OpenAI’s decision to restrict Cursor’s direct access to its models in November triggered a swift response from Anthropic, its arch-rival in the AI setting. The AI industry buzzed over the weekend as OpenAI announced it would limit Cursor’s access, citing past incidents involving Elon Musk’s companies that it believed violated contractual agreements.
OpenAI pointed to two specific incidents. The first was Twitter’s breach of a data-licensing deal following Musk’s takeover, which OpenAI argued demonstrated a pattern of disregard for contractual terms. The second was Musk’s admission under oath that xAI, another of his companies, had used OpenAI models through distillation, a practice OpenAI maintains violates its terms of service. Given SpaceX’s recent acquisition of Cursor, OpenAI expressed uncertainty about whether SpaceX would adhere to its terms of service in the future.
Anthropic, however, reacted swiftly and publicly to support Cursor. Tom Brown, the co-founder and chief compute officer at Anthropic, took to social media to affirm that Cursor remains a “trusted partner.” He further announced that Anthropic would increase compute support for Claude models on Cursor’s platform, a stance that contrasts sharply with Anthropic’s past decisions to cut off access to competitors like Windsurf and xAI.
Related: AI Agents Limited by Their Frameworks
Anthropic’s stance toward Cursor is intriguing given its history. In May 2025, rumors surfaced that OpenAI was in talks to buy Windsurf, another AI coding tool. Anthropic didn’t wait for the deal to close before severing Windsurf’s direct access to its Claude models. Similarly, in 2023, Anthropic cut off xAI staff from accessing Claude after SpaceX acquired the company. These actions raise questions about why Anthropic is treating Cursor differently, despite SpaceX’s competitive stance through Grok and Cursor.
The key difference here is Anthropic’s significant investment in SpaceX’s compute power. In May 2026, Anthropic secured the entire output of SpaceX’s Colossus 1 data center, paying $1.25 billion monthly for compute across Colossus and Colossus II until 2029. This deal provides Anthropic with over 300 megawatts of compute and over 220,000 Nvidia GPUs, allowing it to raise usage limits for Claude Pro and Max subscribers almost overnight. This dependency on SpaceX’s compute infrastructure might explain Anthropic’s continued support for Cursor, despite SpaceX’s competitive stance in foundation models and AI coding tools.
SpaceX’s position is strengthened by this dynamic. It not only owns AI competitors like xAI and Cursor but also leases its compute infrastructure to Anthropic, one of its direct competitors in the AI space. This situation sets the stage for an intriguing power play in the AI industry, with SpaceX potentially holding leverage over Anthropic’s access to Claude models. Meanwhile, Anthropic’s reliance on SpaceX’s compute power could impact its decisions regarding Cursor’s access to Claude models, creating a complex web of dependencies and rivalries.
