Full Replay: SpaceX Q2 2026 Earnings Call & Investor Q&A

以下是SpaceX 2026年第二季度财报电话会议的摘要，涵盖所有重要新闻项目： **I. 2026年第二季度财务亮点 (对比2025年第二季度):** * **营收:** 78亿美元，较41亿美元增长92%。 * **净亏损:** 5.41亿美元，亏损额减少了4.67亿美元。 * **调整后息税折旧摊销前利润 (Adjusted EBITDA):** 35亿美元，较12亿美元增长191%。 * **业务板块细分:** * **航天业务 (Space Segment):** * 营收：9.62亿美元 (环比增长55%，同比增长29%)。 * 调整后EBITDA：亏损2.05亿美元 (主要由于星舰研发投入加速)。 * **互联互通业务 (Starlink):** * 营收：43亿美元 (环比增长32%，同比增长66%)。 * 调整后EBITDA：26亿美元 (同比增长64%)。 * 营业利润率：提升近3个百分点。 * **AI业务:** * 营收：26亿美元 (环比增长213%，同比增长247%)。 * 调整后EBITDA：实现11亿美元正值 (该业务板块首次实现正向盈利)。 * 云服务协议贡献了16亿美元的增量收入。 **II. 主要业务进展:** * **星舰 (Starship):** * 在过去90天内，成功完成了星舰B-3的两次飞行任务 (第13次和第14次飞行)。 * 第13次飞行展示了核心入轨能力，并成功返回星际基地进行捕获。 * 第14次飞行将是首次将V3星链卫星送入运行轨道。 * 目标是在今年内同时捕获星舰B-3的第一级和第二级。 * 最快在下一次飞行中就可能实现对飞船的捕获 (暂定本月底，待监管批准)。 * **愿景:** 每年向轨道运送超过100万吨载荷，最终达到每年1000万吨 (相比之下，猎鹰系列目前每年约2500吨，世界其他火箭每年约300吨)。 * 在第13次飞行表现之后，隔热罩问题被认为是“已解决”；全面快速可重复使用不存在技术障碍。 * 预计在明年年底前达到载人安全标准。 * 目标是在一年内每天至少进行一次星舰飞行。 * 正在建设每年数千次发射所需的基础设施，包括猛禽发动机/飞行器生产、超级工厂，以及多个发射台 (星际基地、卡纳维拉尔角39A和37号发射台)。 * 轨道燃料转移对于HLS (载人登月系统) 和SpaceX的内部目标至关重要。 * 明年将执行阿尔忒弥斯3号任务 (与猎户座飞船对接)，随后是无人月球货物直运任务，然后是2028年实现“载人登月”。 * **星链 (Starlink) (互联互通):** * 第二季度全球净增170万订阅用户 (有史以来最好的一个季度)，高于第一季度的140万。 * 每用户平均收入 (ARPU) 稳定在每月66美元。 * 服务已覆盖167个市场。 * 卫星星座：约10200颗运行卫星 (其中约9600颗用于宽带)，提供约800太比特/秒的总下行容量。 * **星链V3卫星:** * 能力比V2卫星高出一个数量级 (约10倍)。 * 预计将部署数量级 (约10倍) 更多的V3卫星。 * 预计即使每比特的变现能力下降10倍，交付带宽仍将大致增加两个数量级，星链收入将增长10倍。 * 在最近的星舰发射中成功测试了V3星链卫星的激光链路。 * V3卫星将在即将到来的星舰任务 (第14次飞行) 中投入运行部署。 * 预计到明年第二季度，大约1000颗V3卫星的临界数量将显著改善服务。 * 在不到10年的时间内，可以提供全球大部分地区的互联网服务 (在法规允许的情况下)。 * **企业与政府业务:** * 同比增长108%。 * 与美国航空签署了重大协议。 * 新合作伙伴：西南航空、维珍大西洋航空、伊比利亚航空、爱尔兰航空。 * 第二季度赢得了超过60亿美元的美国政府合同 (太空部队项目)。 * 在航空领域渗透率不足10%，增长空间巨大。 * 企业收入预计将大幅超过消费者收入。 * 客户正在选择搭载星链服务的短途航班。 * **星链移动直连手机服务 (Direct to Cell):** * 与软银、NTT Docomo、Spark新西兰建立了新的合作伙伴关系。 * 近期优先事项：明年通过星舰发射移动V2卫星。 * 明年晚些时候将整合65 MHz的EchoStar频谱 (已获FCC批准)。这将与V2卫星结合，使当前移动服务的能力提高约100倍。 * 预计明年年底前开始提供服务。 * 计划利用EchoStar频谱构建地面组件，通过星链天线安装的毫微微蜂窝基站实现资本支出效率。 * **AI (Grok、算力、合作):** * Grok进展迅速。 * 上个月发布了Grok 4.5 (巨大改进)，Grok 4.6下周推出，Grok 4.7将在3-4周内推出。 * Grok 5 (年底前) 将整合SpaceX长达25年的全部数据，旨在使Grok成为“迄今为止最好的工程师”。 * 正在以比任何人都快的速度建设和部署算力，效率最高。 * 预计到2026年底，算力将超过2吉瓦 (GW)。 * 到2027年底，累计在线算力预计将接近10吉瓦，而非5吉瓦 (电源/冷却基础设施目标是到2027年底达到15吉瓦至20吉瓦)。 * 与英伟达建立了Veriruban架构的独家合作伙伴关系 (被认为是最好的AI计算机)。 * **星智AI卫星 (StarMind AI Satellite):** 优化后的Veriruban NVL 72计算机，预计明年发射。 * 还计划在地面部署NVL 72设计，因为它被认为比典型的机架式系统更有效、更便宜。 * 算力分配：预计约10%的算力用于Grok训练，其余用于推理并出租给第三方。 * 算力供需不平衡持续存在；需求每年增长约200%，而内存供应增长约20%。 * 由于高溢价定价，新部署的算力投资回报期不足一年。 * Grok 4.5的企业反馈非常积极；发布后Token消耗量增加了两倍。 * 收购Cursor团队的交易预计很快完成 (监管障碍几乎清除)，将整合工程和销售能力。 * 已在第三季度初额外签署了67亿美元的云服务收入合同 (从10月开始，为期六个多月)。 **III. 财务展望与资本配置:** * **资本支出 (第二季度):** 约184亿美元，其中约158亿美元用于支持AI计算基础设施。其余用于星舰、卫星生产和地面站。 * 支付了与EchoStar频谱信贷协议相关的8.56亿美元。 * **资本配置重点:** 大力投资星舰开发/生产、下一代星链和AI计算基础设施，专注于资本效率和投资回报率 (ROI)。 * **资本市场:** * 完成了首次公开募股 (IPO)，净募集资金约857亿美元。 * 发行了250亿美元的投资级优先票据 (部分偿还了200亿美元的过渡性贷款)，加权平均利率为5.855%，平均到期日为11.7年。 * 第二季度末，现金、现金等价物和有价证券总额为1000亿美元。 * 积压订单达475亿美元。 * **收入目标:** * 预计到2026年底 (基于12月份收入)，年化收入运行率 (ARR) 将达到1000亿美元，包括Cursor。 * 实现1万亿美元*收入*的内部预测从2031年提前到2030年 (有并非零概率在2029年实现)。

