The Rate Scare, According to the Blair Witch
Rising interest rates may matter less to an AI-driven economy than conventional macro analysis suggests, according to investor Jordi Visser. He argues that the more important question is whether higher yields can actually slow growth when nominal GDP is strong and AI investment is expanding rapidly.
Visser also points to Treasury policy as evidence that the government has incentives to resist materially higher long-term rates because of the growing interest cost attached to a large federal deficit.
Visser's core argument is that the rate fear is an artifact of an older economy (one built on housing, autos, and the 2008 crisis) that no longer describes where growth comes from. Frontier AI labs like OpenAI and Anthropic, which he describes as an enterprise-facing duopoly with "enormous" margins, are largely insulated from borrowing costs: on his analysis, a 200-basis-point rise in 10-year yields would move their margins only from about 70% to 69%.
To make the point land, he asked ChatGPT for an analogy and got "The Blair Witch Project": a manipulated story, hyped as real, in which everyone braces for disaster that never arrives. He suggests the widely shared chart of 20-year yield highs functions the same way on X.
His evidence runs deeper than margins. The TLT, the ETF tracking 20-plus-year Treasuries, is down only about 3% in total return this year, versus a decline of over 30% in 2022, yet investors are treating this as a crisis. Meanwhile, he notes, the Nikkei has climbed roughly 70% since last May even as 30-year Japanese yields made all-time highs.
For Visser, both data points suggest people are extrapolating history into a regime of exponential demand where earnings are growing around 30%.
What Could Derail the Thesis
- Significant shocks, not noiseVisser says a significant change would be required; oil spikes and short-term events would not derail the AI story.
- Cyber attacks compound doubtHe expects cyber attacks that "freak people out" and worsen the questioning of terminal value beyond three years.
- Multiple compression riskHe is not bullish on both earnings and multiples, he expects multiples to keep compressing, with periodic 20% drawdowns.
- Reward-free fundamentalsEven crushing guidance from Nvidia went unrewarded for days, a market-structure warning he dates to July of this year.
Speed Matters, and So Does the K
The host pushed back hard: AI companies may employ tens of thousands, but there are 300-plus million Americans, a housing market, business loans, and R&D. Visser's reply narrowed the claim rather than abandoning it.
He is not saying rate rises touch nothing, he concedes asset prices could fall, the S&P 500 multiple has already compressed this year, and a K-shaped economy would not improve, since "just because Anthropic and OpenAI still have demand from enterprises, it doesn't mean we wouldn't lose tons of small businesses." His position is that a slow move from about 4.75% to 5.25% would not push nominal GDP down much.
Timing is central to his case. If yields jumped from 5% to 7% overnight, he says, small businesses would fail and job losses would follow; gradual moves would not. He points to 2022, when aggressive hikes that "everyone thought" would cause a recession did not.
And he invokes a darker rationale: rates going higher "cleanse the system" of zombie businesses kept alive by cheap borrowing, though he stresses he is describing, not endorsing, and that Bessent's yen interventions, buyback expansions, and refunding-announcement language all signal a government leaning the other way.
Even his bullishness carries warnings. Nvidia guided next-year revenue growth far above the street's 44% estimate (Visser reads the real message as closer to 100%) and the stock still fell for two days afterward. He takes that as evidence that July marked a structural shift: fundamental investors are no longer paid for being right, which he attributes to AI agents trading with more information and less regard for being rewarded.
He expects 7% annual stock returns against 20% earnings growth, with periodic 20% drawdowns, which is why he prefers assets "that agents embrace but humans don't, " like scarce crypto.
The Deleveraging Nobody Expects
The host arrived with his "stat of the year": Bitcoin, gold, and land up about 170% over five years against roughly 70% for the S&P 500, with national debt at $40 trillion and annual interest expense above $1 trillion. Visser answered with a take he says he has never heard elsewhere.
Global money supply sits around $100 trillion against roughly $700 trillion in asset values, a 7:1 leverage ratio born of the fractional reserve system, which he frames as society's insurance against repeating the Great Depression his grandmother lived through.
Most people, he argues, assume deleveraging must happen through asset prices collapsing back toward the money supply. His contrarian view is the reverse: the money supply keeps growing, but asset values stop growing as fast (or drift lower) because AI-driven deflation erodes what things are worth. If humanoids can build a house cheaply and materials fall in value, he asks, how does the value of an existing house rise?
He expects this to "start in 3, 5 years" as physical AI becomes real, pointing to Tesla's Cybercab fleets being marketed directly to buyers.
Doubt as Fuel: The Third Wave Thesis
Bitcoin tagged just over 82, 000 (a level Visser had flagged) before settling near 79, 000–80, 000, and his framing of what comes next is explicitly technical. From an Elliott Wave perspective, he believes crypto is "entering a third wave... the most powerful wave in crypto, " and he insists the prevailing doubt is a feature, not a bug: the energy of a third wave requires skepticism as it rises.
He is heartened, he says, by former believers quietly giving up, including one strong believer at a dinner that week and Mark Cuban saying on television that he is "pretty much done with it."
His structural view of the space has consolidated into three roles: Bitcoin as collateral and store of value, Ethereum as the trust layer for building, and Solana for speed. The catalyst, in his telling, will not be falling rates but profits and the economy looking through rates while the government manages the K-shaped fallout (the "Jeff Booth theory" he calls real: printing money to offset K-shape damage, incompletely, with crypto benefiting from the manipulation.
He cites analyst Stan Druckenmiller-style reasoning that 10-year yields "should be more like 8%") about 200 basis points above nominal GDP, as they ran before 1997, implying suppressed rates that should, in his view, favor crypto.
What changed his conviction most is agents. Crypto, he argues, is no longer a binary bet on debasement: agents need to transact in seconds and cannot operate on legacy rails, and tokenization is already official policy elsewhere, South Korea has announced full tokenization by February 2027.
At a $3–4 trillion market cap against roughly $700 trillion in traditional assets, he calls crypto "so underowned" and says a truly diversified investor can no longer justify a 0% weighting.
Dispatches From the Agentic Frontier
The host framed his own week as a trip to the frontier, "like Daniel Boone coming back to tell people what I found." A YC founder he backs had sold a SaaS data product to humans; ten weeks ago he flipped the model so that only AI agents could sign up. He now has thousands of agents paying as they go, averaging over $100 per year each, with revenue that looks like "every investor's dream."
His arithmetic: a million agents would mean roughly $100 million in revenue, ten million agents a billion, and a million agents, he reasoned, is far easier to reach than a million people, since each person runs dozens of agents.
Visser sees the same shift inside his own operations. He runs daily standups for his agents (each reporting yesterday's work, today's plan, and blockers) and noticed the reports read identically to human ones. He even held a formal staff meeting with agents, led by a chief of staff he named Jarvis, because he wanted the same diversity of opinions he prized managing people at Morgan Stanley.
His conclusion: friction is still good (compute demand outstrips supply and data centers slip, which slows a potentially dystopian acceleration) but he expects agents to change "the entire economy over the course of the next 12 months. Not 12 years, 12 months."