The loudest worry in this market is return on investment at the capex level for the hyperscalers. We know the supply side. They have spent about a trillion dollars over the last three and a half years, most of it this year. The return is trickier to measure. The best attempt I have seen, a report out of London running fifty or sixty charts, puts AI revenue at $175 billion so far this year with a run rate heading over $300 billion. Some of that is subjective and parts of it will be wrong, but it is a good enough number that I expect the spending to continue. How it gets financed is a separate risk.
The core of the whole issue is the cost of compute. Goldman Sachs publishes a proxy for the cost of computing and data processing at the large language model level. It started in December of 2025, rose 100 percent through June 1, and has come down about 25 percent since. That is a good thing. Jevons paradox is well documented: as the price of a technology goes down, more people use it. The moonshot moment this week was China announcing a new model it says rivals the leading American labs. It is untested, and they said the same thing about DeepSeek, which sits at the lower end of compute and which plenty of people use. More compute is better. There are real risks in the AI space. My view is that this is not one of them.
The gauge I would actually watch is the price of oil. It was $65 a barrel on March 1, the first day of Ramadan, when the conflict started. The stock market bottomed on April 1, the day Artemis 2 took off from Cape Canaveral. Oil printed $65 again into the July 2 holiday weekend, a full round trip, and it is up 25 percent since. Major portions of the stock market are down about the same amount. The inverse correlation right now is extremely high. Why? Very simple. Stocks are expensive and oil is cheap. Until there is free flow of oil through the Strait of Hormuz, the price is not coming down, and I think the conflict continues to escalate. If oil keeps rising, I would be less than enthusiastic about equities. I could be wrong on that.
There is a truism in global markets that the FX investors get it first, the bond market gets it second, and equity investors get it third. I believe that. The dollar is stable and rising, money is coming into this country, and that is not a risk. Investment grade spreads are at 77, against 95 on March 1 and 170 when Russia invaded Ukraine. High yield is near 267, against 335 in March and 600 at that invasion. Credit is telling you the risk of recession is low. Hyperscaler spreads have widened a bit over the last month or two, which does not alarm me, and they should tighten as revenue trickles in.
Inflation swaps a year out are at 230, and this week’s CPI was the lowest reading in months and the largest negative reading in several years. I do not expect rates to go much higher from here, and rates at this level are no impediment to corporate profitability. Liquidity is ample and rising. Global GDP is about $100 trillion, global stock markets are worth about $100 trillion, and global money supply is about $300 trillion, which has never been higher. Half the increase since the pandemic came from China, which is printing to keep GDP out of negative territory. China is in trouble and I would not invest there. Bond investors do not care how good things get. They care about getting their money back, and right now they are telling you the risk is low.
