Debt Spiral or NEW Golden Age

All-In Podcast, Episode 261, "Debt Spiral or NEW Golden Age?" hosted by Jason Calacanis, Chamath Palihapitiya, David Sacks, and David Friedberg. Duration: 73 minutes. This episode swings between genuine alarm about America's fiscal trajectory and infectious optimism about an AI-driven economic boom, with the besties landing on very different sides of whether we are headed for a debt death spiral or the beginning of a new golden age.

Section 1. AI Acceleration and the Agent Revolution

The episode kicks off with a bang. Jason opens with a Harvard Business Review study from two UC Berkeley researchers who spent eight months embedded at a 200-person tech company studying how AI tools affect work. The findings are counterintuitive. Employees who used AI worked faster, took on a broader scope of tasks, and actually extended their work into more hours of the day. They felt more productive but also reported more stress and burnout. It is not replacing workers. It is supercharging them.

Sacks jumps right in, reminding everyone that his most contrarian prediction for the year was that AI would increase demand for knowledge workers, not put them out of business. He points to the study as evidence. Workers are getting upleveled, offloading menial tasks to AI and doing more purposeful, meaningful work. He references Jensen Huang's framing of moving from task-based jobs to purpose-based jobs. Sacks is particularly excited about the bottom-up adoption wave he sees coming. He argues that while large enterprises are running slow top-down AI transformation initiatives with RFPs and committees, the real change will come from early adopter employees bringing consumerized AI tools into the workplace, just like SaaS tools spread from the bottom up a decade ago. He cites a viral article by Matt Schumer called "Something Big Is Happening" about this career opportunity for AI-native workers.

Jason then goes full evangelist mode. He talks about a tweet he posted that got two million views telling people laid off by Amazon or Microsoft to learn OpenClaw and automate their previous jobs. He describes his venture firm as being in the middle of a total transformation. They have what he calls replicants, essentially AI agent personas, each with their own Notion account, Slack account, Google Docs, and email. He says four out of his twenty employees are focused on agent work, and their leverage is ten to twenty times the other sixteen. He describes agents that go through podcasts finding the best moments, clip them automatically, analyze YouTube and TikTok stats, and suggest strategies for making content go more viral. And here is where it gets wild. Jason says they have built something called "OpenClaw Ultron," which is one meta-replicant that manages the other four agents, checks their work, talks to them all day, and summarizes everything. He also mentions taking about thirty percent of what his Athena virtual assistant was doing and handing it to a replicant.

Chamath pivots the conversation in a fascinating direction with a question nobody else is really asking yet: "Is on-prem the new cloud?" He argues that since 2008, everyone migrated to the cloud for economies of scale, and that is how AWS and GCP built massive businesses. But now, with AI, companies are leaking their proprietary data, their strategy documents, their models, every prompt and response, back to the model providers. He points to a legal ruling that confirmed there is no attorney-client privilege for data processed through cloud AI tools. His conclusion is stark. You need AI to survive, but if you use the tools at a public endpoint, you give up all control, all security, all confidentiality. The only solution might be to swing the pendulum all the way back to private provision networks. On-prem is back.

The token budget discussion is one of the most forward-looking moments in the episode. Chamath reveals that his company hit three hundred dollars a day per agent using the Claude API, which works out to about a hundred thousand dollars a year per agent. He is now asking what token budget he is willing to give his best developers, and when you aggregate across all employees, you can clearly see a trend where each person needs to be at least two times as productive as another employee to justify the cost. Jason calls it a trend nobody else is talking about: "When do tokens outpace the salary of the employee? Because you're about to hit it. I'm about to hit it." Chamath says superstar developers are probably already there, while rank-and-file employees are at ten to twenty percent. He suspects token costs will eventually come down dramatically thanks to Nvidia, Groq, Google, and AMD all working to push costs lower, but the trend and the incentives around confidentiality remain.

Section 2. Prediction Markets Hit Critical Mass

The Super Bowl weekend turned prediction markets into a mainstream phenomenon. Jason lays out the numbers. More than a billion dollars was bet on Kalshi, seven hundred million on Polymarket, nearly two billion total in wagering. But the real story is about what happens when insider information meets open betting platforms. A day-old anonymous Polymarket account correctly predicted seventeen out of twenty halftime show bets, including the surprise appearances by Lady Gaga and Ricky Martin, though it only profited seventeen thousand dollars. Another account created less than twenty-four hours before the game correctly bet on Bad Bunny's set list.

Then Jason drops a bombshell from the Wall Street Journal. Israeli soldiers were accused of using Polymarket to bet on military strikes. Israel arrested several people, including army reservists, for allegedly using classified information to place bets on Israeli military operations. The account, hilariously named Rico Suave 666, racked in more than a hundred and fifty thousand in winnings before going dormant for six months, then resumed trading last month betting on when Israel would strike Iran. Jason quips that Rico Suave 666 was also the alias Chamath used in Vegas at his hotel.

