Michael Burry, known as the "Big Short" investor, continues to steer clear of the overcrowded and heavily leveraged AI computing trade, instead hunting for value elsewhere. His latest portfolio includes stakes in copper miner Ero Copper (ERO.US), Brad Jacobs's QXO (QXO.US), and three deeply discounted stocks in home goods, food retail, and animal health.

Burry has long questioned the AI computing expansion and the associated infrastructure spending frenzy. The most notable logic behind his newest positions is his separation of "AI company valuations and earnings quality" from "physical procurement demand generated by AI construction." In essence, he evaluates the investment merit of the computing chain versus the resource needs of AI buildouts, maintaining a skeptical view on the returns of certain AI stocks while backing the copper demand driven by long-term expansion of data centers, power grids, and distribution systems.

His recent criticism of the AI computing rally centers on extreme valuations, capital expenditure returns, equipment depreciation, and future payment commitments tied to popular AI stocks. In a September 9 filing, he reduced portfolio risk by selling put options on Nvidia and Palantir expiring in December 2026 without rolling them over, while keeping short positions in both companies and put options on Palantir and the QQQ ETF, the latter tracking the Nasdaq 100, expiring in 2027. At that time, his top three stock shorts were Oracle, Palantir, and Nebius, a leader in the "AI new cloud" space.

On September 19, he widened his focus to Amazon, Meta, Alphabet, Microsoft, and Oracle, estimating that the five companies hold over $3 trillion in undisclosed off-balance-sheet commitments, including leases, purchases, and guarantees. He warned of a "catastrophic mismatch" between long-term fixed expenses, rapidly evolving AI hardware, and unproven commercial returns, stating that "when the music stops, these off-balance-sheet commitments will quickly become real liabilities." These moves underscore his bearish stance on AI computing leaders, while he adjusts his short tools and position durations.

Turning Away from the AI Party Hype to Find Overlooked Value

While shunning crowded and leveraged AI computing stocks, Burry is indirectly betting on AI through Ero Copper, arguing that data center construction will boost copper demand, while new mines take 18 years to develop. "The party is crowded, and today it's pushing AI stocks higher, but I largely ignore the cheers," the "Big Short" investor wrote in a Substack subscription post. He named five companies: Ero Copper, QXO, Temple & Webster, Sprouts Farmers Market (SFM.US), and Zoetis (ZTS.US). He also noted holding QXO common shares and its mandatory convertible preferred stock.

Through Ero, a Brazil-focused copper and gold producer, Burry is placing a direct bet on AI's physical needs. "The people in that room are all going to need a lot of copper," he said. COMEX copper futures settled at $6.6865 per pound on Monday, about 2% below the record high set on September 9. Driven by mine supply disruptions, low Chinese inventories, and pre-holiday stockpiling ahead of China's National Day, LME copper futures, the international benchmark, are up 19% this year and 46% over the past 12 months. That annual gain outpaces the SPDR S&P 500 ETF Trust (SPY.US), the Invesco QQQ Trust (QQQ.US), and the SPDR Dow Jones Industrial Average ETF Trust (DIA.US), which have returned roughly 13%, 18%, and 9%, respectively.

Burry acknowledges that part of the current rally reflects temporarily weaker supply. His long-term thesis rests on a severe mismatch between copper demand and the time required to develop new mines. Citing Apollo chief economist Torsten Slok, he notes that discoveries of large copper deposits with at least 500,000 tonnes of contained copper have fallen from double-digit annual figures in the 1990s and 2000s to just one or two per year recently, with none in 2025. New deposits take 18 years to bring into production, while AI data centers can create copper demand within two to three years. S&P Global projects global copper consumption will rise from 28 million tonnes in 2025 to 42 million tonnes by 2040.

Burry views Ero as a higher-cost producer whose margins could expand significantly if copper prices stay elevated. "Ero common stock is right up my alley," he wrote, describing it as a medium-sized position. The company has maintained its 2026 production guidance of 67,500 to 77,500 tonnes, with C1 cash costs of $2.15 to $2.35 per pound. Burry also praised the management team that took over in early 2025, saying it delivered on "three rather difficult commitments," and added that a forensic review of the company's accounting "found no issues." Despite ERO shares more than doubling over the past year, Burry still sees a significant valuation discount relative to larger copper miners.

On Stocktwits, retail sentiment toward ERO fluctuated between "neutral" and "bearish" over the past week, even as posting volume doubled and follower count rose 5% in the past month, signaling growing retail interest. Burry also expressed strong support for Brad Jacobs's acquisition and consolidation strategy at QXO. He holds QXO common shares and its Series B mandatory convertible preferred stock with a 5.5% stated dividend rate. "The stock has fallen sharply, and I see an opportunity to participate," he said.

