Why This Trader Thinks Stocks Have Topped for a Correction

TL;DR

The stock market has likely topped and is due for a 10% to 15% correction as the AI trade reaches an inflection point. AI stocks drove 75% of the S&P's upside over the last two years, but stretched valuations, halted data centers, circular financing between chipmakers, and unrealistic chip depreciation schedules suggest the rubber band has stretched too far.

Transcript

This is a classic kind of Ponzi in a way to keep the positives going here. We are teetering on a level of a retrace or a pullback that could be a lot bigger than expected. What I think is happening right now is so much bigger and will end in so much of a worse scenario that I don't think any of us can even fully grasp it. And at least 60 to 70% of ... Read More

Key Insights

• AI stocks are directly responsible for 75% of the S&P's upside over the last two years, meaning a small group of names has been carrying the entire market higher rather than broad-based strength across the economy.
• The weak labor market is not primarily caused by AI, according to Soloway, but by persistent inflation that is higher than reported, forcing consumers to slow spending and companies to stall hiring in response.
• AI stock valuations are pricing in revenue the companies will not receive until 2030, essentially pulling forward five years of future price action into today's stock price and inflating current levels.
• Circular financing among AI companies functions like a Ponzi dynamic, where firms give each other money and warrants so chips can be bought, keeping the positive narrative going despite weaker underlying reality.
• Data center construction is being halted because of insufficient power, with Microsoft pausing two data centers and Micron halting one, undercutting the core thesis that endless chip demand justifies high semiconductor stock prices.
• Hyperscalers depreciate chips over seven years even though a chip bought at full price is worth only about 10% of that after two years, causing companies to overstate profits by a significant margin.
• The semiconductor ETF SMH has now stretched 102% above its 200-week moving average, the same extreme reached before a 45% retrace and later a 40% correction, signaling a rubber-band snap-back is due.
• Most investors privately agree the market is a bubble yet keep buying because they fear missing out and believe they can sell before the top, the same behavior Soloway heard in crypto during 2021.

Questions & Answers

Q: Why does Gareth Soloway think the stock market has topped?

Soloway believes the market has topped for at least a 10% to 15% correction because the AI trade has reached an inflection point. He points to valuations pricing in revenue that companies will not earn until 2030, circular financing between AI firms, data centers being halted for lack of power, and unrealistic chip depreciation. His semiconductor chart shows prices stretched 102% above the 200-week moving average, the same level seen before prior large corrections.

Q: How much of the stock market's gains came from AI stocks?

According to the data Soloway cites, over the last two years 75% of the upside in the S&P has been directly related to the AI stocks. This means a concentrated group of AI names has been carrying the broader market higher, rather than gains being spread across the whole economy. He treats this concentration as a sign of fragility, because if the AI trade stalls, the main engine lifting the index weakens too.

Q: Is AI causing the weakness in the labor market?

Soloway does not agree that AI is the main reason the labor market is currently semi-weak, though he expects that to become more true over time. He argues the bigger cause is persistent inflation that he believes is higher than reported, which makes consumers slow their spending. When consumers pull back, companies such as restaurants have to stall hiring, so the softness comes from inflation and business uncertainty rather than AI productivity gains alone.

Q: What is the problem with how AI companies are valued?

The problem is that stock prices are reflecting revenue the companies will not receive for the next five years, effectively pricing in what they might earn by 2030. Soloway says this pulls forward a large amount of future price action and upside into the present, leaving little room for further gains. Combined with companies admitting they have trouble monetizing AI right now, he questions whether the current valuations are truly justified.

Q: Why are AI data centers being halted?

Data centers are being put on hold because there is not enough power to run them. Soloway notes Microsoft has halted two data centers and Micron has halted one for this reason. Power cannot simply be pulled from the grid, because doing so would sharply raise electric bills for ordinary people, potentially tripling them. Since the AI stock surge depends on data centers being built and filled with chips, these halts undercut the core bullish thesis for the sector.

Q: What is the issue with how hyperscalers depreciate chips?

Hyperscalers spread the depreciation of chips over seven years, which Soloway calls unrealistic. He points to research, echoing a point made by Michael Burry, showing that a chip bought at full price is worth only about 10% of that value after two years, due to rapid technological advancement and heavy continuous use. Stretching depreciation over seven years makes the yearly cost look small, so reported earnings appear larger and profits are overstated by a significant margin.

Q: What does the semiconductor chart signal about a coming correction?

Soloway looks at the weekly chart of SMH, the VanEck Semiconductor ETF holding names like Nvidia, AMD, and Broadcom, and measures distance from the 200-week moving average. Historically, when the ETF reached 102% above that average it snapped back, once with a 45% retrace and later a roughly 40% correction. A few weeks ago the semis again topped out at 102% above the average, which he treats as a rubber-band signal that a sizable correction is due.

Q: Why do investors keep buying if they think it is a bubble?

Soloway says most people privately agree the market is a bubble but keep buying because they do not want to miss out and believe they can sell before the music stops. He compares this to what he heard repeatedly in crypto in 2021 and in many other assets as tops were being put in. The danger is that everyone assumes they can exit before everyone else, which is rarely how a top actually plays out.

Summary & Key Takeaways

• Gareth Soloway argues the stock market has topped and is set for a 10% to 15% correction. He notes AI stocks drove 75% of the S&P's upside over two years, but the labor market is weak mainly due to persistent inflation squeezing consumers and forcing companies to slow hiring.

• Several warning signs point to an AI inflection: valuations price in revenue not due until 2030, chipmakers shuffle money and warrants between each other in a Ponzi-like way, and data centers are being halted because there is not enough power without tripling household electric bills.

• Soloway highlights that hyperscalers depreciate chips over seven years despite chips losing most value within two, overstating profits. His semiconductor chart shows SMH stretched 102% above its 200-week average, matching prior extremes that preceded 45% and 40% corrections, so a snap-back is due.