Intel Joins Musk's Terafab Chip Factory; Anthropic Deploys Offensive AI for Defense

Intel joins Musk's Terafab chip project as Anthropic launches Project Glasswing — deploying a frontier cybersecurity model with Amazon, Apple, Microsoft, and Google before it can reach public hands.

🗄️ Intel Joins Musk's Terafab AI Chip Project Alongside SpaceX and Tesla

Decoded: Intel confirmed on April 8 that it will join Terafab — Elon Musk's planned AI chip fabrication complex in Austin, Texas — alongside SpaceX, xAI, and Tesla. Terafab was officially announced March 21, 2026, at a special event at the defunct Seaholm Power Plant in Austin. The full-scale facility is projected at approximately $5 trillion in total investment. Intel's role centers on manufacturing processors for Musk's humanoid robotics operations and AI data center buildout. Terafab would be the largest chip fabrication complex in U.S. history if constructed at announced scale. (Reuters, April 8, 2026)

Why it matters: Intel's participation in Terafab is a strategic inflection point for a company that has struggled to win advanced foundry contracts against TSMC and Samsung. Intel Foundry, launched in 2024, has signed limited external customers; Terafab gives Intel a committed anchor tenant at enormous projected scale. For Musk's ecosystem — Tesla, SpaceX, xAI, and the Optimus humanoid robot program — domestic chip fabrication eliminates the Taiwan supply-chain dependency that the U.S. defense and intelligence community has flagged as a national security risk. The $5 trillion projected cost also signals a timeline spanning two-plus presidential terms, making it a policy bet as much as an industrial one. For investors, the announcement is structurally bullish for domestic semiconductor equipment suppliers and grid infrastructure operators that would service a Texas fabrication hub at gigawatt scale.

🔐 Anthropic Launches Project Glasswing — Deploying Claude Mythos Preview for Cyber Defense

Decoded: Anthropic announced Project Glasswing on April 9, a new initiative bringing together Amazon Web Services, Apple, Broadcom, Cisco, CrowdStrike, Google, JPMorganChase, the Linux Foundation, Microsoft, NVIDIA, and Palo Alto Networks to deploy a model Anthropic says it cannot safely release publicly. Claude Mythos Preview — an unreleased frontier model — has already found thousands of high-severity vulnerabilities, including in every major operating system and web browser. Anthropic determined the model's offensive capability is too advanced for general availability. Project Glasswing channels it exclusively into defensive security: launch partners will use Mythos Preview to scan and harden critical software. Anthropic is committing $100 million in usage credits across the program and $4 million in direct donations to open-source security organizations. More than 40 additional organizations maintaining critical software infrastructure have been granted access to scan first-party and open-source systems. (Anthropic official, anthropic.com/glasswing, April 9, 2026)

Why it matters: Anthropic's decision to withhold Mythos Preview from public release is significant: it is the first time a frontier lab has explicitly acknowledged a model is too capable for general deployment and structured controlled access through an industry consortium rather than a staged rollout. The partners list is the de facto A-list of U.S. cyber infrastructure — every major hyperscaler, the two largest pure-play cybersecurity vendors (CrowdStrike and Palo Alto Networks), and JPMorganChase as a financial system proxy. For CrowdStrike (CRWD) and Palo Alto Networks (PANW), early access to a model that can find vulnerabilities across every major OS before it proliferates is a durable competitive moat — the ability to patch before the offensive version reaches adversaries. Anthropic's framing is explicit: AI models have crossed the threshold where they outperform all but the most skilled human researchers at exploiting software. Project Glasswing is a structured attempt to win the defense race before that capability diffuses.

Stay decoded. See you tomorrow.

— The Get AI Decoded Team