Latest / Finance TL;DR Daily / Warsh Rewires the Fed's Playbook as Anthropic Loses Its Court Fight
Transcript
- Warsh Rewires the Fed's Playbook as Anthropic Loses Its Court Fight 2026-09-25 Alex: Welcome to Finance TL;DR Daily, the podcast where we break down the biggest stories in finance. It's Friday, September twenty-fifth. And today the Fed's top job just got a lot less predictable, on purpose. Riley: Here's what's moving today. Alex: So Kevin Warsh, a hundred and twenty-seven days into the chairmanship now, is basically tearing up the old rulebook. He's dropping forward guidance — that's the Fed telling markets in advance roughly where rates are headed — and instead he wants to lean on what markets themselves are already signaling. Riley: Which sounds almost circular. The Fed watches the market, so it can decide policy, which the market then reacts to. Alex: It is a bit of a hall of mirrors. UBS economist Jonathan Pingle put it well — he said Warsh's framework is less rooted in economic measurement and more reflective of market narratives. That's a real shift from the Bernanke-to-Powell era of telegraphing everything. Riley: And this isn't just theoretical. Last week the Fed raised rates a quarter point, the first hike since 2023, with inflation still sitting at three point seven percent in July. Alex: Right, and the bond market's already pricing in a seventy percent chance of another hike in October. The two-year Treasury yield is trading nearly a full point above the actual fed funds rate — that almost never happens unless investors think the Fed is behind the curve. Riley: What about the balance sheet? He came in wanting to shrink that six point seven trillion dollar pile. Alex: Progressing slowly, is the polite way to put it. Inflation's the priority, colleagues are reluctant, and the ten-year yield pushing above five percent isn't helping. He's also floated cutting FOMC meetings from eight a year down to six. Riley: Fewer meetings, less guidance, more market-watching — that's a genuinely different Fed to trade around. Alex: It is. Investors used to know roughly what the Fed would say next. Now they're being asked to read the tape and guess what the tape is telling the Fed. Riley: Right, well, staying with things the courts are now deciding rather than the market — a federal appeals court in Washington just upheld the Pentagon's decision to blacklist Anthropic as a supply chain risk. Alex: That's a big one. What's the actual mechanics of the ruling? Riley: The D.C. Circuit ruled two to one — Judges Katsas and Rao in the majority, Judge Henderson dissenting. The root of it is that Anthropic refused to loosen its contractual bans on using Claude for lethal autonomous weapons or domestic surveillance. The Pentagon argued that refusal created uncertainty about how the model would perform in military use. Alex: So this isn't about Anthropic doing something wrong with the tech — it's about them holding a line on how it's used, and that line itself getting treated as the risk. Riley: Exactly, and Judge Katsas's opinion is pretty blunt about it — he wrote the Department had, quote, ample support for its conclusion that continued integration of Claude into its information systems presented a statutorily covered national-security risk. Alex: What does that actually do to Anthropic commercially? Riley: It cements the blacklist designation, which makes government contracting harder and puts a legal precedent on the books — the first time a court's affirmed that an AI company's own usage restrictions can be read as a national security liability. Alex: That's the bit that could ripple outward. Any AI firm with ethical guardrails around military or surveillance use now has a court ruling saying those guardrails themselves can trigger a security-risk label. Riley: Which puts commercial values on one side of the scale and Pentagon integration on the other, and this ruling just told everyone which side the government's willing to weigh more heavily. Alex: Meanwhile the Fed's had a busy week outside of rates too — they've put out their first real stablecoin rulemaking under the GENIUS Act. Riley: Remind us what the GENIUS Act actually covers. Alex: It's the law from July 2025 that set up federal rules for payment stablecoins — those are the tokens meant to hold a steady one-dollar value, typically backed by cash or short-term Treasuries. The Fed just opened formal public comment on two proposals to implement it. Riley: What's actually in them? Alex: The first requires full backing with things like short-term Treasury bills, sets capital requirements, and bans paying yield just for holding the stablecoin. The second sets up an application process for Fed-supervised banks that want to issue their own. Riley: Sixty-day comment period, and this is already well past the statutory deadline, isn't it — that was supposed to be finalized by July of this year. Alex: It is, so regulators are playing catch-up. Fed Governor Michael Barr's line on it was pretty direct — stablecoins will only be stable if they can be reliably and promptly redeemed at par in a range of conditions. Riley: Which is the whole point, really — a stablecoin that can't be redeemed at par under stress isn't stable, it just looks stable until it isn't. Alex: And banks now have an actual pathway to get into this business under Fed supervision, which matters for how crypto and traditional banking start overlapping. Riley: Right, and while we're on things the Fed is quietly worried about — Kansas City Fed President Jeff Schmid raised a question this week that's been sitting in the back of a lot of people's minds. Is the AI buildout becoming too big to fail? Alex: That's a loaded phrase given what it usually refers to. Riley: It is, he drew the comparison directly to the 2007 to 2009 crisis bailouts. His point is about the web of contracts and financing tying data-center builders, chipmakers and hyperscalers together — if one piece wobbles, does it take the rest down with it? Alex: Is there any actual sizing on how big this is relative to something like the housing boom? Riley: Apollo's Torsten Slok has looked at that — he reckons the data-center buildout is still less than half the size housing was at its 2005 peak, which hit six point six percent of GDP. But it's growing faster relative to GDP than housing did. Alex: And the Fed itself doesn't agree on how worried to be, from the sound of it. Riley: Not remotely. Mary Daly at the San Francisco Fed called the investment growth very worrisome, while John Williams in New York said flatly he doesn't see a bubble. So you've got the same data producing opposite conclusions inside the same institution. Alex: Which tells you this is now genuinely a financial-stability conversation, not just a valuation one. Riley: That's the shift worth watching — if the Fed starts treating hyperscaler AI spending the way it treats systemically important banks, that's capital rules and oversight questions that don't exist yet. Alex: So today it's a new Fed playbook with fewer signals, a court ruling that redraws the line on AI and national security, and two live regulatory fronts — stablecoins and now potentially AI infrastructure itself. Riley: All four stories point the same direction, really — Washington is writing more rules for finance and AI at once, and markets are being asked to read between fewer lines than they're used to. Alex: This has been Finance TL;DR Daily. We'll see you tomorrow.