Friedberg frames the issue as a marketplace problem. He shows that a few accounts with huge amounts of money make almost all the profits, while many small accounts burn through their capital quickly. The sharps eat up all the liquidity. Over time, the platform could burn through its customers if retail bettors keep getting destroyed by insiders.

Chamath delivers the most provocative take. He defines the two types in betting markets, sharps who know what is going to happen and squares who are grist for the mill, then draws a parallel to the securities market. He pulls up a chart of Warren Buffett's returns before and after Reg FD was introduced in 2000, the regulation that made it illegal for company executives to share material information with select investors. Before Reg FD, Buffett's returns were double the market. After Reg FD, when everyone had the same information, his returns dropped to market returns. Zero alpha. Chamath's conclusion is blunt. Markets thrive when there is asymmetry. Prediction markets today, unless they are regulated out of existence, will look like the stock market pre-Reg FD. There is nothing you can do except choose not to bet.

Section 3. The CBO Report and the Debt Death Spiral

This is the meaty center of the episode. The Congressional Budget Office released its long-term budget forecast on February 11th, and Friedberg, wearing his self-proclaimed Doctor Doom hat, is alarmed. The numbers are sobering. The 2026 deficit is 1.9 trillion, nearly six percent of GDP, well above the three percent target Treasury Secretary Bessent described on the podcast. Social Security trust fund runs out in 2032, one year earlier than expected. Debt grows from 31 trillion today to 56 trillion in 2036, averaging 2.5 trillion per year in new debt. The US is at 120 percent debt to GDP, projected to hit 135 percent by 2036.

Friedberg zeroes in on the interest rate assumption. The CBO models short-term rates at about 3.1 percent, but if rates climb closer to five percent, that adds another 650 billion dollars a year in interest expense, taking total interest payments to nearly two trillion a year. Because the country is running a deficit, that new interest expense increases the debt every year, creating the death spiral he has been warning about. But his biggest worry is actually state and local obligations. California alone has nearly a trillion dollars in unfunded pension obligations. If Democrats win the House and the presidency in 2028, Friedberg predicts a federalization of those obligations, the federal government bailing out state and local pensions. He calls this not the straw but the concrete that breaks the camel's back.

Sacks pushes back with the growth argument. He points out that the CBO projects only 2.2 percent real GDP growth for 2026, which he calls unrealistically low given that Q3 last year was over four percent and Q4 was over five percent. All the besties predicted five percent plus growth for the year. Sacks says if you believe the massive AI capex investment will pay off, growth rates could be much higher, and that is ultimately the way out of the debt spiral. He pulls up two key charts: federal net outlays as a percent of GDP and federal tax receipts as a percent of GDP. Historically, receipts bounce around 17 percent and outlays around 20 percent. Before COVID, we were close to that. Then COVID spending pushed outlays to 30 percent, and we have never fully recovered, currently trending around 23 percent. His prescription is simple: freeze federal spending until the economy grows enough that spending as a percent of GDP drops back to 20 percent. He also notes that federal employment is at its lowest level since 1966, dropping from roughly three million to under 2.7 million employees under Trump, a cut of over 300,000.

Chamath takes the longest historical view of anyone. He pulls up 300 years of debt-to-GDP data for the world's largest economies and argues that the trend since 1700 has been up and to the right. Wars spike it up then bring it down, but the general direction is clear. His key insight is that debt-to-GDP moves in unison across countries. If the entire world is at 200 or 300 percent, nothing really changes that much. The real problem is if one country decouples. He is blunt about Congress. He notes that even the most conservative Congress in United States history has failed to solidify the spending cuts that DOGE identified. If they cannot do it, a future Democratic Congress certainly will not. His practical advice: own real durable assets like gold because the underlying currencies of all these economies will erode.

Jason, characteristically, brings it back to the jobs data. Employment is strong. The January report added 172,000 new private sector jobs, blowing away expectations of 70,000, while the government shed 42,000 jobs, bringing unemployment down to 4.3 percent. He highlights that 615,000 new private sector jobs have been created during Trump's second term while over 300,000 government jobs were cut. He sees a potential wildcard: Trump raising the federal minimum wage from seven dollars an hour as a populist move, arguing it could help with affordability concerns ahead of the midterms. Sacks pushes back hard, citing the standard economic finding that minimum wage increases create unemployment in lower-wage segments and accelerate automation. Jason counters with real-world examples from Seattle, San Francisco, Australia, and Scandinavia where higher minimum wages correlate with higher happiness despite ten to twenty percent higher prices.

Sacks brings it home with what becomes the title-worthy moment: "I suspect we'll look back on this time period as the beginning of a new golden age." He compares the current moment to the late 1990s, when incredible economic growth was happening but all anyone talked about was the Clinton-Lewinsky scandal. The parallel today is that everyone is doom-scrolling about Epstein or the CBO report while a genuine economic boom, with 600 billion in hyperscaler capex alone providing a two percent tailwind to GDP, is unfolding right in front of us. Jason agrees enthusiastically. "We're going to print six percent."

