Two days after the 30-year yield hit a nineteen-year high, the Treasury blinked.
Stocks rebounded Wednesday after the Treasury Department said it would at least double its buybacks of long-dated government bonds — an announcement that arrived barely two weeks after it had published this quarter’s buyback schedule. Markets read it as exactly what it looked like: a government watching borrowing costs climb toward multi-decade highs and signaling it would rather not let them go further. The 30-year yield fell more than 10 basis points and the 10-year dropped over 6 to 4.637%.
The S&P 500 rose 0.21% to 7,708 and the Nasdaq Composite added 0.16% to 26,331, with advancing issues outnumbering decliners by better than two to one on the NYSE. Semiconductors stayed weak. But the session belonged to healthcare, where Moderna posted the best day in its history on a result that will be remembered well beyond financial markets.
Index Performance
Rates, Dollar, Commodities
The Treasury Buyback: A Quiet Operation Twist
The mechanics deserve explaining, because the Treasury buyback expansion is a more consequential policy move than its low-key delivery suggested.
The department said it would at least double repurchases of bonds in the 10-to-30-year range. It did not specify how the buybacks would be funded, but the Treasury has historically leaned on short-dated bill issuance to meet variable financing needs. If that is the plan here — retiring long bonds and replacing them with bills — the effect closely resembles the Federal Reserve’s Operation Twist, in which the central bank sold short-term securities and bought long-dated ones to press down on the long end of the curve. Different institution, same objective: reduce the supply of duration the market must absorb, and long rates fall.
The fiscal backdrop explains the urgency. Total US public debt has surpassed $40 trillion for the first time, having grown by roughly a third in under five years. With yields where they are, interest costs have reached $1.17 trillion in fiscal 2026 to date — now the third-largest line in the federal budget, behind only healthcare and Social Security. Every basis point on the long end compounds that burden.
LPL Financial’s Lawrence Gillum framed it precisely: the buybacks are “more of a Band-Aid than a panacea,” but they serve as a reminder that the Treasury is watching market conditions and will do what it can to prevent rates from rising too far, too fast.
That framing matters for how much comfort to take. The buyback addresses the symptom — the supply-demand imbalance at the long end — without touching the causes, which are a widening deficit, sticky inflation, oil that has stopped falling, and a corporate bond calendar swollen by AI infrastructure financing. This column identified long-end disorder as the market’s most persistent structural fragility over the weekend. Wednesday didn’t resolve it. It bought time.
Moderna’s 177%: A Medical First and a Market Record
Moderna closed at $174.38, up 176.97% — the best single day in the company’s history — after it and Merck announced that their personalized mRNA cancer vaccine met its primary endpoint in a Phase 3 melanoma trial. Volume reached 185.1 million shares, roughly 1,819% above the three-month average.
The science is genuinely historic. The therapy, intismeran autogene, is manufactured individually for each patient from mutations found in their own tumor, encoding as many as 34 neoantigens to train T cells to recognize that specific cancer. In the trial, 1,137 patients with high-risk melanoma that had been surgically removed received either intismeran plus Merck’s Keytruda or Keytruda alone. The combination significantly extended recurrence-free survival and reduced the risk of the cancer spreading to distant organs. It is the first mRNA-based cancer therapy ever to succeed in a Phase 3 trial.
“It’s a big moment for medicine, a big moment for patients,” Moderna CEO Stéphane Bancel said. Merck Research Laboratories president Dean Li framed the logic of the approach: intervening earlier, when cancers are most treatable, to increase the possibility of cure. Earlier Phase 2b data had pointed this direction — at 60.3 months, recurrence or death risk fell 49% and distant metastasis or death risk fell 59% — but a Phase 3 readout is a different order of evidence.
The market impact rippled outward. Merck rose 12.6% to a record close and was the Dow’s biggest contributor. BioNTech gained 21%, Novavax 6%, and Eli Lilly 5.3% — pushing Lilly’s market capitalization to $1.21 trillion, making it the first pharmaceutical company in history to clear $1.2 trillion. The S&P 500 healthcare sector rose 2.9% to a record high, providing the single largest support to the index.
The disparity in the two headline moves is worth understanding: Merck entered the day at roughly $333 billion in market value, Moderna at about $25 billion. Same trial result, wildly different leverage to it.
Two caveats belong in any honest account. Full trial data — hazard ratios, confidence intervals, subgroup consistency — will only be presented at an upcoming medical meeting, and neither company has disclosed when they intend to file for approval. Personalized manufacturing economics, including turnaround time and capacity at scale, remain unresolved questions for commercial value.
Marvell Wins Google’s TPU Business — at Broadcom’s Expense
Marvell Technology gained close to 10% after disclosing an agreement with Google related to its tensor processing units, paired with an unusual financial structure: Marvell issued Alphabet a warrant to purchase up to $12.2 billion of Marvell common stock — as many as 58.97 million shares at an exercise price of $206.58.