Here's a summary of SpaceX's Q2 2026 Earnings Conference Call, covering every significant news item: **I. Q2 2026 Financial Highlights (vs. Q2 2025):** * **Revenues:** $7.8 billion, up 92% from $4.1 billion. * **Net Loss:** $541 million, an improvement of $467 million. * **Adjusted EBITDA:** $3.5 billion, up 191% from $1.2 billion. * **Segment Breakdown:** * **Space Segment:** * Revenue: $962 million (up 55% sequentially, 29% YoY). * Adjusted EBITDA: Loss of $205 million (due to accelerated R&D in Starship). * **Connectivity Segment (Starlink):** * Revenue: $4.3 billion (up 32% sequentially, 66% YoY). * Adjusted EBITDA: $2.6 billion (up 64% YoY). * Operating Margin: Increased by nearly 3 points. * **AI Segment:** * Revenue: $2.6 billion (up 213% sequentially, 247% YoY). * Adjusted EBITDA: Positive $1.1 billion (first time segment positive). * Cloud services agreements contributed $1.6 billion incremental revenue. **II. Key Business Updates:** * **Starship:** * Completed two successful flights of Starship B-3 (Flight 13 & 14) in the past 90 days. * Flight 13 demonstrated core orbital capabilities and return to Starbase for catch. * Flight 14 will be the first to fly V3 Starlink satellites to operational orbit. * Goal to catch both the first and second stage of Starship B-3 this year. * Could possibly catch the ship as soon as the next flight (tentatively scheduled end of month, pending regulatory approval). * **Aspiration:** Deliver well over 1 million tons to orbit per year, eventually 10 million tons/year (compared to Falcon's current ~2,500 tons/year and rest of world's ~300 tons/year). * Heat shield problem considered "solved" following Flight 13's performance; no technical obstacles to full and rapid reusability. * Expects to achieve human-rated safety levels by end of next year. * Targeting at least one Starship flight per day within a year. * Building infrastructure for thousands of launches per year, including Raptor/vehicle production, Gigafactories, and multiple launch pads (Starbase, Cape Canaveral Pad 39A & 37). * Propellant transfer in orbit is critical for HLS and internal SpaceX ambitions. * Artemis 3 mission (dock with Orion) next year, followed by uncrewed direct-to-lunar cargo mission, then "boots on the moon" in 2028. * **Starlink (Connectivity):** * Added net 1.7 million subscribers globally in Q2 (best quarter to date), up from 1.4 million in Q1. * Average Revenue Per User (ARPU) stable at $66 per month. * Service available in 167 markets. * Constellation: ~10,200 operational satellites (~9,600 broadband) delivering ~800 terabits/second total downlink capacity. * **Starlink V3 Satellites:** * About an order of magnitude (10x) more capable than V2 satellites. * Expect to launch an order of magnitude (10x) more V3 satellites. * Anticipate a roughly two-order-of-magnitude increase in delivered bandwidth and a 10x increase in Starlink revenue even if monetization per bit drops 10x. * Successfully tested V3 Starlink satellites' laser links on the most recent Starship launch. * V3 satellites will be deployed for operational use on the upcoming Starship mission (Flight 14). * Expect critical mass of ~1000 V3 satellites for noticeable service improvement by Q2 next year. * Could deliver the majority of the world's internet in less than 10 years (where allowed). * **Enterprise & Government:** * Revenue grew 108% YoY. * Signed major agreement with American Airlines. * New partners: Southwest, Virgin Atlantic, Iberia, Aer Lingus. * Won over $6 billion in US government contracts in Q2 (Space Force programs). * Less than 10% penetrated in aviation, large headroom for growth. * Enterprise revenue expected to substantially exceed consumer revenue. * Customers are flying