--- title: "Bernstein \"Pours Cold Water\": Musk's Super Chip Factory Needs $5-13 Trillion in Capital Expenditure, \"Harder Than Landing on Mars\"" type: "News" locale: "en" url: "https://longbridge.com/en/news/280323484.md" description: "Bernstein believes the \"Terafab\" project remains largely at the level of hype. According to estimates, achieving the target of 1 terawatt of annual compute capacity would require $5 trillion to $13 trillion in capital expenditure and new capacity equivalent to the current global semiconductor total, a challenge \"akin to rebuilding the entire semiconductor industry.\"" datetime: "2026-03-24T13:13:36.000Z" locales: - [zh-CN](https://longbridge.com/zh-CN/news/280323484.md) - [en](https://longbridge.com/en/news/280323484.md) - [zh-HK](https://longbridge.com/zh-HK/news/280323484.md) --- > Supported Languages: [简体中文](https://longbridge.com/zh-CN/news/280323484.md) | [繁體中文](https://longbridge.com/zh-HK/news/280323484.md) # Bernstein "Pours Cold Water": Musk's Super Chip Factory Needs $5-13 Trillion in Capital Expenditure, "Harder Than Landing on Mars" Elon Musk's ambitious "Terafab" plan is facing a sober assessment from Wall Street. According to Zhuifeng Trading Desk, Bernstein Research pointed out in its latest report with detailed quantitative calculations that **to achieve Musk's envisioned annual production target of 1 terawatt of compute power, the required capital expenditure would reach $5 trillion to $13 trillion, with the necessary wafer capacity equivalent to the total current global semiconductor capacity—a challenge "harder than landing on Mars."** Musk announced the launch of the "Terafab" project last weekend, aiming to expand human compute production scale to 1 terawatt annually, approximately 50 times the current global compute supply (around 20 gigawatts). The plan includes building an advanced wafer fab in Austin, Texas, covering the entire production chain for compute chips, logic chips, memory, packaging, and photomask production. A team led by Bernstein analyst Stacy A. Rasgon promptly released a report providing a systematic quantitative assessment of the plan's feasibility. Bernstein pointed out that if investors truly believe Musk can achieve this goal, semiconductor equipment stocks would be the most direct beneficiaries; however, for now, **the substantive impact of the plan on the industry "might not be significant, remaining more at the level of hype."** ## Astronomical Capital Expenditure Estimates Using the current compute architecture (Nvidia racks) as a baseline, Bernstein provided a rough estimate of the wafer demand for three core chip types: GPUs, HBM memory, and CPUs. The report indicates that **producing 1 terawatt of compute power annually would require a starting volume of 7 million to 18 million 300mm wafer starts per month, with HBM memory demand being the dominant factor.** **Converting this demand to factories with a capacity of 50,000 wafers per month, it equates to building 140 to 360 such factories. Based on a capital expenditure of $35 billion per factory, the total investment would range from $5 trillion to $13 trillion.** Bernstein explicitly noted in the report that these calculations are "very rough estimates" and do not yet include other semiconductor categories such as networking, optical, analog, and power chips. ## Capacity Demand Akin to Rebuilding the Global Semiconductor Industry Bernstein's estimates reveal the staggering scale of this plan in terms of capacity. The report states that **the wafer capacity required for 1 terawatt of compute power is roughly equivalent to the total current global semiconductor capacity (approximately 16 million 300mm equivalent wafers per month).** If focusing only on "relevant" semiconductors—namely memory plus 4nm and below advanced logic wafers—the current global installed capacity is about 5 million 300mm equivalent wafers per month, while the 1-terawatt target would require several times this figure. In other words, **Musk's plan essentially demands a multi-fold expansion on top of existing advanced process capacity, and the difficulty of this is self-evident.** ## Industry Impact: Equipment Stocks Benefit, Foundry Model Not Threatened Bernstein believes the Terafab plan has limited near-term substantive impact on the semiconductor industry but offers several directions worth noting from an investment logic perspective. The report suggests that if the market believes Musk can advance this plan, semiconductor equipment (semicap) stocks would be the most direct beneficiaries. Regarding whether Musk's in-house chip manufacturing will impact existing chip manufacturers, Bernstein holds a relatively optimistic view, suggesting that in an environment of such strong compute demand, any participant will face upside potential far exceeding their own capacity to fulfill orders, and the same applies to memory manufacturers. On the business model front, Bernstein points out that Terafab integrates logic, memory, photomask manufacturing, chip design, and packaging, essentially forming a "super IDM" model. However, this model has proven to be far less efficient than the division of labor system involving "foundry + fabless + specialized memory IDM." Therefore, the report concludes that Terafab currently poses limited threat to foundries like TSMC. ## Bernstein: Not Dismissing Musk, but the Challenge is "Extremely Formidable" Despite the staggering calculation results, Bernstein has not entirely ruled out the possibility of Musk's success. The report states that Musk has previously accomplished things that outsiders deemed impossible, adding, "we would not easily dismiss him." However, the report also clearly states, "A true Terafab feels like an extension to us, especially if adhering to the current compute paradigm." Bernstein proposes two potential alternative paths: First, if Musk cannot proceed independently, he might seek partnerships with existing chip manufacturers; Second, Musk might have some "more unexpected" technological path to break through the current paradigm, but the report admits, "we don't know what that would be." ### Related Stocks - [Tesla, Inc. (TSLA.US)](https://longbridge.com/en/quote/TSLA.US.md) - [GraniteShares 2x Long TSLA Daily ETF (TSLR.US)](https://longbridge.com/en/quote/TSLR.US.md) - [iShares Semiconductor ETF (SOXX.US)](https://longbridge.com/en/quote/SOXX.US.md) - [T-REX 2X Long Tesla Daily Target ETF (TSLT.US)](https://longbridge.com/en/quote/TSLT.US.md) - [Direxion Daily Semicondct Bull 3X ETF (SOXL.US)](https://longbridge.com/en/quote/SOXL.US.md) - [VanEck Semiconductor ETF (SMH.US)](https://longbridge.com/en/quote/SMH.US.md) - [Leverage Shares 2X Long TSLA Daily ETF (TSLG.US)](https://longbridge.com/en/quote/TSLG.US.md) - [Simplify Volt TSLA Revolution ETF (TESL.US)](https://longbridge.com/en/quote/TESL.US.md) - [Invesco Semiconductors ETF (PSI.US)](https://longbridge.com/en/quote/PSI.US.md) - [State Street® SPDR® S&P® Smcndctr ETF (XSD.US)](https://longbridge.com/en/quote/XSD.US.md) - [Direxion Daily TSLA Bull 2X Shares (TSLL.US)](https://longbridge.com/en/quote/TSLL.US.md) ## Related News & Research - [Musk found liable to Twitter shareholders in fraud lawsuit over $44 billion takeover, Bloomberg says](https://longbridge.com/en/news/279998260.md) - [Einride secures regulatory approval in Texas](https://longbridge.com/en/news/280314062.md) - [Kulicke & Soffa Expands Memory Solutions Portfolio | KLIC Stock News](https://longbridge.com/en/news/280257681.md) - [Elon Musk Says SpaceX Will Beat 'Everyone Else Combined' In AI, Warns Non-AI Firms Risk Extinction](https://longbridge.com/en/news/279383442.md) - [Arm Expands Compute Platform to Silicon Products in Historic Company First | ARM Stock News](https://longbridge.com/en/news/280350392.md)