Section 4. The Immigration Enforcement Debate

This segment produces some of the most heated back-and-forth in the episode. Jason argues that the most effective immigration enforcement would be going after employers, not conducting raids in cities. He points out that the top two employment categories for undocumented workers are construction at number one and leisure and hospitality at number two, with about two and a half million people working in those categories. His proposal is straightforward: ICE should surveil construction sites, take pictures of workers, then go to the business owners and demand payroll records. If workers are being paid off the books, the businesses face massive fines for tax evasion.

The other three push back. Friedberg asks how you can surveil someone with a camera and figure out if they are illegal. Sacks says if Stephen Miller were doing this, Jason would call him a fascist. Jason fires back, "Once again, incorrect," and insists he has publicly stated on the podcast and on Twitter that this is exactly what Miller should do. He cites a 2017 Justice Department case that recovered 95 million dollars, the largest payment ever in an immigration case. Jason's broader argument is that if there were no off-the-books jobs available, the economic incentive to cross illegally would disappear. He cites an LA Times survey showing 75 percent of immigrants come for better job opportunities, not benefits or crime. Sacks counters by asking about government benefits as an incentive. Jason says that is far down the list. It is a lively, genuine disagreement that shows the four hosts can have real friction without it becoming toxic.

Section 5. Ferrari Goes Electric and the Death of Car Culture

The episode shifts gears, literally, to Ferrari's first all-electric vehicle. Jason walks through the specs: over one thousand horsepower, four electric motors, zero to sixty in under 2.5 seconds, 330-mile range, and at 5,100 pounds, the heaviest Ferrari ever made compared to the iconic F40 at just 3,000 pounds. The interior was designed with involvement from former Apple design chief Jony Ive and his partner Marc Newson. The key is a beautiful square glass piece like an iPhone that you insert, and the yellow Ferrari color drains out and flows into the shifter. Starting the car is like firing up a jet, with a twist-and-press launch button that turns the whole cockpit Ferrari orange.

Sacks likes the interior, calling it a good compromise between Tesla's all-glass cockpit and a traditional analog Ferrari. The buttons are tactile and make pleasing sounds, allowing for muscle memory. But he hates the projected exterior, calling it terrible and saying it looks like a Corvette or a Pontiac Firebird Trans Am. "A Ferrari should look swoopier, curvier." Jason asks when Sacks last actually drove himself. Sacks admits that full self-driving has made him a driver again because Uber takes forever in Texas.

Chamath delivers a bittersweet meditation on car culture. He says there is something unique about the Ferrari experience that is different from every other car, but autonomy and FSD are going to change everything. Driving will become like horse racing, something that happens in smaller and smaller places, less and less often. The American car culture, driving cross-country on vacation, the freedom, the interstate highway system, all of it is going to change. Insurance costs for manual driving will become prohibitive. In the future, there will be high-end cars like Ferrari where you pay for the experience and the luxury, and then everyone else will use FSD or Waymo.

The episode closes with a delightful tangent about luxury minivans. Jason saw a Lexus LM in Abu Dhabi and is obsessed with it. The doors open to reveal first-class airline seats, a full privacy divider between passengers and the driver, a rising monitor, and what Jason compares to Etihad first-class seating. Friedberg is equally enthusiastic. Neither vehicle, the Lexus LM nor the Toyota Alphard, is available in the United States, which drives the besties crazy. They are the number one chauffeur-driven cars in China, Singapore, and the Middle East.

Key Takeaways

First, AI is not replacing workers. It is making them do more work, faster, at a broader scope. The UC Berkeley study confirms that early adopters of AI tools will have massive career advantages, and enterprise adoption will come from the bottom up, not top-down transformation initiatives.

Second, the token budget problem is real and imminent. Chamath's company is spending a hundred thousand dollars a year per agent on API costs. When token costs exceed employee salaries, companies will need to rethink how they allocate AI resources. On-prem computing may make a comeback as enterprises grapple with data leakage and confidentiality concerns.

Third, prediction markets have gone mainstream with nearly two billion dollars bet during the Super Bowl, but insider information is a structural problem that may be impossible to regulate without killing the platforms entirely. Chamath's Buffett-Reg FD comparison is the clearest illustration of why information asymmetry is the lifeblood of these markets.

Fourth, the CBO report paints a grim fiscal picture with the deficit at 1.9 trillion and debt projected to hit 56 trillion by 2036. But the besties split on how worried to be. Friedberg sees a death spiral, especially if state pension obligations get federalized. Sacks says the CBO is using unrealistically low growth assumptions and that we may be at the beginning of a new golden age. Chamath argues that debt-to-GDP has been rising globally for 300 years and what really matters is owning durable assets as currencies erode.

Fifth, the strongest economy in a generation is happening right now with 172,000 private sector jobs added in January, unemployment at 4.3 percent, over 600 billion in hyperscaler AI capex, and GDP growth potentially hitting five to six percent. The question is whether Americans will notice the boom or keep doom-scrolling past it.