Summary of Key Points from the Conference Call Industry Overview - The conference call focuses on the **AI Supply Chain** within the **semiconductor industry**, particularly highlighting the performance and projections for major companies involved in AI semiconductors, including **Google**, **Nvidia**, **Broadcom**, **TSMC**, and others [1][2][3][4]. Core Insights and Arguments 1. **Google's Capital Expenditure (Capex) Increase**: - Google raised its 2025 capital expenditure budget to **US$85 billion** from **US$75 billion**, indicating strong ongoing demand for AI infrastructure [2]. - The company processes over **980 trillion tokens monthly**, which has doubled from **480 trillion** in May, reflecting robust AI demand [2]. 2. **AI Semiconductor Demand**: - The demand for AI semiconductors remains strong, with a positive outlook for both **US Semiconductors** and **Greater China Semiconductors** [2]. - The **H20 chip** shipments are expected to resume, with a total of **1 million units** anticipated from previous inventory and new production [3]. 3. **TSMC's CoWoS Capacity Allocation**: - TSMC's 2026 CoWoS capacity is projected to grow by **40%-50%**, driven by increasing demand for cloud AI semiconductors [4]. - The forecast for CoWoS-S consumption at TSMC for **AMD** has been raised from **50k to 60k units** in 2025, indicating expected recurring orders for its MI308 in the Chinese market [3]. 4. **Nvidia's CoWoS Booking**: - Nvidia's CoWoS-L booking for 2026 is projected at **510k units**, a **31% year-over-year increase**, implying a total of **5.4 million chips** [8]. - The company is also expected to see **CoWoS-R booking** of around **60k units** for various CPU projects [11]. 5. **Broadcom's CoWoS Allocation**: - Broadcom has booked approximately **145k CoWoS wafers** at TSMC, primarily for projects with Google, Meta, and OpenAI [11]. 6. **Overall AI Capex Trends**: - The top four US hyperscalers are expected to generate **US$550 billion** in operating cash flow in 2025, supporting ongoing investments in AI-related data centers [37]. - The average AI capex/EBITDA ratio is projected to be around **50%** in 2025, indicating strong financial capacity for further investments [38]. Additional Important Insights - **Depreciation Trends**: Depreciation as a percentage of total expenses for data center customers has risen significantly, expected to reach **10-14%** in 2025 due to increased capex [38]. - **Cloud Capex Growth**: The cloud capex tracker indicates a **43% year-over-year growth** in 2025, up from a previous forecast of **39%** [45]. - **AI Semiconductor Revenue**: AI semiconductors are projected to account for a significant portion of TSMC's revenue, with estimates suggesting **25%** of total revenue in 2025 [67]. This summary encapsulates the key points discussed in the conference call, providing insights into the current state and future outlook of the AI semiconductor industry.

Partnerships & Technology - Tesla is collaborating with Samsung to manufacture the next-generation AI6 chip at Samsung's Texas facility [1]

Business Agreement - Samsung Foundry 与 Tesla 签订了价值 165 亿美元（$16.5 billion）的 AI 芯片制造协议 [1] Industry Impact - 该协议表明了 AI 芯片制造领域的巨大商业潜力 [1] - 这也反映了 Tesla 在 AI 技术领域的持续投入和发展 [1]

Musk Announces Deal With Samsung For Tesla AI Chips Made In Texas—Touts ‘Strategic Importance’ https://t.co/m2qQyaa0uG ...

Onshoring and Economic Impact - Onshoring to America, especially high value-added tech manufacturing like chips, has a large fiscal and economic impact without overly stressing the labor force [2] - A major investment deal between a leading American company and a foreign technology company aims to onshore industry in the US [1] - The success of major investments by companies struggling in their respective fields remains to be seen in terms of long-term impact [3][4] Semiconductor Industry and Competition - TSMC has an overall technical lead in the semiconductor space, having built an extraordinary business that has taken out competitors like Intel and Samsung [4][5] - Increased competition is coming for both TSMC and Nvidia [5] - The semiconductor industry is experiencing tremendous growth [5] Market and Investment Trends - The US prime book is heavily weighted and overexposed to the semiconductor space and semiconductor equipment [6] - Hedge funds have dialed back exposure to the semiconductor space ahead of earnings [6] - Extraordinary retail participation, a dramatic increase in margin utilization, and a dramatic decline in short interest have been observed [7] - Market conditions are extraordinarily difficult for hedge funds, requiring near-perfect earnings to live up to the hype [9] - Hedge funds have been underexposed to software services [9] - Hedge funds are expected to rotate towards sectors that are performing well, such as cyclicals versus defensives [10]

