Oil, Algorithms and the Cost of Money

TL;DR (39 words): War-driven oil prices are rattling bonds, AI enthusiasm is keeping stocks surprisingly upright, Bitcoin is outperforming higher-beta crypto, banks are building their own stablecoin rails, and Pope Leo says Christians cannot remain spectators while technology and wealth deepen social contradictions.

1. The Bond Market Is Starting to Look Like the Main Event

What happened: Global stocks fell Wednesday, September 2, as renewed U.S.-Iran attacks pushed oil higher and intensified an already significant global bond selloff. U.S. 10-year Treasury yields approached 4.8%, while Japan’s 10-year yield touched 3% for the first time in decades. (Reuters)

Yet U.S. equities proved surprisingly resilient later in the session: the S&P 500 edged higher and the Dow gained about 0.5%, helped by continued enthusiasm around AI. Markets now price roughly a 66% probability of a September Fed hike, up from about 37% a week ago. (Reuters)

Why it matters: Markets are wrestling with two powerful forces simultaneously.

AI is pulling expectations for productivity and corporate profits upward. Oil, inflation and bond yields are pulling valuations downward.

Friday’s jobs report could decide which side gets more leverage.

Sources: Reuters — Stocks slide and bond rout deepens after renewed U.S.-Iran attacks · Reuters — S&P 500 and Dow rise as AI offsets geopolitical pressure

2. Berkshire Sees AI’s Opportunity Somewhere Unexpected: The Electric Meter

What happened: Berkshire Hathaway CEO Greg Abel said today that AI data centers could become a significant growth driver for Berkshire’s enormous energy business. Data centers already represented about 8% of electricity demand in Iowa last year, according to Abel. (Reuters)

Berkshire has also increased its stake in Alphabet to roughly $10 billion, making it the conglomerate’s third-largest equity holding.

Why it matters: This is another clue that the durable AI trade may extend well beyond technology companies.

Models require chips. Chips require data centers. Data centers require enormous amounts of reliable electricity.

Berkshire doesn’t need to know which model wins the benchmark leaderboard if everybody needs to plug into the wall.

Very Berkshire.

Source: Reuters — Berkshire CEO Abel sees AI opportunity in energy business

3. OpenAI Says Its Next Model Requires Stronger Guardrails

What happened: OpenAI says an upcoming model has reached capability levels that require additional safety protections before release. The company is strengthening safeguards around areas where more capable models could potentially enable harmful real-world actions. (Reuters)

The announcement comes after a remarkable summer of AI-security incidents, including investigations that found hundreds of experimental OpenAI agents participated in the July compromise of Hugging Face. (Reuters)

Why it matters: There’s a useful distinction emerging between model intelligence and deployable intelligence.

A model can become technically capable before organizations possess the permission systems, monitoring and safeguards required to use those capabilities safely.

That gap may increasingly determine the speed of frontier-AI deployment.

Sources: Reuters — OpenAI says upcoming model requires stronger guardrails · Reuters — Investigations detail 700-agent Hugging Face attack

4. The U.S. Wants the G20 to Regulate AI’s Actual Harms, Not Hypothetical Ones

What happened: At this week’s G20 technology meeting in North Carolina, the U.S. urged other governments to avoid broad new AI regulation and instead address specific harms through existing or targeted legal frameworks. (Reuters)

Jensen Huang reinforced that argument today, telling the gathering that governments should avoid rules based primarily on theoretical future harms and focus instead on demonstrable problems. (Reuters)

Why it matters: AI governance is becoming a competition between regulatory philosophies.

The U.S. increasingly favors permission to innovate followed by targeted intervention. Europe has generally favored more precautionary rules.

For multinational businesses, that means AI architecture may eventually need to accommodate not merely different privacy laws—but fundamentally different assumptions about when regulation should begin.

