Record numbers on both sides of the ledger this month — record debt, record yields, record IPO proceeds. Records are not the same thing as depth. This letter is about the difference.
Two numbers moved into record territory within days of each other this month, and neither move was really about growth. US national debt crossed $40 trillion. The 30-year Treasury yield touched its highest level since 2007. On the other side of the market, 2026's IPO proceeds have already passed the whole of 2025's — with a single SpaceX offering, on its own, larger than every US IPO of the two previous years combined. Read separately, these look like two unrelated stories: a fiscal one and a capital-markets one. Read together, they're the same question asked twice: is there actually more liquidity in the system right now, or just more paper circling a pool that hasn't gotten any deeper?
The 30-year Treasury yield has been grinding higher for weeks, and by mid-August it had reached levels not seen since 2007 — a genuinely long round trip, not a routine wobble. The proximate trigger was straightforward: national debt crossing $40 trillion, more than doubling in under a decade, landing at a moment when the market was already nervous about who absorbs the next round of issuance. The US Treasury tried to answer that nervousness directly, with a plan specifically built to ease pressure on long-dated yields. It failed within days — yields resumed climbing shortly after, erasing the brief relief the announcement had bought. That is the clearest single data point in this letter: when the market's most direct, most deliberate intervention against thin bond liquidity doesn't hold, the thinness is not a messaging problem. It's structural.
Oil has been compounding the same pressure from a different angle. Brent spent the back half of the month grinding from the mid-$80s toward $93 a barrel — closing above $93 on an intraday basis for the first time since late July — as the US-Iran standoff showed no sign of the diplomatic resolution markets had spent months pricing toward. Higher energy costs feeding into an already-nervous bond market is not a coincidence of timing; it's the same inflation-and-fiscal-space anxiety showing up in two markets that price off each other constantly.
On paper, 2026 is the strongest new-issue year in over a decade. US IPO proceeds had crossed $251 billion by late July — already ahead of the whole of 2025 — and the second quarter alone raised $104.8 billion, the largest quarterly total for US listings ever recorded. SpaceX's own listing accounted for $75 billion of that on its own, the largest IPO ever completed, priced at a peak valuation near $2.66 trillion in mid-June.
The number that actually matters for this letter's argument sits underneath those headlines: float size. A standard IPO offers 15-25% of a company's shares to the public. This year's mega-caps are structurally incapable of doing that — the deals are simply too large relative to what public markets can absorb in one sitting — so floats are landing in the 3-8% range instead. That is not a footnote. A $2.66 trillion company trading on an 5%-ish float has a genuinely small pool of tradable shares relative to its headline size, and thin floats amplify every subsequent move, up or down. SpaceX's own post-listing trajectory makes the point without needing commentary: shares that peaked near that $2.66 trillion valuation on 16 June were back near the original offer price by early August. A company can raise a record sum and still hand its new shareholders a genuinely illiquid, volatile position, because raising the money and creating a deep trading market for the resulting stock are not the same achievement.
OpenAI and Anthropic are the two names every desk is watching next, and both illustrate the same fragility from a different angle. OpenAI has reportedly moved its target toward a Q4 2026 listing near a $1 trillion valuation, with Goldman Sachs, Morgan Stanley, and JPMorgan said to be lining up investor meetings for as early as October — though other reporting suggests the timeline could still slip to 2027 depending on how the rest of the year's debuts land. Anthropic closed a $30 billion round in February at a $380 billion valuation and has been positioned toward an October target of its own. Neither number is the risk. The risk is sequencing: industry analysts covering this cycle have been explicit that a single disappointing debut from either company — even just a modest first-day reception rather than an outright failure — could shift sentiment quickly enough to push the next tier of filers back onto the sidelines, echoing how 2021's late-year IPO wobbles bled into a two-year freeze. A market this concentrated in a handful of outsized, thin-float names is a market where one bad print does more damage than it would in a broader, deeper field of debuts.
Record debt and record IPO proceeds look like opposite stories — one is the government struggling to place its paper, the other is private companies rushing to sell theirs. They're the same story. Both are testing how much real depth sits underneath a market that has spent this year setting size records without proving it has actually gotten any deeper.
This is checkable, not just a mood. If the 30-year yield stabilizes below its current multi-decade highs over the next quarter without another failed intervention, the bond side of this thesis weakens. If OpenAI and Anthropic both price and trade steadily through their first months — floats notwithstanding — the IPO side of it weakens too, and 2026 genuinely earns its "banner year" label rather than a "record headline, thin foundation" one. If either fails — a stalled Treasury response to yields, or a wobbly mega-cap debut that cools the pipeline behind it — this letter's actual argument is that the two would not be a coincidence. They'd be the same underlying liquidity constraint showing up in the two places most people aren't looking at together.
$40 trillion in debt and a failed yield-suppression plan on one side; $251 billion in IPO proceeds built on floats a third the size of a normal cycle on the other. Both are real, verified, current numbers — and both describe the same condition from opposite directions: more paper, not more depth. Neither market has to break for this thesis to matter. The question worth tracking isn't whether the records keep coming. It's whether either market can absorb its next real shock without the thinness showing.