Broadcom's CEO said AI demand is 'robust,' but Druckenmiller reportedly sold his entire Broadcom stake. Who should I believe?
2026-09-15
An Asymmetry Between Winners and Losers: Same Stock, Two Conflicting Convictions
Two signals collided head-on on the same day. Broadcom CEO Hock Tan reaffirmed that demand for AI compute infrastructure is "very strong and I believe very durable." Meanwhile, Druckenmiller's Duquesne Family Office fully liquidated its stakes in Broadcom, Intel, and Micron in its Q2 13F filing (as of June 30, disclosed August 14) (Motley Fool, 2026-08-24 / 2026-09-12). Yet the actual 13F data shows that capital didn't flow into Hyperliquid or mining stocks — it went into a new AMD purchase, an expanded TSMC stake, new positions in Lam Research and Equinix, and a new Alphabet purchase (Investing.com/Motley Fool, 2026-09). In other words, Druckenmiller hasn't "left AI" — this looks more like a selective reallocation, narrowing from legacy hardware and earlier-generation AI bets (Broadcom, Intel, Micron, Coherent, Lumentum) into foundry, equipment, and data-center infrastructure (TSMC, AMD, Lam Research, Equinix).
Who Wins and Who Loses
This divergence is backed by a structural signal, not just a single-stock issue. In the BofA Global Fund Manager Survey, the share citing "long global semiconductors" as the most crowded trade plunged from 82% in July (a record high) to 53% in August — a 29-percentage-point drop in a single month (BofA survey cited via Seeking Alpha/Investing.com, 2026-08). When consensus thins out like this, participants typically split into two camps. Corporate executives with solid earnings and guidance (like Broadcom's CEO, who has direct visibility into revenue numbers) have every incentive to stay optimistic, while macro investors who assess valuation across the whole portfolio (like Druckenmiller, managing aggregate portfolio risk) tend to trim first once they judge the consensus has become overheated. Today's signal from SMH tracks the same fault line.
| Date | Close | MA10 | MA50 | trail_buf | Verdict |
|---|---|---|---|---|---|
| 09-09 | $574.29 | $560.18 | $572.05 | 5.48% | entry_ok, caution |
| 09-10 | $560.28 | $560.63 | $570.14 | 3.4% | entry_ok, risky |
| 09-11 | $568.53 | $560.19 | $569.10 | 4.62% | entry_ok, risky |
| 09-14 | $541.50 | $559.03 | $568.09 | 0.6% | entry_ok, risky |
The fact that trail_buf (the cushion left before a trailing stop triggers) shrank from 5.48% to 0.6% in a single week means the price itself has moved right up against the stop-loss line, regardless of management's optimism.
So What Should Investors Do?
There's no need to believe only one of the two signals — management commentary and the macro investor's positioning shift operate on different time horizons. Broadcom's CEO's remarks reflect company-level 2028 guidance, while Druckenmiller's rebalancing is a defensive adjustment ahead of a possible valuation pullback over the next one to two quarters. Rather than relying on conviction about a single stock (AVGO, etc.), it's safer to also track whether the BofA survey's crowding reading falls further below 53% in August and whether SMH's trail_buf rebounds above 3% after the FOMC (9/16), instead of judging purely on earnings-side optimism. Detailed charts and signals for other tickers are available on the /signals dashboard.