Scion Asset Management's Michael Burry covered his short positions in Nvidia (NASDAQ: NVDA), Micron Technology (NASDAQ: MU), and Palantir Technologies (NYSE: PLTR) and replaced them with put options.
Understanding the gap between his reputation and his post-2008 returns is part of maintaining financial fitness in a market where celebrity trades move sentiment before the data does. The more instructive signal sits in hyperscaler depreciation schedules, not in any single fund manager's quarterly filing.
Now Available: The Definitive Guide to Retirement Income
Many successful investors eventually reach the same moment. The saving is done, the portfolio is built, and the question quietly changes from how much can I grow this to how much can I take out? Get that second question wrong and decades of good investing can come apart in a handful of years.
That is exactly what The Definitive Guide to Retirement Income helps answer.1 <p class="">Investing in securities involves a risk of loss. Past performance is never a guarantee of future returns. Investing in foreign stock markets involves additional risks, such as the risk of currency fluctuations. Guides are provided by Fisher Investments. The publisher may be compensated when readers request a guide.</p> It covers what your retirement could actually cost, which income sources are worth using, and the withdrawal math that decides whether the money lasts. It is free today from Fisher Investments. Read More Here ›
Burry shifted from outright shorts to put options on Nvidia, Micron, and Palantir
Burry held short positions in all three stocks before switching to puts, a transition from a strategy with unlimited theoretical loss to one that caps downside at the premium paid, The Motley Fool documented. Puts preserve bearish exposure without the margin calls that force premature exits during rallies.
The move aligns with Burry's warning of a "1987-style crash" driven by crowded momentum positions in artificial intelligence (AI) stocks and unsustainable leverage among growth investors. Defined-risk options suggest he expects a prolonged correction.
The 2008 housing bet earned $100 million, but the post-crisis record is mixed
Burry earned $100 million personally and generated $700 million for his investors by shorting subprime mortgage bonds before the 2008 crisis, FinMasters confirmed. The trade established his reputation as a contrarian willing to absorb years of losses before a thesis materializes.
Scion managed roughly $238 million across four clients as of the third quarter of 2023. Burry also publishes Cassandra Unchained, a paid newsletter on Substack with more than 300,000 subscribers, The Motley Fool noted, a platform that creates financial incentives to maintain bearish calls regardless of accuracy.
A $1.6 billion S&P 500 bet and a $47 million chip put both closed at a loss in 2023
Burry purchased $1.6 billion in puts against S&P 500 and Nasdaq 100 exchange-traded funds (ETFs) in 2023 and closed them at a loss as the S&P 500 gained 24% and the Nasdaq 100 rose 54%, FinMasters confirmed. A $47 million semiconductor ETF put also closed at a loss.
The pattern of directional instinct paired with timing so early that positions expire before the thesis materializes has defined his post-2008 record, and the Nvidia put carries the same structural risk if AI spending accelerates before the depreciation reckoning.
Released: The Definitive Guide to Retirement Income
Most people spend their entire working lives focusing on one thing: growing a pile of money. Then they retire and discover the skill they actually need is close to the opposite. Turning that pile into steady income, year after year, through good markets and bad, without ever running out. It may be the least discussed and most consequential transition in personal finance, and plenty of otherwise careful investors walk into it with no real plan.
That is exactly what The Definitive Guide to Retirement Income helps answer.1 <p class="">Investing in securities involves a risk of loss. Past performance is never a guarantee of future returns. Investing in foreign stock markets involves additional risks, such as the risk of currency fluctuations. Guides are provided by Fisher Investments. The publisher may be compensated when readers request a guide.</p> It is a free guide from Fisher Investments covering what your retirement will really cost, where the cash flow should come from, and how much you can safely withdraw each year. Learn More Here ›
The Big Four hyperscalers guided $610 billion to $655 billion in 2026 capital spending
The four largest cloud infrastructure providers are on track to spend a combined roughly $610 billion to $655 billion in capital expenditures in 2026, according to Dubach's analysis. The estimates break down as follows.
Amazon guided about $200 billion in 2026 capital spending.
Meta was estimated at $60 billion to $65 billion.
Microsoft was estimated at $100 billion to $120 billion.
Alphabet guided $175 billion to $185 billion.
Amazon shortened server depreciation from six to five years and took a $920 million charge
Amazon took a $920 million charge in 2024 to early-retire servers, then shortened useful lives for certain hardware from six years to five years effective January 1, 2025, reducing operating income by $700 million, Dubach's analysis documented. The adjustment acknowledged that graphics processing unit (GPU) hardware evolves faster than traditional servers.
Alphabet, Microsoft, and Meta have not matched Amazon's depreciation change, and carrying five-year or six-year schedules on hardware that may become obsolete in three years creates an earnings gap that widens as spending accelerates.
Understated depreciation could overstate hyperscaler earnings by $176 billion through 2028
Burry estimates that hyperscalers will understate depreciation by approximately $176 billion between 2026 and 2028, overstating reported earnings by more than 20% during that period. The calculation assumes true useful lives are shorter than the five-year schedules providers carry.
Depreciating $610 billion to $655 billion in 2026 capital spending over five years generates $122 billion to $131 billion in annual depreciation charges. Shorter useful lives would pull those charges forward, compressing the earnings that justify valuations for Nvidia's largest customers and GPU demand forecasts.
AI revenue covers roughly 25 cents of every capex dollar, while GPU costs fall 280-fold
Direct AI revenue covered approximately 0.15x of what the Big Four spent on AI infrastructure in 2025, rising to a projected 0.25x in 2026, the analysis documented. Sequoia Capital estimated the AI ecosystem needs $600 billion in annual revenue to justify the spending trajectory, a threshold five to seven times above projected levels.
GPU inference costs fell from $20 per million tokens in 2020 to $0.07 by early 2026, a 280-fold decline that accelerated after January 2024. Falling prices undermine pricing power for providers selling inference, meaning revenue may grow more slowly than capacity.
Bottom line
Burry's shift from shorts to puts on Nvidia, Micron, and Palantir reflects a thesis about AI overcapitalization, but his post-2008 timing has been wrong often enough that the trade alone is not a portfolio signal. The depreciation math underneath his bet is a separate and more trackable data point.
The $176 billion depreciation gap, the $600 billion revenue threshold, and falling GPU prices are signals you can monitor independently, and tracking whether hyperscalers shorten depreciation schedules carries more weight for your decision to start investing in AI stocks than any single fund manager's quarterly filing.
This article is for informational purposes only and should not be considered investment advice.
More from FinanceBuzz:
- Retire like the rich: 14 ways you could build wealth in your 50s.
- Find out if you could pay less for car insurance in just a few clicks.
- Make these 7 savvy moves when you have $1,000 in the bank.
- 14 moves seniors could benefit from but often forget about.
Add Us On Google