QXO, led by Brad Jacobs, founder of United Rentals and XPO, is consolidating the fragmented building materials distribution market through a series of acquisitions. Burry believes scale and solid execution could drive a valuation re-rating, despite the stock being "affected by the housing market and interest rate disruptions, not to mention diesel price swings." He favors QXO's preferred shares, noting a current dividend yield of 7.4% with cumulative dividends. Unless converted early, these securities will automatically convert into common stock on or around May 15, 2028. "The common and preferred are roughly equally attractive, but I like the downside protection the dividend provides," he said. On Stocktwits, retail sentiment on QXO has moved from "bullish" a week ago to "extremely bullish," with monthly posting volume surging 467% and follower count up 4%.

Burry increased his stake in Australian online furniture retailer Temple & Webster, which trades over-the-counter under the ticker TPLWF, calling it "a fairly large position." The stock has fallen about 82% over the past year on concerns about housing and consumer spending. For fiscal 2026, revenue rose 11% to A$665 million, while net profit dropped 62% to A$4.3 million. Burry said he plans to publish his full investment thesis soon.

The short-focused investor also "added slightly" to Sprouts Farmers Market and Zoetis to complete this round of buying. Sprouts shares are down about 43% over the past year, while Zoetis has halved amid weak companion animal demand and lowered guidance. Burry said Zoetis "has always appealed to me" but is now "extremely out of favor," lacking any COVID-era boost, AI relevance, or current market "buzz." On Stocktwits, retail sentiment on Zoetis remains "bearish" over the past week, with posting volume down 75%, though follower count has grown 9% in the past month. Meanwhile, sentiment on Sprouts has improved from "bearish" a day ago to "neutral," though still below the "bullish" level from a week earlier; posting volume fell 67%, and follower count was roughly flat, up just 0.1% in the past month.

The 'Big Short' Investor's Contrarian Playbook: Shorting AI Valuations, Betting on Copper as the Foundation of Power

While questioning the valuations and earnings quality of certain popular AI computing stocks, Burry is betting on the power grid and distribution system demands driven by the rapid expansion of AI data centers. His core positive thesis centers on slow supply response: new copper mines take about 18 years from discovery to production, while data centers can be built and operational in two to three years, creating a clear time gap between incremental procurement demand and mine supply. S&P Global projects global copper demand will grow by about 50% from 28 million tonnes in 2025 to 42 million tonnes in 2040, encompassing contributions from AI, electrification, and other economic activities. Thus, Burry's copper investment seeks resource segments where construction demand can materialize but supply cannot quickly expand.

For Ero Copper, he is betting on earnings elasticity from both copper prices and operational improvements. The company has confirmed 2026 copper production guidance of 67,500 to 77,500 tonnes with C1 cash costs of $2.15 to $2.35 per pound, and expects higher output in the second half with costs declining as operations improve. According to Burry's latest assessment, Ero remains undervalued compared to larger copper miners, and its higher cost structure means thinner initial margins, but when volumes and costs are controlled, rising copper prices can drive stronger profit gains. Management's delivery on operational commitments adds company-level support to this price elasticity. This forms an investment approach combining "resource supply-demand, operational improvement, and valuation discount."

Copper miners with producing assets, executable expansion projects, and stable supply networks are poised to gain more durable demand support as AI inference scales. As inference calls, agent tasks, and online service volumes expand, computing demand translates into server capacity, power supply, and distribution buildouts. The International Energy Agency (IEA) now projects global data center electricity consumption will nearly double from 485 terawatt-hours in 2025 to 950 terawatt-hours by 2030. Copper enters the picture through power cables, transformer windings, distribution equipment, and rack busbars, where its conductivity helps manage resistive losses in high-power equipment. Within racks, short-distance high-speed interconnects also use copper cables and related components, such as the copper cable trays in Nvidia's GB200 NVL72 for internal rack connections.

Power infrastructure constitutes the primary copper volume demand, while high-speed copper interconnects reflect AI technological value-added in signal integrity, high-speed connectors, and cable manufacturing for data centers. Meanwhile, longer-distance links within data centers are accelerating the adoption of optical interconnect components and co-packaged optics (CPO) technology. In terms of corporate gains, miners capture profit elasticity from resource scarcity, volumes, and copper prices; traders benefit from sourcing, logistics, financing, and spread management; and high-speed cable makers gain from product upgrades.

The other four companies reflect value recovery and capital allocation logic. QXO's appeal stems from Brad Jacobs's track record in consolidating fragmented building materials distribution markets and the potential efficiency gains and valuation re-rating from expanded scale. Burry holds both common shares and the Series B mandatory convertible preferred stock with a 5.5% stated dividend rate, while the 7.4% figure represents the current dividend yield at prevailing prices. Temple & Webster, Sprouts Farmers Market, and Zoetis represent contrarian plays in home goods, food retail, and animal health, respectively.