A Semi-Theory of Almost Everything — Michael W. Green - a summary
Green’s full article is long and technical, so here’s the short version. The message: leveraged ETFs are poor long-term holdings. Obvious to you, perhaps. It’s lost on a retail crowd that can’t get enough of them.
Michael Green argues that the semiconductor rally in spring 2026, and its abrupt reversal, was driven substantially by mechanical ETF flows. Improving AI fundamentals and investor enthusiasm explain much less of it than people assume.
The Philadelphia Semiconductor Index rose 88% during Q2, with realised volatility around 60%. Micron rose about 230%. Nvidia gained only around 15%. Green traces the size, volatility and concentration of these moves largely to leveraged semiconductor ETFs and one very large newly launched sector fund.
Leveraged ETFs reset their exposure every day. A 3× fund buys more after the market rises and sells after it falls. Inverse leveraged funds trade in the same direction: both bullish and bearish products chase the day’s movement at the close.
That structure amplifies momentum. Until recently, investors in these products took profits after rises and added money after declines. Those flows offset the ETFs’ compulsory momentum trading, which made the overall complex stabilising.
From early April, investors stopped taking profits as aggressively. The ETFs retained more assets as prices rose and had to buy progressively more exposure.
Green calculates that the leveraged semiconductor complex generated roughly $76 billion of gross rebalancing demand during Q2, leaving about $33 billion of net momentum buying once investor withdrawals are accounted for. Under the old investor behaviour the net effect would have been close to zero.
A model using only mechanical trading rules, with no sentiment, narratives or psychology, explains roughly 60% of the change in market behaviour. Volatility rose sharply, daily reversals weakened and trends extended instead of correcting. When the market turned down in July, the same rules ran in reverse and produced about $13 billion of momentum selling in three weeks.
Volatility drag. Leveraged ETFs multiply daily returns, not long-term returns, so alternating gains and losses steadily destroy value.
At normal semiconductor volatility, a 3× fund already needs very strong underlying returns just to break even. At the 60% volatility of the rally, Green estimates the underlying index would need to return roughly 54% annually to overcome the drag. Once a strong trend goes sideways or reverses, losses compound fast.
SOXS, the 3× bearish semiconductor ETF, has lost virtually its entire value since 2012. It still operates because fresh investors keep replacing the capital that has been destroyed.
Green says the leveraged ETFs set the character and violence of the market, while a separate $26 billion sector fund helped set which shares benefited most. Its rules generated around $5 billion of exposure demand for one mid-cap semiconductor company, Micron, during the quarter, whether that company’s shares were rising or falling.
The growing crowd of single-stock leveraged ETFs mattered much less than assumed, with one exception. Their assets were mostly too small to move the market.
Green’s wider conclusion: modern market prices increasingly reflect the operating rules of investment products and the behaviour of their holders.
The semiconductor boom was real. Its magnitude, volatility and leadership were substantially manufactured by ETF mechanics. Once investor behaviour changed, a system that had cancelled out its own momentum started reinforcing it. Caveat emptor!
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.