Broadcom fell more than 4% on the read-across, and the reason is straightforward. Broadcom has been the manufacturing partner behind Google’s TPU program for years; a deal that brings Marvell into that supply chain dilutes what had been a near-exclusive relationship. It is a notable turn from just a week ago, when both Alphabet and Broadcom weakened on skepticism that TPUs could meaningfully erode Nvidia’s share. The competitive question has shifted from whether custom silicon can challenge Nvidia to who gets to build it.
The warrant structure is itself part of a pattern worth tracking. Equity-linked arrangements between AI buyers and suppliers — Nvidia’s stakes in its customers, Alphabet’s warrant in Marvell — keep multiplying, and each one further entangles the balance sheets of the companies selling AI infrastructure with those buying it.
OpenAI Completes a Three-Day Reckoning
The chip weakness had a specific source. The Wall Street Journal reported, citing sources, that OpenAI’s disclosed second-quarter results disappointed investors: revenue grew 18% quarter over quarter to $6.7 billion from $5.7 billion, while losses widened.
Set that against Anthropic’s $11.6 billion in the same quarter — more than double its first-quarter figure — and the picture that has driven three consecutive sessions comes into focus. Monday, chips rallied on Anthropic’s revenue. Tuesday, they fell when Anthropic’s $65 billion run rate came in below the $70–80 billion whisper. Wednesday, they fell again as OpenAI’s numbers landed well short of that. The semiconductor trade is a leveraged bet on frontier-lab spending, and for three days running, the labs have been the ones setting the price.
Broadcom lost more than 4%, AMD nearly 4%, and the iShares AI Innovation and Tech Active ETF slipped about 2%. One mitigating note: OpenAI’s second quarter closed before it launched GPT-5.6 last month, and growth is reported to have reaccelerated since.
FOMC Minutes: Hawkish, but Less Than Feared
July’s meeting minutes confirmed a divided committee. Several officials favored raising rates at that meeting, and many said further tightening could become necessary if inflation fails to slow. Markets read the tone as less hawkish than the dissent count had implied.
The more interesting passage concerned artificial intelligence. Some participants assessed that AI’s effect on prices has so far been confined to specific categories. Several others judged that AI is increasing aggregate demand and is having — or is likely to have — broader price effects. A few said it remains too early to determine whether AI will drive shifts across a range of goods and services or produce a more persistent and widespread impact on inflation.
That three-way split is the Fed publicly working through a question this column has tracked since the memory shortage first appeared in consumer prices: is AI an inflationary force in its own right? The committee does not yet agree, and the minutes make the disagreement explicit.
Separately, the minutes revealed that Chair Kevin Warsh proposed reducing the number of FOMC meetings from eight per year to six — roughly one every two months. No decision was reached.
Around the Market
SK Hynix fell 9.75% in Seoul trading, then announced after the Korean close a $28.6 billion buyback and share cancellation covering roughly 3.3% of shares outstanding, while raising its 2025–2027 shareholder return target from “up to 50%” of free cash flow to “at least 50%.” The stock opened firm in US trading, but the strength failed to spread across the semiconductor complex.
Lowe’s dropped nearly 2% after issuing lackluster full-year revenue guidance — a less reassuring consumer read than Home Depot’s the previous session.
Iran escalated its rhetoric, with the military warning Gulf states against supporting American attacks. Reports indicated Tehran had considered striking US assets in Europe, including in Bulgaria and Cyprus. The UAE suspended all trade, commercial, and financial dealings with Iran for the remainder of the year, citing escalating tensions. Meanwhile CNN reported the US Navy has stepped up patrols in the Strait of Hormuz, somewhat weakening Iranian control — and crude appears to keep flowing.
Tariffs delivered a small relief: Trump paused the 50% levy on Canadian goods hours before it was due to take effect.
Gold broke above $4,490, up 3.54%.
Bottom Line
Wednesday was a rebound built on an intervention and a discovery. The Treasury’s decision to double long-bond buybacks pulled yields back from nineteen-year highs and gave equities room to breathe — but as LPL noted, it treats the symptom rather than the disease. Public debt above $40 trillion and interest costs above $1.17 trillion do not get solved by changing which maturities the government retires.
The Moderna result is the more durable story. A personalized mRNA cancer vaccine succeeding in Phase 3 is the kind of outcome that reshapes an industry’s expectations for a decade, and the healthcare sector’s record high suggests the market grasped that immediately. It also marks something notable about 2026’s rotation: the money leaving crowded AI trades has to go somewhere, and it keeps finding healthcare.
For semiconductors, the three-day sequence from Anthropic’s revenue to OpenAI’s shortfall has been clarifying. The trade’s fate now visibly depends on how fast the AI labs grow — not on chip demand in the abstract. Nvidia reports on August 26, and it will be the first major supplier to answer the labs directly.
Leave a Reply