SpaceX Q2 Earnings Was CRAZY: Recap & Highlights

以下是SpaceX 2026年第二季度财报及财报电话会议的综合摘要： **2026年第二季度财报亮点：** * **财务表现：** * 营收同比增长92%。 * 营收：78亿美元（增长92%）。 * 净亏损：5.41亿美元（较此前10亿美元的亏损大幅改善）。 * 调整后EBITDA：35亿美元（增长近200%）。 * 大力投资：星舰项目、星链和AI算力。 * **重要交易与收购：** * 签署了多项行业领先的云服务协议，带来了141亿美元的合同销售额。 * 宣布以600亿美元收购Cursor（全股票交易），以加速AI企业业务机会。 * 获得超过60亿美元的Starshield多年期美国政府合同。 * **产品与技术更新：** * 星舰V3试飞取得更多进展。 * 7月发布Grok 4.5（被评价为“极其出色，具备最佳智能成本效益”）。 * 星链业务“呈爆发式增长”。 * **首席财务官评论：** * 所有业务部门营收增长加速。 * 凭借新的AI算力协议，实现了强劲的经营杠杆和显著的利润率扩张。 * AI算力业务被视为未来几个季度和几年内营收爆发式增长的主要驱动力。 * 在发射、星链用户增长、新企业及政府合作伙伴关系，以及一流AI基础设施（AI算力）方面拥有无可匹敌的领导地位。 * **资产负债表与积压订单：** * 季度末现金、现金等价物和有价证券达1000亿美元。 * 积压订单达475亿美元。 * **主要业务数据（截至8月初）：** * 78次发射。 * 超过1000吨有效载荷送入轨道。 * 1200万星链用户。 * 超过10000颗星链卫星在轨。 * 星链服务已覆盖167个国家。 * **AI算力：** 名义算力消耗达1.4吉瓦（GW）（Q1为1.0 GW，去年为0.4 GW）。 * **各业务部门营收明细（第二季度）：** * **航天（发射）：** 三大业务中规模最小，但支撑了互联互通和轨道AI数据中心。 * **互联互通（星链）：** 营收超过40亿美元。其中，消费者营收24.85亿美元，企业和政府营收18.6亿美元。 * 互联互通营收环比增长32%，同比增长66%。 * 企业营收同比增长63%。 * 政府营收同比增长108%。 * 星链平均用户收入（ARPU）保持在66美元（与上季度持平）。 * **AI业务：** AI解决方案及基础设施营收21.94亿美元。广告营收可能来自X平台。预计吉瓦级（GW）算力上线和变现将加速。 * **各业务部门调整后EBITDA：** * 航天：亏损较小。 * 互联互通：超过25亿美元。 * AI：超过11亿美元。 * **资本支出（CapEx）（第二季度）：** * 航天：11.74亿美元。 * 互联互通：13.67亿美元。 * AI：158.28亿美元（AI算力投入巨大）。 **2026年第二季度财报电话会议重点摘要：** * **埃隆·马斯克开场致辞：** * **总体情况：** 业绩出色，星舰取得进展，星链实现惊人增长，AI算力增长速度超越所有竞争对手。 * **星舰：** 本季度成功进行了几次试飞。第14次飞行将首次将V3星链卫星送入运行轨道。最早在下一次飞行中，将尝试捕获第一级和第二级（包括飞船本身）。目标是每年将100万吨以上（最终达1000万吨以上）有效载荷送入轨道。 * **星链V3：** 每次发射的带宽容量大约增加20倍。马斯克认为，星链可能是唯一能满足（因AI和人形机器人驱动的）带宽巨大增长需求的解决方案。 * **Grok AI：** Grok 4.5已于上月发布，4.6即将推出，4.7也很快会到来。预计改进节奏将大幅加快。未来的Grok模型将整合SpaceX所有历史数据，有望使Grok成为“最佳工程师”。 * **AI算力：** 建设、部署和增长速度比任何人都更快、更高效。预计今年底AI算力上线将超过2吉瓦（GW）（可能远超此数字）。将完全基于英伟达（NVIDIA）使用Rubin芯片进行建设。 * **星智（Starmind）AI卫星：** 采用英伟达Rubin芯片优化，预计明年开始发射。“既然我们要把它放到太空，为什么不也放到地面上呢？” * **格温·肖特韦尔评论：** * SpaceX是全球领先的发射服务提供商。 * **星链用户：** 第二季度新增170万净消费者用户（Q1为140万）。 * **ARPU：** 平均用户收入（ARPU）保持在66美元不变。 * 即将到来的星舰任务将部署V3星链卫星投入使用。 * **星链企业/政府业务：** 认为该业务板块有望达到或超越星链消费者营收。第二季度启动了多项移动运营商合作。随着时间推移，星链的“直连手机”（direct-to-cell）服务客户数量将远超家庭宽带客户。 * **AI算力：** 正在迅速扩展算力容量，以满足内部需求并服务其他公司（如谷歌、Anthropic）。Grok 4.5由于其成本效益，上线后令牌使用量立即增长了两倍。 * **收购Cursor：** 期待尽快整合Cursor团队（尚待监管批准）。 * 正在扩大和深化AI合作。 * **布雷特（首席财务官）评论：** * 营收快速增长，净亏损缩小近一半。 * 目前正在建设基础设施，以实现“每年发射数千艘星舰”的目标。 * **星链航空：** 该网络在航空市场的渗透率不足10%，没有竞争，机会巨大。 * **AI业务增长：** 营收同比增长247%（主要来自AI算力）。调整后EBITDA转为正值，达到11亿美元。 * **资本支出：** 本季度总资本支出约180亿美元，其中158亿美元用于AI算力基础设施建设。 * **AI云服务：** 看到巨大市场兴趣，经济效益也越来越好（每吉瓦（GW）收入更高）。 * **回报期：** 新AI算力的回报期不足一年（被形容为“荒谬”、“疯狂”、“前所未有”）。 * **合同收入：** 已在六个月内签署了额外67亿美元的算力合同收入，即将加速增长。 * **年度经常性收入（ARR）目标：** 预计今年12月年度经常性收入（ARR）将达到1000亿美元，主要由AI算力驱动。 * **投资者问答环节：** * **星链企业业务：** 企业营收“非常稳定”，从未流失过客户。鉴于星链已展现出高可靠性，将积极组建企业销售团队。马斯克预计星链的企业营收将远超其他客户类型。 * **AI算力容量（关键细节）：** * 暂定目标是到明年底，在上线项目中获得**20吉瓦（GW）的电力和冷却能力**（不是GPU，而是基础设施）。 * 芯片是限制因素，而非电力和冷却。他们的目标是让电力和冷却能力远超现有芯片。 * SpaceX预计明年将获得“大量”英伟达GPU。 * 疑因排他性和快速变现能力，英伟达给予了优先待遇。 * **带宽需求：** 由于AI，对互联互通的带宽需求大幅增加。 * **月球基地/机器人：** 人形机器人将有助于建造月球基地和工厂。马斯克表示，这听起来“完全疯了”，但他们可以将“发射到太空的智能规模扩展到地球经济的一千倍”。月球基地是一个严肃的目标。 * **资本支出回报（重申）：** 首席财务官证实AI算力资本支出回报期不足一年。 * **ARR与营收预测（重申与扩展）：** * 埃隆：内部预测**今年12月ARR将达到1000亿美元或更多**，他表示“如果他们现在什么都不做，这基本上也会实现”。 * SpaceX内部此前预测**年营收达1万亿美元**的时间是2031年，现在提前至**2030年**，并且“有非零的可能性”在**2029年**实现。 * 所有三个业务部门的增长都将做出贡献，但AI云服务是迄今为止最大的部分（7月新交易、谷歌/Anthropic加速、Grok/Grok建设、Cursor收购）。 * **发射频率：** 未来发射频率将至少达到每天一次。 * **AI算力定价：** 马斯克对Rubin芯片每瓦变现的“最佳猜测”是30-50美元每瓦（强调这只是猜测）。每瓦智能（intelligence per watt）正在快速增长。 * **内存：** 内存生产短缺（提及Terrafab项目）。 * **星链V3影响：** 预示一旦V3卫星发射，星链营收将发生巨大飞跃。 * **星链移动业务：** 卫星将于明年开始飞行，首批客户服务将于明年底启动。下一代星链移动卫星（明年发射）性能比现有卫星好约100倍。肖特韦尔指出，美国“三大”电信提供商合计每年营收约6000亿美元，SpaceX预计将因更优的服务/覆盖而获得大量客户。 * **Grok训练与租赁：** 随着时间推移，Grok训练占算力的比例将下降，更多算力将租赁给其他公司。从长远来看，Grok训练可能只占用10%的AI算力。 * **星链移动基站：** 构想是部署更小、更频繁的星链移动基站（“几乎就像一个星链碟形天线”），而非传统的、大型且不频繁的移动基站。 * **Cursor收购：** 由于监管限制（交易尚未完成），目前无法讨论合并后的AI产品路线图。

Here's a comprehensive summary of SpaceX's Q2 2026 earnings report and subsequent earnings call: **Q2 2026 Earnings Report Highlights:** * **Financial Performance:** * 92% Year-over-Year (YoY) growth in revenue. * Revenue: $7.8 billion (up 92%). * Net Loss: $541 million (massive improvement from prior losses of $1 billion). * Adjusted EBITDA: $3.5 billion (up nearly 200%). * Aggressive investment: Starship program, Starlink, and AI compute. * **Key Deals & Acquisitions:** * Closed multiple industry-leading cloud service agreements, resulting in $14.1 billion of contracted sales. * Announced agreement to acquire Cursor for $60 billion (all-stock transaction) to accelerate AI enterprise opportunity. * Awarded over $6 billion in multi-year U.S. government contracts for Starshield. * **Product & Technology Updates:** * More progress on Starship V3 test flights. * Released Grok 4.5 in July (described as extremely good and best cost per intelligence). * Starlink is "printing and growing like crazy." * **CFO Commentary:** * Revenue growth accelerated across all business segments. * Delivered strong operating leverage with significant margin expansion, led by new AI compute agreements. * AI compute business is seen as the primary driver of explosive revenue growth in coming quarters and years. * Unparalleled leadership in launch, Starlink subscriber growth, new enterprise and government partnerships, and best-in-class AI infrastructure (AI compute). * **Balance Sheet & Backlog:** * Ended the quarter with $100 billion of cash, cash equivalents, and marketable securities. * $47.5 billion backlog. * **Key Business Statistics (YTD early August):** * 78 launches. * Over 1,000 tons to orbit. * 12 million Starlink subscribers. * More than 10,000 Starlink satellites in orbit. * Starlink available in 167 countries. * **AI Compute:** 1.4 gigawatts (GW) of nameplate compute draw (up from 1.0 GW in Q1 and 0.4 GW last year). * **Revenue Breakdown by Segment (Q2):** * **Space (Launch):** Smallest of the three, but enables connectivity and orbital AI data centers. * **Connectivity (Starlink):** Over $4 billion in revenue. $2.485 billion from consumer revenues, $1.86 billion from enterprise and government revenues. * Connectivity revenues up 32% sequentially and 66% YoY. * Enterprise revenue up 63% YoY. * Government revenue up 108% YoY. * Average Revenue Per User (ARPU) for Starlink remained $66 (identical to last quarter). * **AI:** AI solutions and infrastructure revenues of $2.194 billion. Advertising revenues presumably from the X platform. Expect acceleration of GW online and monetization. * **Adjusted EBITDA by Segment:** * Space: Fairly small loss. * Connectivity: Over $2.5 billion. * AI: Over $1.1 billion. * **Capital