shorter-hop flights to ensure they are on Starlink-activated flights. * **Starlink Mobile (Direct to Cell):** * Launched new partnerships with SoftBank, NTT Docomo, Spark New Zealand. * Near-term priority: Launch Mobile V2 satellites on Starship (next year). * Will integrate 65 MHz of EchoStar spectrum later next year (FCC approved). This, combined with V2 satellites, represents a ~100x improvement in capability over current mobile service. * Expect to start providing service by end of next year. * Intends to build out terrestrial components using EchoStar spectrum, leveraging Starlink dish-mounted femtocells for CapEx efficiency. * **AI (Grok, Compute, Partnerships):** * Rapid progress on Grok. * Grok 4.5 released last month (huge improvement), Grok 4.6 coming next week, Grok 4.7 in 3-4 weeks. * Grok 5 (before end of year) will incorporate the entire 25-year corpus of SpaceX data, aiming to make Grok "by far the best engineer." * Building and deploying compute capacity faster than anyone else, with highest efficiency. * Expect to end 2026 with over 2 gigawatts (GW) of compute capacity. * Cumulative compute online by end of 2027 expected to be closer to 10 GW than 5 GW (power/cooling infrastructure targeting ~15 GW to 20 GW by end of 2027). * Exclusive partnership with NVIDIA for their Veriruban architecture (considered the best AI computer). * **StarMind AI Satellite:** Optimized Veriruban NVL 72 computer, expected to launch next year. * Also plan to deploy the NVL 72 design on the ground, as it's seen as a more effective and cheaper design than typical rack-style systems. * Compute allocation: Expect ~10% of compute for Grok training, with the rest for inference and leasing to third parties. * Compute supply/demand imbalance persists; demand growing ~200% annually, while memory supply grows ~20%. * New compute deployments have a less than one-year payback due to premium pricing. * Grok 4.5 enterprise feedback very positive; token consumption tripled post-release. * Acquisition of Cursor team expected to close soon (regulatory hurdles almost cleared), integrating engineering and sales capabilities. * Already contracted an additional $6.7 billion of cloud services revenue in early Q3 (over six months starting October). **III. Financial Outlook & Capital Allocation:** * **Capital Expenditures (Q2):** Approximately $18.4 billion, with roughly $15.8 billion supporting AI compute infrastructure. Remainder for Starship, satellite production, ground stations. * Paid $856 million related to the EchoStar Spectrum Credit Agreement. * **Capital Allocation Priorities:** Aggressively invest in Starship development/production, next-gen Starlink, and AI compute infrastructure, focusing on capital efficiency and ROI. * **Capital Markets:** * Completed IPO, raising approximately $85.7 billion in net proceeds. * $25 billion inaugural investment-grade senior notes offering (partially repaid $20 billion bridge loan), with weighted average interest rate of 5.855% and average maturity of 11.7 years. * Ended Q2 with $100 billion in cash, cash equivalents, and marketable securities. * Backlog of $47.5 billion. * **Revenue Targets:** * Projected to reach $100 billion Annualized Revenue Run Rate (ARR) by end of 2026 (based on December revenue), including Cursor. * Internal projection for reaching $1 trillion in *revenue* moved up from 2031 to 2030 (with a non-zero chance of 2029).