The Machine That Doesn't Need Cheap Money

There's a comfortable story that's been told for two years about the Fed and markets: hike rates, and speculative excess gets squeezed out. Sell the trade. Watch the leverage unwind. It's a story that worked reasonably well in 2022, and investors have been waiting for a sequel ever since. It isn't playing out that way this time, at least not for artificial intelligence.

On September 16th, the Federal Reserve voted 12-0 to raise its target rate a quarter point, to a range of 3¾ to 4%. Sixteen of eighteen policymakers see another hike coming before the year is out. By the old playbook, that should be putting real pressure on the most speculative corner of the market — and AI, with its eye-watering capital commitments and uncertain near-term payoff, should be exhibit A.

Instead, the number keeps getting bigger. Goldman Sachs now projects roughly $1 trillion in AI-related capital investment globally this year. JPMorgan and Goldman both see that figure climbing past $1 trillion again in 2027, with trillions more stacked up behind it. This isn't retail money chasing a meme. It's hyperscaler operating cash flow, sovereign wealth, pension allocations, private equity, and foreign capital, moving in a volume that dwarfs almost anything a quarter-point rate move can meaningfully dent. The AI buildout isn't financed the way a leveraged bond trade is financed — it isn't sitting on overnight repo, bleeding out with every tick up in the effective fed funds rate the way we discussed with Treasury carry. It's largely self-funded by companies posting some of the highest operating margins in corporate history. Rate hikes are a headwind for a trader borrowing to hold a 10-year note. They're a rounding error for a company deploying its own free cash flow into a data center it expects to still be running in 2035.

That resilience is starting to look like leverage of its own — not financial leverage, but geopolitical leverage. And this is where the story gets more interesting than a simple capex chart.Chip access has quietly become one of the more useful bargaining chips in the US-China relationship. Washington has repeatedly delayed planned tariffs on Chinese semiconductor imports, most recently pushing the timeline out to mid-2027, explicitly to preserve room for negotiation. The administration has also eased restrictions on advanced AI chip sales to China — approving Nvidia's H200 and AMD's MI325X for export under new conditions — a real reversal from the "presumption of denial" posture that defined the prior approach. None of this means the AI capital cycle single-handedly dragged both countries back to the table. The trade truce was already moving that direction for its own reasons — rare earths, tariff fatigue, a fragile ceasefire neither side particularly wants to break. But the AI trade has given both governments an additional reason to keep talking rather than escalate: China wants the compute, the US wants the market for its chips, and neither outcome survives a full rupture. Call it an accelerant rather than the spark.

Meanwhile, Europe is absorbing pain that has nothing to do with any of this and everything to do with it at the same time. Germany's 10-year Bund yield just hit its highest level since 2009. France's yield spread against Germany has widened to nearly 100 basis points — a gap last seen during the 2012 eurozone debt crisis — driven by a fiscal deficit north of 5% of GDP and a government that's burned through prime ministers at an alarming clip. Some of this is genuinely self-inflicted; France's politics would be shaky in any global environment. But a real share of it is imported: oil prices pushing above $100 a barrel, feeding straight into inflation expectations, feeding straight into how aggressively the ECB has to respond. And the reason oil is behaving that way isn't abstract market psychology. It's a shooting war.

This is worth being precise about, because it's easy to talk about "Middle East tension" as background noise. It isn't background noise right now. The US struck Iranian positions in the Strait of Hormuz; Iran retaliated against US assets in the region and, by some accounts, against Kuwait and Bahrain; the Houthis have struck Saudi Arabia. Bond market commentary this month refers matter-of-factly to events happening "since the Iran war began." That's a materially different situation than pre-conflict brinkmanship, and it changes what "working it out with Iran" would actually require. A bilateral US-Iran deal, even a generous one, doesn't automatically stand down Kuwait, Bahrain, Saudi Arabia, and the Houthis. De-escalation now needs more signatures on it than it would have needed six months ago.

Whether that de-escalation happens before midterms, and whether it happens because the political incentives point that way, is a genuinely open question rather than a foregone conclusion. There's a reasonable case that an unresolved, costly conflict — one visibly showing up in gas prices and bond yields at the same time — creates real domestic political exposure heading into an election. There's an equally reasonable case that a sitting president can consolidate rather than lose support around a live conflict, and that midterm outcomes get decided by a dozen factors that have nothing to do with the Strait of Hormuz. Anyone telling you they know how that resolves is telling you their politics, not a fact.

What's less ambiguous is the shape of the multipolar pressure building underneath all of this. The BRICS summit that concluded in New Delhi this month wasn't a footnote — the bloc produced more than fifty concrete outcomes under India's chairship and adopted a declaration explicitly condemning unilateralism and coercive economic measures, all while representing roughly 40% of global GDP. That's not nothing. But it's also not a unified anti-Western trade bloc marching in lockstep; negotiators reportedly had to bridge real divisions between Iran and the UAE just to get a joint statement out the door. BRICS is less a battering ram against US hegemony and more a slow, uneven diffusion of leverage away from any single center of gravity — a trend, not an event.

Put all of it together and you get something less like a single unified thesis and more like several currents running in the same rough direction. AI capital is big enough and self-funded enough to shrug off a rate hike that would have flattened a more leveraged trade. That resilience gives the US and China a shared reason to keep the chip-and-tariff negotiation alive rather than let it collapse. Europe is paying a real price for a war that's more active than casual conversation gives it credit for, and that war's resolution is now a multi-party problem, not a two-party one. And underneath all of it, BRICS is chipping away at the assumption that Washington sets every table alone — not by declaring independence, but by quietly building its own tables.

None of that guarantees a clean return to cooperative trade diplomacy on any particular timeline. It does suggest that the forces pushing toward negotiation — capital too large to walk away from, war too expensive to sustain indefinitely, a multipolar world too diffuse to dictate to — are real and currently outweighing the forces pushing toward further fracture. That balance can tip either way. It's just currently tipped toward the table.

* References: U.S. Federal Reserve monetary policy and FOMC interest rate decisions; Goldman Sachs and JPMorgan global artificial intelligence capital expenditure projections ($1 trillion scale); U.S.-China semiconductor trade negotiations and chip export policies (Nvidia H200, AMD MI325X); European sovereign debt and bond market metrics (German 10-year Bund yields, French yield spreads); Middle East geopolitical conflicts (Strait of Hormuz, Iran, Houthis, Saudi Arabia); and BRICS summit outcomes and multilateral economic policy.