Market Trends & Industry Dynamics - Samsung's chip manufacturing gains confidence despite losing market share [1] - Samsung aims to compete with TSMC in the high-margin foundry business [2] - The US government is trying to rebuild the semiconductor industry with the CHIPS Act, offering $39 billion in incentives [8] - The industry desires competition in advanced chip manufacturing to avoid reliance on TSMC [14] Investment Opportunities & Strategic Partnerships - Tesla is a marquee customer for Samsung's foundry business, trusting them with high-end chips for self-driving technology [3] - The deal with Tesla could increase Samsung's foundry business growth by approximately 10% [4] - Elon Musk indicates the deal with Samsung is bigger than $16.5 billion over multiple years [5] - Samsung is expanding capacity in Texas to supply US customers and circumvent Trump tariffs [8][9] Company Performance & Competitive Landscape - Samsung's semiconductor business has been a primary profit driver recently [2] - SK Hynix has surpassed Samsung in HBM (High Bandwidth Memory) chip technology [11] - Samsung is striving to regain ground and secure authorization to sell HBM chips to Nvidia [11] - Intel has not made significant progress in the foundry business, positioning Samsung as a strong alternative to TSMC [13]

Strategic Partnership - Tesla has selected Samsung to manufacture its next-generation AI6 chip [1] - The multi-year deal is valued at $165 billion for Samsung [1] - Tesla's CEO, Elon Musk, signals a focus on driverless cars and robots, potentially impacting short-term profits [1] Technological Advancement - The A16 chip is integral to Tesla's self-driving vehicle initiative [2] - Tesla collaborated with Samsung to optimize the chip's manufacturing [2] Industry Impact - This contract is significant for Samsung, aiding in its expertise development [3] - Samsung has been facing challenges recently [2]

Tesla CEO Elon Musk announced late on Sunday that Samsung will manufacture the car maker’s next-generation AI chip at its upcoming Texas semiconductor plant as part of a deal worth $16.5 billion. (Photo: Apu Gomes via Getty Images)https://t.co/MljKrYXd0S https://t.co/OghX75vOYi ...

Strategic Partnership & Manufacturing Advantage - Tesla gains real-world foundry experience at an exceptionally low cost, enhancing chip design capabilities and manufacturing knowledge [1] - Acquiring core manufacturing expertise becomes a strategic advantage for Elon Musk's businesses due to increasing demand for advanced chips [1] - Samsung's new Texas fab will be dedicated to making Tesla's next-generation AI6 chip, highlighting its strategic importance [4] - Partnership presents manageable downside and strong upside potential for both Tesla and Samsung [3] AI Chip Development & Production - Tesla's AI6 chip is scheduled for mass production in 2027 using Samsung's 2nm node (SF2) [2] - Samsung's SF2 yield is currently 40-45%, lower than TSMC's N2 (over 70%) and Intel's 18A (50-55%) [2] - TSMC will make AI5, which just finished design, initially in Taiwan and then Arizona [5] Risk Mitigation & Alternative Scenarios - If AI6 production falls short of expectations, Tesla could shift the order back to TSMC, absorbing resulting delays [3] - Tesla's edge in real-world AI could significantly reduce the risk of AI6 delays [3] Competitive Landscape - Chip design and manufacturing could become a core competitive advantage across Elon Musk's businesses if AI6 reaches mass production smoothly [4] - Samsung may not fully catch up with TSMC in advanced nodes but has discovered a new business model involving customers in the manufacturing process [4]

Strategic Collaboration & Technological Advancement - Elon Musk and Tesla have a significant opportunity to engage in wafer fabrication at a very low cost, enhancing chip design capabilities and future bargaining power with foundries [1] - Mastering core chip manufacturing technologies is a strategic advantage for Elon Musk's ventures in the long term, given the increasing demand for advanced chips [1] - Tesla's AI6 chip, expected to be in mass production in 2027, will utilize Samsung's 2nm (SF2) advanced process [1] - Samsung's new Texas fab will be dedicated to making Tesla's next-generation AI6 chip, highlighting its strategic importance [2] Manufacturing & Production Risks - Samsung's SF2 process currently has a yield of 40-45%, lower than TSMC's N2 (over 70%) and Intel's 18A (50-55%) [1] - It is difficult to predict whether Samsung can smoothly mass-produce AI6 with SF2 as scheduled, despite the use of GAA technology similar to SF3 [1] - The worst-case scenario for Tesla if AI6 production is not as expected is to transfer orders back to TSMC and bear the impact of AI6 delays [2] Competitive Advantages & Business Models - If AI6 can be mass-produced smoothly, chip design and manufacturing will be a core competitive advantage for Elon Musk's ventures, allowing for more flexible planning and lower costs [2] - Samsung has identified a new business model that allows customers to participate in manufacturing, even if it cannot fully catch up with TSMC in advanced processes [2] - TSMC will manufacture AI5, which has just finished design, initially in Taiwan and then Arizona [3]