Good morning.

OpenAI has a date now, or something close to one. Chief financial officer Sarah Friar told employees at a Wednesday all-hands that the company "will be a public company in 2027," and possibly sooner if "our business continues to inflect," CNBC reported. She framed the listing as "not a finish line, it is a milestone, another fundraise," and pointed to the $122 billion the company raised in March. Asked whether it matters if Anthropic gets to the tape first, she said OpenAI is "running our own race."

A $122 billion round, and the IPO is the fundraise. Somewhere a Series A founder just quietly closed their laptop.

Biotech

For four years Moderna was the stock that had already peaked. On Wednesday it remembered it owned a platform.

Moderna (MRNA) closed at $174.38, up 176.97%, by a wide margin the largest single-day gain in the company's history. The prior record was a 27.81% jump in February 2020, per Benzinga. Partner Merck (MRK) closed up 12.60%.

The trigger was topline data. Merck and Moderna said their Phase 3 INTerpath-001 trial met its primary endpoint of recurrence-free survival and its key secondary endpoint of distant metastasis-free survival, testing intismeran autogene plus Keytruda against Keytruda alone in patients with completely resected stage IIB to IV melanoma.

This is not a drug in the way the word usually works. A surgeon removes the tumor, a lab sequences it to find the mutations specific to that one patient, and an mRNA therapy is manufactured for that person alone, then given alongside Merck's immunotherapy. The companies describe it as the first positive Phase 3 readout for an individualized neoantigen therapy and for an mRNA-based cancer therapy of any kind.

The supporting evidence has been building for three years. Five-year follow-up from the Phase 2b KEYNOTE-942 study, presented at this year's ASCO meeting, showed a 49% reduction in the risk of recurrence or death and a 59% reduction in the risk of distant metastasis or death versus Keytruda alone, according to the companies' release. Not everyone was positioned for it to work: Bloomberg reported short sellers were sitting on roughly $5.5 billion of losses by the close.

Remission Statement: No efficacy magnitude has been released yet, and that number decides everything downstream: the label, the payer conversation, and whether a therapy that takes weeks to build for one person can be manufactured at national scale. Melanoma was also the friendliest possible venue for this technique, which makes the read-across to other tumors a hypothesis rather than a result. What is settled is narrower and still significant. The mRNA platform has now cleared a Phase 3 bar in oncology, which is the thing the market had spent four years pricing as unlikely.

Artificial Intelligence

Stripe spent 15 years taking a cut of money moving around. Now it would like a cut of tokens moving around.

Stripe agreed Wednesday to acquire OpenRouter, the gateway that lets corporate customers route their AI workloads across roughly 400 models from Anthropic, Google, OpenAI and others, and meter what all of it costs. Stripe declined to discuss terms. CNBC noted that The New York Times put the price at $7.5 billion.

The number is the story. OpenRouter is a young company sitting between enterprises and their model vendors, and it just got valued like infrastructure.

Every AI narrative so far has been about who builds the models. This deal is a bet on who meters them. As frontier capability converges and per-token prices fall, the model itself starts to look less like a moat and more like a commodity input, which makes the switching layer, the layer that decides which model gets which request at what price, more valuable rather than less.

Toll Position: Token spend has quietly become a real line item at large companies, and nobody has been managing it particularly well. Stripe's argument is that inference is just another payment rail, with the same problems of routing, reconciliation and cost control it already solved once for card transactions. If that framing holds, the interesting question stops being which lab wins and starts being who sits at the interchange. Stripe has now paid a reported $7.5 billion to find out.

Economics

When the 30-year is smoldering at a two-decade high, you do not wait for the smoke to clear on its own.

The Treasury said Wednesday it will double the size of its liquidity-support buyback operations in longer-dated nominal coupon securities, to at least $4 billion per operation from $2 billion, covering the 10-to-20-year and 20-to-30-year sectors from Sept. 9 through Nov. 4.

Bonds took the hint immediately, and equities took it second-hand. This came a day after the 30-year yield closed at its highest level in nearly two decades, TheStreet reported.

Liquidity-support buybacks are a plumbing tool. The Treasury repurchases older, thinly traded off-the-run issues to tighten spreads and improve trading conditions. They are not new financing, they do not retire net debt, and $2 billion of extra capacity per operation is a rounding error against a market this size. The size is not what moved anything. The willingness was.

By the numbers: the 30-year yield fell almost 10 basis points to about 5.187% and the 10-year slipped 4.9 basis points to 4.65%, per CNBC's read of the session. Equities snapped a three-day losing streak, with the S&P 500 (SPX) up 0.21% to 7,708, the Dow up 0.22% to 53,463 and the Nasdaq Composite up 0.16% to 26,331, TheStreet reported. Ed Ludlow opened Wednesday's Bloomberg Tech by calling the buyback plan the morning's big catalyst.

Yield Sign: The awkward part is that a buyback aimed at liquidity gets read as a buyback aimed at yields, and once the market reads it that way, the reading is self-fulfilling until it isn't. Long-end yields at multi-decade highs are the binding constraint on everything from mortgages to the hyperscaler bond calendar. Wednesday bought some room. It did not buy a solution, and the operations do not even begin until September.

Refund Wars: Lowe's (LOW) trimmed full-year sales guidance to the $92 billion bottom of its range and now sees comparable sales flat after a quarter in which comps rose just 0.2% and transactions fell 2.1%, per CNBC, while Target (TGT) posted comps up 3.8% on traffic up 3.6% with $994 million of tariff refunds in the quarter; Bloomberg Intelligence's Lindsay Dutch said on Wednesday's episode that rivals took June refund money and spent it on July price cuts, and Lowe's lost share.

Memory Loss: SK Hynix's board approved a 40 trillion won buyback and cancellation of up to 24.07 million shares, about 3.3% of those outstanding, running from today through Nov. 19, after the stock shed roughly half its value since June, the Korea JoongAng Daily reported.

Warrant Officer: Marvell (MRVL) granted Google a warrant to buy almost 59 million shares at $206.58 apiece, worth about $12.2 billion, vesting in tranches tied to every $500 million of chips Google buys, BNN Bloomberg reported; the stock popped about 10%, per CNBC.

A therapy that has to be built one patient at a time nearly tripled a large-cap stock in six and a half hours. A company that decides which chatbot answers your API call sold for a reported $7.5 billion. And the Treasury said it would buy back some old bonds, and the entire equity market exhaled. Nothing about Wednesday rhymed. It all cleared anyway.

That's the tape. We'll see you at the open. — AllThingsWallSt