Nvidia's $60B Quarter and What It Means

Ninety-six billion dollars passed through Nvidia's books in a single quarter this fall, and sixty billion of it stuck as pure profit (09:38). That is not simply a strong result. Host Jason Calacanis calls it the most profitable quarter any public company has ever posted, once one-time gains like asset sales are set aside (09:38). Nvidia also told investors to expect 70 percent revenue growth next year, far above the 45 percent Wall Street had modeled (09:44). For months, a slice of Wall Street had bet that AI spending, or capex, meaning capital expenditure on chips, data centers, and infrastructure, was a bubble about to pop. These numbers did not just answer that bet. They buried it, and even after an 8 percent stock jump, Nvidia still trades at only 12 times earnings, cheap by the standards of a company this dominant (30:45).

"This AI CapEx is gonna continue well into the future. It's got real legs." — David Sacks [30:33]

Salesforce told a related story. Its stock jumped more than 20 percent after an earnings beat and a raised full-year guidance of $46 billion (10:52). The company had been the poster child of what the hosts call the SaaSpocalypse, a fear that AI coding tools would let anyone rebuild their own business software and make traditional SaaS, or software as a service, obsolete. David Sacks argues that fear was overblown for one specific kind of software: the system of record, meaning the canonical database a company trusts as its single source of truth. Enterprises do not want to gamble their compliance and data integrity on something vibe-coded over a weekend, he says (18:11). What changed the picture is Salesforce's deal with Anthropic: Salesforce is letting Anthropic's AI assistant, Claude, act as the front-end interface, with full access to Salesforce's data and workflows (19:20). Sacks calls this a bet that pays off precisely because Salesforce still owns the underlying system agents have to query. David Friedberg, another host, draws the line differently: horizontal platforms like Salesforce, Workday, or Excel are hard to replace because they hold real data across a whole company, while vertical, industry-specific software mostly just manages workflow on top of data stored somewhere else, which makes it the more exposed target (25:08).

But the same week Nvidia and Salesforce delivered good news, a much less cheerful number was quietly getting worse: the price the United States government pays to borrow money.

The bill comes due

The 30-year Treasury yield hit 5.3 percent last week, a 19-year high (33:43). In response, Treasury Secretary Scott Bessent doubled the government's long-bond buyback program, essentially using public money to buy up Treasury bonds and push their prices up, which pushes yields down (34:07). His former mentor, investor Stan Druckenmiller, published a Wall Street Journal op-ed calling this the wrong fix: the real problem, Druckenmiller argued, is not the bond market's pricing but Congress's spending (34:07). Friedberg walks through why the arithmetic is unforgiving.

"For every 1% change in the interest rate, the US government has to pay 1.25% of GDP in excess interest each year." — David Friedberg [38:18]

The government's average cost of debt is currently 3.4 percent, well below the 5.2 percent now being charged on new 30-year bonds (37:54). That gap matters because roughly $10 trillion of existing federal debt has to be refinanced, meaning rolled over into new bonds, in the next twelve months (38:48). Every dollar of that debt that gets refinanced at today's higher rates raises the government's annual interest bill, which raises the deficit, which is already running at roughly $2 trillion a year and adding a trillion dollars in new debt roughly every five months (34:30). Friedberg's read is that Druckenmiller's op-ed functions as cover for Bessent: it tells markets that the yield problem is Congress's to solve, not the Treasury's, because no amount of bond buying can offset $10 trillion in refinancing at higher rates (39:15). Chamath frames the stakes starkly: if the 30-year yield reaches 6 percent, he calls that the start of a slow-moving crisis that could take years to play out, solvable only through a congressional deal to cut spending (44:32).

The hosts point to one place where different choices produced a visibly different outcome: housing. Home prices in Austin, Texas are down 27 percent from their 2022 peak, largely because the state allowed more building (59:46). Jason Calacanis, who moved from New York, Los Angeles, and San Francisco to Texas, argues that affordability followed directly from a policy choice to let supply grow, in contrast to cities that restrict new construction (59:12).

Who owns the cure

The most detailed segment of the episode has nothing to do with markets. Moderna's market value roughly tripled in two weeks, from about $20 billion to $60 billion, after positive results from what is being called a cancer vaccine (83:02). Friedberg, whose company works in DNA sequencing, walks through why that label is misleading and how the underlying science actually works. The idea, dating to the 1990s, is to find a neoantigen: a small protein unique to a specific patient's cancer, one the immune system will recognize as foreign (83:26). Doctors take a sample of a tumor, sequence its DNA to find the mutation that makes it unique, and use that sequence to build a matching protein (83:59, 85:16). In Moderna's version, instead of manufacturing the protein outside the body, doctors inject mRNA, the molecular instructions that tell a cell what protein to build, so the patient's own cells manufacture the neoantigen and the immune system learns to attack matching cancer cells throughout the body (86:04, 86:49). It is the same basic mechanism used in the COVID-19 vaccine, but aimed at cancer cells instead of a virus.

Friedberg's objection is not to the science, which he calls well understood after decades of trials largely funded by NIH and other public research dollars. It is to the price. Moderna is reportedly planning to charge around $500,000 for the treatment (88:56), while CAR T therapy, a related but distinct approach that reprograms a patient's own immune cells to hunt down cancer, already runs up to $1 million per treatment (94:23). Friedberg says clinics already offer a version of the neoantigen approach for a small fraction of that cost.

"It's a process... Why is Moderna saying that they're gonna charge $500,000 for this?" — David Friedberg [88:56]

His argument is that a therapeutic process built on public science should not become the exclusive property of one company through patents and FDA approval, especially when the core technique, sequence the tumor, build a matching protein, provoke the immune system, is neither secret nor new (89:00). He expects the approach to spread overseas at far lower cost as more providers adopt it (93:26).

Laid side by side, the three stories share a question the hosts never quite state outright: who gets to charge for a process that mostly already exists. Nvidia and Salesforce are charging premiums for owning the pipes that AI agents now have to run through. The federal government is discovering that trust, priced daily in the bond market, is not free either. And Moderna is charging half a million dollars for a delivery method wrapped around science the public already paid to develop. The specifics differ. The pattern, watching who ends up holding the toll booth, does not.