AI's World-Changing Breakthroughs, Bond Market Mayhem & US-China Power Moves | The Tom Bilyeu Show

The 10-year Treasury yield closed at 5.22%, a level Tom Bilyeu said had not been seen in nearly two decades, while the 30-year yield reached 5.5%, last seen in 2004, and mortgage rates were above 7%. The 10-year yield’s Wednesday jump was its largest single-day rise since the tariff turmoil of April 2025. Tom warned that high borrowing costs were freezing housing and could expose the wider economy if AI investment fails to deliver. Oil above $103, record diesel prices, and August inflation of 3.4% added pressure; a purchasing-managers survey showed business activity at a five-year high as input prices rose at their fastest pace in four years.

Tom argued that rate hikes cannot reopen the Strait of Hormuz and questioned treating war-driven energy inflation as evidence of excessive demand. Higher rates could suppress spending and inflation, he said, but might first damage the economy; markets were pricing a 75% chance of another Fed hike in October. A speaker estimated that roughly $10 trillion of debt would roll over in the next 12 months, potentially doubling interest payments, and speculated that Kevin Warsh’s selection may reflect a desire to lower debt costs. The speaker said Treasury Secretary Scott Bessent had doubled long-bond buybacks to $4 billion per operation, but the rally lasted only a day, and described shifting issuance toward lower-rate two-year debt as a strategy that depends on inflation staying controlled.

Oracle’s shares fell about 5–6% after progress on its New Mexican data center slowed. The project faces at least a six-month pause after Oracle cited circumstances outside its control in a force majeure filing; the company expects revenue later than planned and may pause some debt payments. Blue Owl, involved in the project’s debt, also took a stock-price hit. A speaker said difficulty securing gas and obtaining permits could keep data centers from earning revenue in time to service debt. He warned that AI infrastructure is being financed with debt at a scale relative to company revenue exceeding earlier railroad and internet buildouts, and that investors may be treating the debt as safe based on anticipated future cash flows. Whether Oracle signals systemic trouble remains unclear, but speakers warned that weakening confidence could shut off overnight funding; a Fed and Treasury rescue might stabilize markets while adding inflationary pressure.

A reported Anthropic experiment used Claude and roughly 950 AI agents for 21 hours to search a large DNA database, identifying a previously unknown bacterial molecular system called ART, or Array Associated Reverse Transcriptases. A speaker said it resembles CRISPR, but its function and medical potential have not been confirmed. By comparison, the speaker said CRISPR can be programmed to target and cut DNA and has advanced treatments for sickle cell disease and cancer. The episode presented ART as a potentially important discovery, not a proven treatment, and noted that promising research can stall without FDA approval or investment, particularly when rare diseases offer limited financial incentives.

Speakers argued that falling AI costs could broaden research: Emad Mostak reportedly said some models he had seen were already 1,000 times cheaper per token, while a January OpenAI model might approach that reduction but had not been released. One speaker illustrated the potential with a hypothetical research effort falling from $20 million to $20,000, then $2,000 or $200, and suggested that 10,000 agents working in parallel could compress 100 years of progress into 21 hours. AI is also contributing to radiology and cancer research, with human clinicians still involved; earlier and more accurate scan review could improve outcomes and lower costs, though increased usage could offset savings and providers may not pass reductions to patients.

US-China competition was framed as a contest between the only two superpowers, with the United States retaining unmatched military strength and control of the global financial system, while China is expanding energy supply faster and has greater manufacturing capacity. Both countries face economic difficulties, which one speaker said currently helps balance their power, and both seek to avoid direct conflict even as they compete economically and in AI. Elon Musk warned that the US may not bring enough compute online by 2027 because it cannot produce enough energy. Speakers linked the AI buildout’s prospects to energy supply, permitting, and public opposition to data centers; Musk’s idea of moving computation into space was discussed as a longer-term possibility, not an immediate solution.

One speaker said China’s export sales bring in dollars despite a weak economy, and suggested that rising gold, silver, or Bitcoin could signal investors seeking refuge from US debt. The speakers said the US could lose ground if it fails to address debt and internal divisions, while China could pull ahead if it addresses demographic challenges. AI-driven productivity might help the United States grow into a more sustainable fiscal position; without it, entitlement cuts could become necessary and sharpen political conflict. They supported continued AI competition to accelerate innovation, but argued for guardrails around weapons-grade AI and cautioned that poorly designed or excessive regulation could impede progress.