Anthropic vs Apple AI

The Failed Siri Deal and What It Reveals

In late 2025, Apple evaluated both Anthropic and OpenAI to power a dramatically improved Siri, testing their models on Apple's Private Cloud Compute infrastructure. The negotiations with Anthropic collapsed over pricing — Anthropic reportedly demanded several billion dollars annually, with the fee doubling each year for three years. Apple ultimately chose Google's Gemini at a reported ~$1 billion annual fee, white-labeled so users would never see Google's branding.

This episode reveals the fundamental tension between these companies. Anthropic, growing at breakneck speed toward $20B in annualized revenue, priced its technology at what it believed frontier AI was worth. Apple, accustomed to being the platform that extracts value from partners, wasn't willing to create a dependency on an AI startup at those terms. The result: Apple gets good-enough AI from Google at a fraction of the cost, while Anthropic continues to grow explosively through its own channels.

Model Capability vs. Distribution Power

On raw AI capability, there is no contest. Claude Opus 4.6 is among the most capable large language models in existence, with a million-token context window, state-of-the-art coding performance, and autonomous computer use that lets Claude open apps, navigate browsers, and fill spreadsheets on your behalf. Apple Intelligence, by contrast, is deliberately modest — optimized for on-device tasks like writing assistance, notification summaries, and image generation, with heavier lifting outsourced to Google Gemini or OpenAI's ChatGPT.

But distribution is Apple's superpower. Two billion active devices mean Apple Intelligence reaches more people on day one than any API-first company can dream of. When Apple ships a feature, it becomes the default for hundreds of millions of users who never chose it — they just updated their iPhone. Anthropic must win users one by one, through superior quality and word of mouth.

Competing Visions of the Agentic Future

Both companies are building infrastructure for agentic AI, but their approaches could not be more different. Anthropic created the Model Context Protocol (MCP), an open standard that lets any AI model connect to any tool or data source. With over 17,000 MCP servers deployed, MCP is becoming the TCP/IP of the agentic web — a neutral protocol layer that benefits the entire ecosystem while reinforcing Anthropic's position at the center of it.

Apple's answer is App Intents, a proprietary framework that lets iOS apps expose their capabilities to Apple Intelligence and Siri. Every app that adopts App Intents becomes a service that Apple's AI agents can discover and invoke. Combined with Apple Pay for transactions, this creates what may be the most seamless agentic commerce layer in consumer computing — but one that only works within Apple's walled garden.

The philosophical divide is stark: Anthropic is building an open protocol for a multi-platform agentic web; Apple is building a closed service mesh for its own ecosystem. Both strategies can succeed, but they serve different futures.

Privacy: Two Different Definitions of Safety

Anthropic and Apple both claim safety as a core value, but they mean entirely different things. Anthropic's safety work centers on alignment — ensuring powerful AI systems behave as intended through Constitutional AI, mechanistic interpretability, and its Responsible Scaling Policy. This is about making sure the AI itself is trustworthy.

Apple's safety philosophy is about data privacy — ensuring that user information never leaves the device unnecessarily, that cloud processing happens on Apple Silicon with cryptographic guarantees, and that no one (including Apple) can access your data. Apple doesn't need its AI to be the smartest; it needs it to be the most private.

These aren't competing approaches so much as complementary ones addressing different threat models. The ideal AI system would be both aligned (Anthropic's concern) and private (Apple's concern). Today, users must choose which they prioritize.

Developer Ecosystems and the Tools Layer

Anthropic's developer story is Claude Code — an agentic coding tool that has reached $2.5 billion in annualized revenue and now accounts for roughly 4% of all GitHub commits. The Claude Agent SDK, combined with MCP, gives developers a complete toolkit for building autonomous AI applications. This is a horizontal play: Claude Code works everywhere, on any platform, with any codebase.

Apple's developer story runs through Xcode 26.3, which now integrates both Claude Agent and OpenAI's Codex for agentic coding capabilities. This is notable — Apple is a customer of Anthropic's technology even as it competes with Anthropic's vision. For developers building iOS and macOS apps, Xcode's AI integration is deeply convenient. But it only serves Apple's platform, and Apple controls which AI providers are available.

The irony is rich: Anthropic's Claude powers coding inside Apple's own IDE, meaning Anthropic is simultaneously a competitor to Apple's AI strategy and a critical supplier to Apple's developer tools.

Scale, Revenue, and Strategic Independence

Anthropic closed a $30 billion Series G at a $380 billion valuation in February 2026 — the second-largest venture deal in history. Its revenue has grown from ~$9 billion annualized at end of 2025 to nearly $20 billion by March 2026, with 80% coming from enterprise customers. This is extraordinary growth, but Anthropic remains dependent on Amazon and Google for compute infrastructure, a vulnerability that Apple — which designs its own silicon and runs its own data centers — does not share.

Apple's AI revenue is harder to isolate, but the company earned roughly $900 million from generative AI apps through the App Store in 2025, and its broader services business (where AI increasingly drives engagement) generates over $80 billion annually. Apple doesn't need AI to be its primary revenue driver; it needs AI to keep people buying iPhones and subscribing to services. This difference in business model fundamentally shapes each company's AI incentives.