Nvidia's Best Quarter Ever, and a 12x Multiple
The market isn't doubting $60 billion in net income — it's betting competition is coming. The most profitable quarter in history paired with the cheapest valuation in history is the most counterintuitive thing in this episode.
The video won't play here. Listen to the audio instead:
The argument · tap a timestamp to hear it
The best earnings ever come with the cheapest valuation ever
Nvidia did $96.2 billion in revenue this quarter, more than double year over year, against Wall Street expectations of just $92 billion; more important is the guidance — 70% growth next year, when the market had expected only 45%. Net income of $60 billion makes it the most profitable core-business quarter of any public company in history. Yet even after the stock rose 8% that day, Nvidia trades at just 12 times earnings. The reason isn't that the market disbelieves the numbers — it's that the market believes competition is coming. That combination is the most counterintuitive thing in this episode: the strongest earnings ever, paired with the cheapest valuation ever.
— Jason CalacanisStage two of AI is giving the brain hands and eyes
Chamath splits AI into stages: stage one is the model, essentially a brain; stage two is the harness, meaning agents, essentially giving that brain eyes, hands, a notepad and a keyboard. But he judges stage two isn't enough — what enterprises are really missing is making that autonomous agent more informed and more contextual, so it can actually do the work of a lawyer, a support rep, a salesperson. Whoever owns that context owns the next stage. That's the underlying logic behind his May call that Salesforce was oversold and his insistence on staying long.
— Chamath PalihapitiyaAgents are Salesforce's power users
Benioff himself raised an underrated point: Salesforce contains a lot of ‘trap value’ — the platform's features are so vast that ordinary users can never find them, but an agent can. The agent will proactively tell the user ‘I see these actions in Salesforce that match what you asked for, want me to run them’, and once the user clicks ‘always allow’, the agent can unlock the full power of Salesforce. That means enterprises no longer need to hire a fleet of expensive consultants to figure Salesforce out. Freeberg's read: Benioff knew that wiring Salesforce into Claude carried disintermediation risk and still handed over the customer relationship, because he knows what customers want is to build agents inside AI and let those agents run across every SaaS platform.
— David FriedbergThe SaaS apocalypse is a product of naive extrapolation
Freeberg reduces this whole narrative cycle to one error: naive extrapolation. The SaaS apocalypse argument runs ‘AI agents will write code, therefore all software goes to zero’; the jobs apocalypse argument runs ‘AI can do knowledge work, therefore all knowledge workers lose their jobs’. Both ignore that people can turn around and use the tools. His conclusion is blunt: those who adopt the technology live, those who don't die. He also notes that Leopold's SaaS short may not be a judgment problem but a leverage problem — to get enough leverage, you can only short the most liquid large caps.
— David FriedbergNvidia is grabbing open source's distribution layer
Nvidia reportedly acquired Hugging Face for $12 billion and did an acqui-hire of Poolside for $6 billion, roughly $20 billion combined, less than 1% of Nvidia's market cap. Chamath reads this as Jensen's counterpunch after being provoked by Sam Altman's AMD deal and by in-house inference chips — he wants to own the distribution layer for open source. The bigger judgment is that the boundary between customer and supplier is melting: hyperscalers build their own chips, so Nvidia builds its own models, runs a cloud, serves APIs — eventually every company will do everything.
— Chamath PalihapitiyaBessent can't save the long end
Over the next 12 months the US federal government has to refinance $10 trillion of maturing debt, which means going back to the Treasury market to sell new bonds and borrow new money, and borrowing new money costs more, which makes the deficit bigger. So the speaker argues Bessent cannot materially affect the long end no matter what he does — the most buying he can muster is $1 trillion, while he still has to sell $10 trillion. The point of Druck's memo isn't prediction, it's cover for Bessent: push responsibility back onto Congress and the White House, and say this isn't the Treasury secretary's job — it's the people who control the budget who have to cut spending.
30-year at 6% is where the death spiral starts
The speaker's threshold call: if you see the 30-year hit 6%, that's the start of the death spiral. Not an immediate explosion, no reason to panic, but the beginning of extreme pain, and that pain lasts for years. The only fix is getting the budget under control, and that's an act of Congress, and it has to happen, this year or next, it's only a matter of time. Sacks adds that no one controls spending: 435 representatives and 100 senators each have a vote, plus the president, and no single person controls spending — the power of the purse originates in the House, and the president has no line item veto.
The cancer vaccine is a process, not a new drug
Moderna went from a $20 billion market cap to $60 billion in two weeks on positive data for a so-called cancer vaccine. But Freeberg points out this neo antigen immunotherapy has existed since the late 1990s: take a tumor sample, sequence its DNA, find the mutation sequences unique to that cancer, then build proteins or mRNA from them and inject them to activate immune cells to attack. Melanoma is the biggest target because UV-induced mutations are distinctive enough. It's more like a process than a traditional drug, and the process is customized per person. What bothers him most is that everyone treats this as a one-of-a-kind breakthrough — people have already gone to Montana, printed peptides from their own cancer sequences, made their own neo antigens, injected them and been cured; today you can pay $50,000 to have it done, lots of clinics do it, and it isn't an FDA-approved drug — while Moderna plans to charge $500,000 for the same process.
— FreebergIn their own words · checked verbatim
if the first phase of AI were models and think of like a model as a brain, the second phase of AI is harnesses, right? agents, which is the analogous equivalent of giving a brain a pair of eyes and hands and a notebook for memory and a keyboard to type things on.
Chamath Palihapitiya12:05
the simple extrapolation here was AI agents can code therefore all software is going to zero.
David Friedberg27:16
The other narrative that's getting shredded today is that this AI capex is a bubble.
David Friedberg30:16
this idea that you have this very delineated separation of customer and supplier I think is melting away.
Chamath Palihapitiya32:17
And as that yield goes up, the very simple rubric is trust is going down. Yield goes up, trust goes down.
Because if you see yields, if you see the third year at 6%. It is the beginning of a death spiral. It's not going to be immediate, okay? So, don't freak out, but it is the beginning of some extreme pain. And that pain will last years.
I think not disclosing it is the betrayal. If you want to write, if you want to give me your opinion as interpreted through the AI magic guessing word box, you need to tell me.
You can pay someone $50,000 to do this for you today.
Freeberg1:28:00
Figures
| Nvidia quarterly net income | $60 billion | 9:03 |
| Nvidia's revenue growth guidance for next year | 70% (Wall Street expected 45%) | 9:03 |
| US Treasuries to be refinanced over the next 12 months | $10 trillion | 39:21 |
| Debt growth vs GDP growth | 7% vs 2%–4% | 41:24 |
| Moderna market cap | from $20 billion to $60 billion | 1:22:56 |
Glossary
- harness
- The layer that wires a model to tools, memory and an operating interface, so the brain can actually act.
- acqui-hire
- An acquisition made mainly to get the team rather than the product.
- yield curve suppression
- Government bond buying that pushes bond prices up and implied yields down, lowering the cost of new issuance.
- neo antigen
- A unique protein fragment produced by a tumor's DNA mutations, belonging only to that cancer, that the immune system can recognize.
- CAR-T
- Chimeric antigen receptor T-cell therapy: take a patient's T cells, reprogram them, and infuse them back to hunt cancer-specific proteins.
How to listen
Investors and founders watching AI compute and SaaS valuations, plus macro traders trying to get straight on US long-end rates and fiscal constraints.
The Ukraine war segment (1:17:53) and the AI writing disclosure fight (1:01:39) can be skipped.