Expenditure (CapEx) (Q2):** * Space: $1.174 billion. * Connectivity: $1.367 billion. * AI: $15.828 billion (massive spending on AI compute). **Q2 2026 Earnings Call Highlights:** * **Elon Musk's Opening Remarks:** * **Overall:** Great year, Starship progress, Starlink phenomenal growth, AI capacity growing faster than anyone else. * **Starship:** A few successful test flights in the quarter. Flight 14 will be the first to fly V3 Starlink satellites to operational orbit. Will attempt to catch first and second stages (including the ship itself) as soon as the next flight. Aims to put 1 million+ tons (eventually 10 million+) into orbit annually. * **Starlink V3:** Roughly 20 times increase in bandwidth per launch. Musk believes Starlink may be the only thing that can service the enormously growing demand for bandwidth (due to AI and humanoid robots). * **Grok AI:** Grok 4.5 released last month, 4.6 coming soon, 4.7 not long after. Expects cadence of improvement to increase dramatically. A future Grok model will incorporate all SpaceX data from their lifetime, expected to make Grok the "best engineer." * **AI Compute:** Building, deploying, and growing faster and more efficiently than anyone else. Expecting to end the year with over 2 GW of AI compute online (possibly meaningfully higher). Will build exclusively on NVIDIA using Rubin chips. * **Starmind AI Satellite:** Optimized with Rubin chips from NVIDIA, expected to start launching next year. "If we're going to put it in space, why not also put it on the ground?" * **Gwynne Shotwell's Remarks:** * SpaceX is the leading launch provider globally. * **Starlink Subscribers:** Added 1.7 million net consumers in Q2 (vs. 1.4 million in Q1). * **ARPU:** Average revenue per user remained identical at $66. * Upcoming Starship missions will deploy V3 Starlink satellites into service. * **Starlink Enterprise/Government:** Believes this segment could match or exceed Starlink consumer revenue. Multiple mobile carrier partnerships launched in Q2. Starlink direct-to-cell service customers will far outgrow home broadband customers over time. * **AI Compute:** Rapidly expanding compute capacity for internal needs and for servicing other companies (Google, Anthropic). Grok 4.5 token usage tripled out of the gate due to its cost efficiency. * **Cursor Acquisition:** Looking forward to integrating the Cursor team soon (regulatory hurdles pending). * Expanding and deepening AI partnerships. * **Brett (CFO) Remarks:** * Revenue grew fast, net losses narrowed by nearly half. * Currently building infrastructure to enable the launch of "thousands of Starships per year." * **Starlink Aviation:** Network has sub-10% penetration in the aviation market, no competition, huge opportunity. * **AI Business Growth:** Revenue grew 247% YoY (mostly AI compute). Adjusted EBITDA turned positive to $1.1 billion. * **CapEx:** Of roughly $18 billion total CapEx for the quarter, $15.8 billion was spent on AI compute infrastructure. * **AI Cloud Services:** Seeing enormous interest, increasingly better economics (more money per GW). * **Payback Period:** Less than one-year payback period on new AI compute (described as "absurd," "bonkers," "unprecedented"). * **Contracted Revenue:** Already contracted $6.7 billion