Michael Burry has replaced a direct short position in SOXX with long-dated puts, according to October 1 reporting by Benzinga and Stocktwits. The trade expresses a bearish view on the durability of AI-related chip demand. It does not establish that demand is already falling: the same reporting describes large customer commitments to memory supply.
SOXX puts extend Burry's bearish exposure
The reports describe September 2027 SOXX puts with strikes in the low $400s, alongside put positions in other AI-linked companies. Stocktwits reports different expiries across those positions; they should not all be treated as identical September 2027 trades. The reports attribute these position updates to Burry himself.
Benzinga reported that SOXX fell about 11% in the third quarter, its weakest quarter since March 2025. It put the fund's assets at about $48.4 billion and trailing price-to-earnings ratio near 68. Those figures describe the market backdrop to Burry's trade, not proof that his forecast will be correct.
Burry questions the link between scale and intelligence
Stocktwits describes Burry discussing an 1880 account of reasoning before written language. His argument concerns whether scaling language models establishes understanding, and whether compression or reasoning-first approaches could reduce computing requirements. That is a conditional challenge to the economics of scaling AI, rather than evidence that current chip demand has already weakened.
Customer commitments challenge the bearish case
Benzinga cites an S&P Global estimate that Alphabet, Amazon, Meta, Microsoft and Oracle will collectively spend around $750 billion on capital expenditure in 2026. That is an estimate of combined capital spending, not a disclosed semiconductor order total.
The report also cites Micron customer commitments of $32 billion under long-term supply agreements and agreements covering most of its 2027 high-bandwidth memory output. These commitments provide a counterpoint to the bearish interpretation: customers are still securing supply. They do not settle whether future efficiency gains will reduce demand per task or enable more applications and greater overall usage.
Spending and usage will test the thesis
Future customer spending disclosures and memory-supply updates may help test the competing interpretations. They could clarify whether efficiency gains are reducing hardware requirements or supporting wider adoption. Capital-spending announcements alone would not establish the eventual outcome of Burry's positions.
Share this article
