show. Welcome to another edition of the Practical AI podcast. I am your cohost, Chris Benson. I am an AI and autonomy engineer at Lockheed Martin. And, today with me, our guest is Ben Buchanan, who is an assistant professor at John Hopkins University School of Advanced advanced international studies. He was previously the white house special advisor on AI to the Biden administration. He's been the author of four books. One of them, about to come out called the bitter struggle. And he also authored, a recent article for, foreign affairs magazine called the AI grand bargain, what America needs to win the innovation race. Welcome to the show, Ben. Thanks for having me. Appreciate it. So, I'm kinda curious if you, could tell us a little bit about, that that's that's a fantastic, set of things that you've that you've done professionally. And I often, we like when people have these, kind of amazing backgrounds, we often like people to start off. Just kinda tell us how you got to where you got and and why you're why you're passionate about the topic before we dive into policy issues just to get a little personal spin on your background there. Of course. Well, the the cool thing, at least in my view about a lot of jobs I had is that they didn't exist, before I had them. And there wasn't there wasn't a a White House special adviser, for AI and the like. So it's been a real adventure, but it's not the kind of thing that I've tried to plan out that I would do these. So, the short answer is by accident and by luck. But I do think I I got into AI, really in 2013, 2014, 2015 when I was doing my PhD in cyber operations and how nations hack one another and, what that means for international affairs. And we were just transitioning as a society or as an AI community from an older paradigm to a newer machine learning paradigm then. And I kinda noticed this happening. And there's a period in the PhD when everyone's kinda sick of their own subjects and they're looking for things to procrastinate with. And for me, that was AI at a time when not it was not really a policy subject. And then over the last ten or fifteen years or so, it's just becoming even …

Get the full transcript (9,353 words) + summary by email — free

One-time email with the complete transcript and AI summary of this episode. No account needed.

One email, no spam. We’ll also show you what SignalCast does.

Browse all Practical AI transcripts →

You just read a 3-minute summary of a 45-minute episode.

Get Practical AI summarized like this every Monday — plus up to 2 more podcasts, free.

Pick Your Podcasts — Free

Keep Reading

Less about Models; More about Architecture

Sep 3 · 45 min

The Prof G Pod

The New Playbook for Resisting Authoritarianism — with Julia Angwin & Ami Fields-Meyer

Jul 16

Building the Foundation for the Agentic AI Era

Aug 28 · 45 min

Stay Tuned with Preet

Democrats Done Playing Nice (with Susan Rice)

Feb 19

Books, tools, and gear mentioned in this episode

SignalCast may earn commission on purchases via these links. As an Amazon Associate, SignalCast earns from qualifying purchases.

Books

Neural Scaling Laws for Neural NetworksBy guest
by OpenAI
“A 2020 OpenAI paper called "Neural Scaling Laws for Neural Networks" established that AI capability scales primarily with compute, not data.”
Bookshop Amazon Audible trial