Двухнедельный обзор фондовых рынков №317. «Медвежий» капкан Трампа

A lot of investors are looking at the S&P 500 rally with understandable caution. September is historically one of the weakest months for equities, yet the index still finished higher. Then, on October 6, it set a fresh all-time high at 7,819. What makes that more striking is what is happening beneath the surface. At the start of October, half of S&P 500 constituents were still 17% below their all-time highs, while only about 20% were trading above their 50-day moving averages. The third-quarter picture was no broader: the seven largest technology companies gained 11%, compared with just 2% for the S&P 500, while the equal-weighted index fell 2% and the Russell 2000 lost 7%.

To me, the explanation is fairly straightforward: earnings are just as concentrated as market performance. Expected S&P 500 earnings growth for Q3 stands at 27%, but more than two-thirds of that increase is coming from only ten companies. Nvidia and Micron alone account for more than one-third. For the typical company, meanwhile, earnings growth is slowing from 14% to 9%. The same concentration is visible in capital spending. By our estimates, capex at the five largest cloud companies will rise from more than $800bn in 2026 to $1.1tn in 2027. That still implies 37% growth next year, although it follows an almost twofold increase in 2026. For now, it remains much easier to measure the benefits accruing to the companies building AI infrastructure than to the companies using it. In 2Q, only 2% of S&P 500 companies quantified AI’s impact on earnings. Clearly, the gains from this revolutionary technology are not showing up everywhere yet.

But I would not sound the alarm too early. The earnings are real. The S&P 500 is trading at around 19x expected 12-month earnings, almost exactly in line with its 10-year average of 19.1x. The obvious risk is that expectations become too volatile or too optimistic, but at this stage that remains more of a concern than an established fact. Semiconductors are a good example. The largest producers understand the cyclicality of their business perfectly well. If they overproduce, they risk losing a meaningful volume of orders over the following two to three years. Memory producers are trying to reduce that risk through multi-year agreements. Micron has already signed 26 contracts that, according to the company, will cover more than 35% of revenue through 2030. These agreements include purchase-volume commitments, and many also contain price floors.

I see those contracts as an additional layer of protection for individual companies. They do not eliminate cyclicality, but they can meaningfully soften it. That is why I think a correction is entirely possible, but my base-case year-end target is still closer to 8,000 than 7,500. I do not expect the index to fall below 7,000 or decline by more than 10% before year-end. There is also an upside scenario. With each quarterly earnings season, more companies may begin to quantify the contribution of AI to earnings. If that happens, the market could get a second leg of support, not from even greater concentration, but from broader evidence that AI is translating into productivity and profits. In that scenario, the concentration debate we started with would gradually become less important. For more cautious investors who did not follow my earlier recommendation, buying protection through put options on the S&P 500 ETF (SPY) may still make sense. The cost of that insurance remains relatively low.