additional compute revenue over a six-month period, ramping soon. * **ARR Target:** Expects to reach $100 billion Annual Recurring Revenue (ARR) by December this year, primarily driven by AI compute. * **Investor Question & Answer Session:** * **Starlink Enterprise:** Enterprise revenue is "very sticky," never lost a customer. Will proactively build an enterprise sales team as Starlink has demonstrated high reliability. Musk expects enterprise revenue from Starlink to far exceed other customer types. * **AI Compute Capacity (Critical Detail):** * Tentative target to get **20 gigawatts (GW) of power and cooling** in projects coming online by the end of next year (not GPUs, but infrastructure). * Chips are the limiting factor, not power and cooling. They aim to have power and cooling far in excess of available chips. * SpaceX expects to receive "fuck tons" of NVIDIA GPUs next year. * Suspected priority treatment from NVIDIA due to exclusivity and fast monetization capability. * **Bandwidth Demand:** Huge increase in demand for connectivity due to AI. * **Moon Base/Robots:** Humanoid robots will be useful for building moon bases and factories. Musk says it sounds "totally nuts" but they can scale up to "a thousand times the economy of Earth in terms of intelligence launched into space." Moon base is a serious goal. * **CapEx Payback (Reiterated):** CFO confirms AI compute CapEx has an under one-year payback. * **ARR & Revenue Projections (Reiterated & Expanded):** * Elon: Internal projections for **$100 billion or more in ARR by December this year**, stating it "would basically happen if they did nothing from here." * Internal SpaceX projections for **$1 trillion annual revenue** were previously 2031, now pulled forward to **2030**, with a "non-zero chance" of **2029**. * Growth in all three business segments will contribute, but AI cloud services are the biggest piece by far (new July deals, Google/Anthropic ramping, Grok/Grok build, Cursor acquisition). * **Launch Cadence:** Future launch cadence of at least one flight per day. * **AI Compute Pricing:** Musk's "best guess" for monetization per watt of Rubin chips is $30-$50 per watt (stressed it's a guess). Intelligence per watt is increasing rapidly. * **Memory:** Shortage in memory production (Terrafab project mentioned). * **Starlink V3 Impact:** Flagged massive step change in Starlink revenue once V3 satellites are launched. * **Starlink Mobile:** Satellites begin flying next year, first customer service at end of next year. Next-gen Starlink Mobile sats (flying next year) are ~100 times better than current. Shotwell noted the "big three" US telecom providers do ~$600 billion/year combined, and SpaceX anticipates acquiring many customers due to better service/coverage. * **Grok Training vs. Leasing:** Grok training as a percentage of compute will decrease over time, with far more leased to others. Long-term, perhaps only 10% of AI compute for Grok training. * **Starlink Mobile Stations:** Idea for much smaller Starlink mobile stations ("almost like a Starlink dish") deployed all over, compared to traditional large, infrequent mobile stations. * **Cursor Acquisition:** Cannot discuss combined AI product roadmap due to regulatory limitations (deal not yet closed).