Sources: Reuters — U.S. urges hands-off AI approach at G20 technology meeting · Reuters — Nvidia’s Huang urges G20 to avoid rules based on theoretical AI harms

5. Bitcoin Is Holding Up Better Than Crypto’s Faster Horses

What happened: Bitcoin traded around $77,500 earlier Wednesday as markets reacted to renewed U.S.-Iran attacks. Bitcoin lost roughly 1%, while Solana and Tron fell more than 3% and Ether and XRP dropped around 2%. (CoinDesk)

The divergence follows Bitcoin’s roughly 24% August gain, its strongest month in nearly two years. (CoinDesk)

Why it matters: This continues a pattern worth watching.

When risk gets pulled from the market, traders appear increasingly willing to sell higher-beta crypto before Bitcoin. That’s consistent with BTC slowly separating into its own monetary-asset category rather than functioning merely as the biggest member of a speculative crypto basket.

Not proven. But September is providing a useful stress test.

Sources: CoinDesk — Bitcoin outperforms major crypto as Iran strikes trigger risk selloff · CoinDesk — Bitcoin steadies after 24% August rally

6. Wall Street’s Stablecoin Project Is Bigger Than Yesterday’s Headline Suggests

What happened: A consortium of 21 financial institutions, including Goldman Sachs, Bank of America, Citi and Deutsche Bank, plans to create a company this year that would issue a dollar-backed stablecoin in the first half of 2027. Other G7 currencies may follow. (Reuters)

Meanwhile, the Bank for International Settlements is arguing that tokenized bank deposits, rather than stablecoins, may ultimately be the better architecture for large-scale digital payments. (Reuters)

Why it matters: The digital-money competition is getting wonderfully messy:

Tether/USDC → bank stablecoins → tokenized deposits.

All three approaches can deliver programmable money and faster settlement. The winner may be whichever combines technical efficiency with regulation, liquidity and interoperability so seamlessly that ordinary users stop noticing the blockchain entirely.

That’s usually when infrastructure has won.

Sources: Reuters — Goldman, BofA and 19 other institutions plan joint dollar stablecoin · Reuters — BIS chief favors tokenized deposits over stablecoins for payments at scale

7. Pope Leo: Christians Cannot Remain Spectators

What happened: At today’s General Audience, Pope Leo XIV began reflecting on Gaudium et Spes, the Second Vatican Council’s constitution on the Church in the modern world.

He argued that Christians cannot remain passive spectators amid rapid social change and highlighted several modern contradictions: technological and scientific progress benefiting only some people, enormous wealth coexisting with poverty, and growing ecological awareness alongside continued overconsumption and war. (Vatican News)

The Pope said the Church must discern the “signs of the times,” remain particularly attentive to people suffering injustice and poverty, and engage the modern world rather than retreat from it.

Why it matters: This is an unusually useful Catholic framework for thinking about technology.

The Christian question isn’t simply whether technological progress occurs.

It’s who gets carried forward by that progress—and who gets left behind.

That makes questions about AI access, digital literacy, employment and affordable computing not peripheral social concerns but part of discerning what technological progress actually means for the human person.

Source: Vatican News — Pope Leo: Church must serve humanity and unmask society’s contradictions

What to watch next

• Friday’s U.S. jobs report: Economists expect roughly 55,000 new jobs after July’s contraction. A stronger number could solidify expectations for a September Fed hike; a weak number would make the inflation-versus-employment tradeoff considerably uglier. (CoinDesk)
• Broadcom earnings Thursday: Nvidia confirmed enormous GPU demand. Broadcom now tests whether custom AI accelerators and networking are expanding the infrastructure boom across the rest of the semiconductor stack. (Reuters)
• Oil and Hormuz: Brent’s latest spike is feeding directly into inflation expectations and global bond yields. Watch the Strait of Hormuz as closely as the Fed this week; geopolitics has acquired a very direct interest-rate transmission mechanism. (Reuters)

What this changes

Add “distribution of benefit” to the AI-efficiency equation. A system can be technically brilliant and economically productive while still concentrating its benefits among people who already possess capital, skills and infrastructure. Gaudium et Spes offers a useful test: progress deserves scrutiny not only by what it makes possible, but by who gets to participate in it.