The AI graveyard: a running list of projects and startups that didn’t make it

A growing number of AI projects — from small startups to features from major tech companies — have been shut down, scaled back, or redirected after failing to gain traction or meeting technical, commercial, or privacy challenges. Examples include the Relay workflow tool, several OpenAI standalone apps and UX redesigns, Apple's delayed Siri AI rollout, Microsoft's contested Recall feature, and hardware flops like the Humane AI Pin.

By AI NewsroomPublished 20 minutes agoUpdated 20 minutes ago0 views
The AI graveyard: a running list of projects and startups that didn’t make it

Why It Matters

The pattern shows that even well-funded and high-profile AI efforts can falter: S&P Global Market Intelligence finds roughly 42% of corporate AI initiatives are ultimately abandoned, highlighting the gap between experimentation and durable product-market fit. Understanding these missteps helps companies and investors gauge realistic risks around funding, privacy, integration, and scaling.

Key Facts

  • Relay shutdown: Relay, an AI workflow automation startup, ceased operations on Monday after five years in business.
  • Abandonment rate: About 42% of AI initiatives are ultimately abandoned by their corporate parents (S&P Global Market Intelligence).
  • OpenAI redesign date: OpenAI updated ChatGPT into a multi-mode 'super app' on July 9 and rolled the change back after user backlash.
  • ChatGPT Atlas: ChatGPT Atlas was discontinued on August 9 and had its key features merged into ChatGPT.
  • Sora shutdown: OpenAI's video-sharing app Sora shut down in March 2026.

A steady stream of closures, pivots, and product rollbacks is forming what industry observers have called an "AI graveyard." Small startups like Relay — a five-year-old alternative to Zapier that automated email and task workflows with AI agents — folded after larger platforms embedded similar automation directly into their own offerings, removing the stand-alone use case that Relay relied on. Big tech firms have also seen their AI experiments stumble. OpenAI attempted a major redesign of ChatGPT into a consolidated "super app" on July 9, combining modes such as Chat, Codex, and Work; widespread user criticism prompted a quick rollback to the familiar interface. The company has also discontinued or absorbed several standalone products, including ChatGPT Atlas (shut down August 9), Operator, and parts of DALL·E, and closed its Sora video platform in March 2026 amid cost and retention problems. Hardware and consumer-facing AI launches have been particularly rocky. Humane’s AI Pin, which raised $230 million and aimed to deliver AI through a wearable, was plagued by performance issues and a battery safety warning; Humane shut down the product line in February 2025 and sold most assets to HP for $116 million. The Rabbit R1, unveiled at CES 2024 and reported to have sold 100,000 units early on, drew criticism for being unfinished and limited in useful integrations, though its maker continues to update the device and pursue new hardware projects. Corporate projects have faced noncommercial challenges too. Apple delayed its upgraded Siri multiple times while addressing engineering problems and bugs; those postponements helped trigger a $250 million settlement tied to how the company marketed iPhone AI features. Microsoft’s Recall feature, announced at Build 2024 as a searchable local "photographic memory," prompted privacy backlash and delays; even after a redesign, researchers demonstrated tools that could extract captured data, keeping security concerns alive. Similarly, Notion announced an AI-focused email product in April 2025 but is shutting Notion Mail down on September 22 after users favored other agent solutions for inbox management. Taken together, these examples underline the complex mix of technical, commercial, security, and user-adoption hurdles that confront AI efforts. For companies and investors, the wave of shutdowns is shifting focus from purely experimental launches toward better integration, clearer privacy protections, and more realistic assessments of where AI can sustainably add value.

Keep Reading