The AI Graveyard Grows: What a 42% Failure Rate Says About the Market

Relay shut down in August. Five years old, built as an AI-powered alternative to Zapier, the workflow automation startup found itself competing against features that OpenAI and Google were quietly building into their own platforms for free. There was no longer a clear reason for it to exist as a stand-alone product, and it folded, with staff joining Google's Chrome team.

Character.ai followed in October, filing for bankruptcy just five years after launch. Founded in 2021 by former Google researchers Noam Shazeer and Daniel De Freitas, the chatbot platform raised $150 million at a $1 billion valuation in 2023. Google later paid a reported $2.7 billion to license its technology and bring the founders back in-house. The underlying business, built around customizable AI characters modeled on celebrities and fiction, couldn't sustain itself on its own.

These are not isolated stories. According to S&P Global Market Intelligence, roughly 42% of AI initiatives are ultimately abandoned by the corporate parents that funded them. The reasons are mundane and familiar: insufficient funding, technical shortfalls, scaling problems, competitive pressure, weak demand. None of this is exotic risk. It's the same attrition that hits any capital-intensive technology cycle, just moving faster because the capital involved is larger and the hype cycle is louder.

Why Even the Biggest Players Aren't Immune

OpenAI, one of the most valuable private companies on earth, has its own list of casualties. In July, the company tried to turn ChatGPT into a broader "super app," splitting the interface into separate "Chat," "Codex," and "Work" modes while rebranding the familiar version "ChatGPT Classic." Users found the redesign confusing and cluttered. OpenAI rolled it back within weeks, restoring the interface people already knew.

The company has had more success folding stand-alone products back into its core app than it has launching new ones outward. ChatGPT Atlas, a stand-alone AI browser, lasted less than a year before its most useful features were absorbed into ChatGPT in August. Operator, an agent built to browse the web and complete tasks independently, met the same fate. Image generation has increasingly been pulled into ChatGPT directly, shrinking DALL-E's role as a separate destination. Sora, the video-sharing platform, shut down in March 2026 after struggling with high operating costs and poor user retention.

Apple's Siri overhaul tells a similar story of delay and cost. Unveiled as part of Apple Intelligence in 2024, the promised Siri was supposed to understand context across apps and actually get things done. Engineering issues and bugs pushed the release back repeatedly. The delay contributed to a $250 million settlement over how Apple marketed the iPhone 16's AI capabilities. Siri AI finally reached the iOS 27 beta this past July, rolling out to English-language users this month, with other languages still pending.

Microsoft's Recall feature, pitched in 2024 as a "photographic memory" for Windows PCs, triggered immediate backlash over privacy and security risks tied to screenshotting sensitive user activity. Microsoft delayed the launch for nearly a year to rework its protections. Even after a redesign, a cybersecurity researcher recently built a tool that could extract data Recall had captured, reopening the question of whether the underlying problem was ever actually fixed.

Smaller bets have fared no better. Notion Mail, an AI-driven inbox tool launched in April 2025, is scheduled to shut down this September after users opted to run their own AI agents against their inboxes instead. Huxe, an AI audio app built by former NotebookLM developers that turned text into podcast-style audio, shut down in May 2026 as Spotify and other large platforms built comparable features into products with existing audiences. Yupp, a free playground letting users compare responses across more than 800 AI models and earn cryptocurrency for voting, shut down in March 2026 after its founders concluded it never found real product-market fit. Figgs AI, a role-play and storytelling chatbot platform that drew more than 1 million users between 2023 and 2024, shut down when keeping the free service running became too expensive to sustain.

Hardware has its own graveyard. The Humane AI Pin raised $230 million and generated enormous early buzz as an alternative to smartphones. Performance problems mounted, then Humane warned customers to stop using its charging case over battery fire risk. The business shut down in February 2025, with most assets acquired by HP for $116 million. The Rabbit R1, unveiled at CES in January 2024, reportedly sold 100,000 units quickly but drew reviews describing it as unfinished and unreliable. Rabbit hasn't given up. It continues updating the R1 and recently announced "Project Cyberdeck," a new device aimed at vibe-coding.

The Bottom Line

The common thread across these failures isn't bad technology so much as bad positioning. Relay, Huxe, and Yupp all built genuinely useful products that larger platforms absorbed or replicated as free features. Notion Mail lost to the agents its own users preferred. Even OpenAI, with essentially unlimited capital and distribution, couldn't force a super app redesign on users who wanted the simpler version back.

For investors, the 42% abandonment rate from S&P Global should be treated as a baseline assumption, not a surprising statistic. Capital allocated to AI ventures, whether inside a public company's R&D budget or a venture fund's portfolio, should be sized and diversified with that failure rate in mind. The products most at risk are narrow point solutions sitting on top of foundation models that platform owners can replicate internally within a product cycle or two. Durable value tends to accrue to whoever owns the distribution, the data, or a genuinely hard technical moat, not to whoever ships the feature first. Watching which startups survive the next twelve months by building something a platform can't easily absorb will tell you more about where this market is heading than any funding round announcement.