Were We Wrong About Interest Rates? | George Gammon

Gammon says the AI investment structure ultimately depends on models such as OpenAI and Anthropic generating substantial profits, with Nvidia also exposed to that outcome. Separately, a speaker describes the AI boom as more dependent on structured credit than the dot-com bubble, warning that financing could stall if the junk market rejects deals.

Gammon says Oracle’s data-center project would require as much power as the entire state of New Mexico uses, underscoring the scale of infrastructure involved in AI investment. He also says the AI investment structure depends on models generating substantial profits, making the economics—not just the buildout—a key condition for the investment thesis.

Gammon argues that nominal GDP, historically correlated about 85% with the 10-year Treasury yield, is a better guide to rates than debt and deficits; with nominal GDP around 6.5%, he says the yield should be higher. He challenges claims that the relationship has entered a new regime, asking proponents to explain why historical patterns no longer apply.

Gammon attributes the pandemic-era price surge more to supply-chain disruptions and supply shocks than to U.S. M2 alone, and says money creation and destruction must be considered globally, including reduced circulation of existing money and credit. He also points to an energy shock—oil at $120 a barrel, record-high crack spreads, and California diesel at $10 a gallon—as an overlooked factor in rising rates.

Gammon says banks expanded their balance sheets to absorb Treasury supply as total debt rose from about $800 billion to $9 trillion between 1980 and 2007, while M2 grew from $1.5 trillion to $7.5 trillion. He argues that higher Treasury yields can attract more demand by improving bonds’ risk-reward relative to nominal GDP and equities, and says banks could buy heavily at 10% yields if nominal growth stayed around 5% or 6%.