SpaceX Aug 17 1

Rating Company Date 17 August 2026 Buy SpaceX Company Update North America United States Reuters Bloomberg Exchange Ticker Price at 14 Aug 2026 (USD) 140.00 SPCX.OQ SPCX US NSM SPCX Price Target 235.00 Industrials 201.80 - Space Technology 52-week range 108.27

Double clicking on Cursor Valuation & Risks

Edison Yu Following the close of the $60bn all-stock acquisition, we take a closer look at Research Analyst Cursor and how it can augment SpaceX’s AI efforts going forward. In short, we see three key benefits: 1) instance presence across many blue-chip enterprises who are Bryan Kraft already using Cursor at large scale; 2) access to a proprietary data flywheel derived Research Analyst from >1m real-world users generating coding interactions which in turn represent high-quality reinforcement-learning data for models; and 3) vertically integrating the application/harness layer (Cursor) with a frontier model (Grok) and large amount Melissa Weathers Research Analyst of compute (>2 GW exiting this year). We also refine our forecasts to account for an earlier than expected close of the Cursor deal and separately, more favorable neocloud pricing next year. Reiterate Buy and $235 price target. Roshan Ranjit, CFA Research Analyst Cursor 101 Cursor is an AI-native software development environment sitting at the coding Laura Li application layer. It began as a Visual Studio Code fork and has since become a Research Associate multi-surface agent platform: desktop IDE (integrated development environment), CLI (command-line interface), and integrated agent (web, GitHub, cloud Key changes background, etc…). The value proposition is simple: a developer states intent in natural language and Cursor searches the repository, plans the changes, edits files, EPS (USD) 0.96 to 1.06 ↑ 10.9%

runs commands/tests, and iterates until the task is done. The user stays in the 46,041 to Revenue (USD)m) ↑ 0.9% 46,439 decision loop reviewing diffs, answering clarifying questions, and setting how Source: Deutsche Bank much autonomy the agent is allowed.

1. Developer inputs in ordinary language: “add a login page” or “fix the checkout bug” 2. Cursor looks through the project’s files to find the relevant pieces 3. It writes a plan, then makes the changes itself: rewriting files, running commands, and checking that nothing obvious broke 4. If a check fails, it tries again; the loop continues until the job is done or the user intervenes 5. User reviews what changed, answers clarifying questions, and sets how much freedom the AI is allowed to have

For the actual product suite, we view it as 4 core offerings. A developer can use one

Deutsche Bank Securities Inc. IMPORTANT RESEARCH DISCLOSURES AND ANALYST CERTIFICATIONS LOCATED IN APPENDIX 1. Deutsche Bank does and seeks to do business with companies covered in its research reports. Thus, investors should be aware that the firm may have a conflict of interest that could affect the objectivity of this report. Investors should consider this report as only a single factor in making their investment decision.

17 August 2026 Space Technology SpaceX

of them or all of them. What changes is how much of the work the AI is allowed to take on.

n Tab is the basic offering for autocomplete or next-edit prediction: as a user writes code, Cursor guesses the next few words or lines. n Agent is the more advanced offering. A programmer describes an action. Then Cursor writes a plan, changes several files, and shows a before-and- after of each change. The user watches and approves instead of typing every line. n Cloud Agents do the same kind of work except on a remote computer, for hours or days, and several can run at once. The user can leave and come back later to review finished work (i.e., more like assigning tasks overnight). n Automations are agents that “wake up” on their own: for example when overnight tests fail, during a certain time period, when someone opens a review. The prompt is “keep the main project healthy,” as opposed to “help me write this function.”

Harness and models One of Cursor’s main advantages is being able to serve as an effective harness. Users can route different tasks to different models to optimize performance…