AI Graveyard: Why Major Artificial Intelligence Projects and Startups Are Failing

The artificial intelligence industry continues to attract unprecedented investment, generate ambitious technological promises and reshape the global digital economy. Yet, behind the rapid emergence of new products and companies, an increasingly significant number of projects are struggling to survive. From highly anticipated consumer devices to sophisticated software platforms, several initiatives have been discontinued, restructured or forced to abandon their original ambitions.

The growing number of setbacks illustrates a fundamental reality: technological innovation and commercial success do not necessarily go hand in hand. According to figures from S&P Global Market Intelligence, approximately 42% of corporate AI initiatives are eventually abandoned. Limited financing, technical shortcomings, fierce competition, difficulties expanding operations and insufficient customer demand are among the main factors behind these decisions.

This trend is creating what could be described as an AI graveyard, populated by projects that initially generated considerable excitement but failed to establish sustainable business models.

Even Technology Giants Are Struggling to Get AI Products Right

The difficulties facing AI developers are not limited to emerging startups. Some of the world's most influential technology companies have also encountered obstacles when attempting to introduce new artificial intelligence experiences.

OpenAI provides a particularly revealing example. Its efforts to transform ChatGPT into a broader digital platform encountered resistance after a major interface redesign combined several functions into a single environment. The changes generated criticism from users who considered the new experience unnecessarily complicated, prompting the company to restore the more familiar interface.

Other initiatives followed a different path. ChatGPT Atlas, an independent AI-powered browser, was discontinued, while Operator's capabilities were increasingly integrated into ChatGPT. The company's image-generation strategy also moved towards incorporating tools directly into its main platform rather than maintaining separate destinations.

Sora, OpenAI's video-sharing platform, also joined the list of discontinued projects in March 2026, following difficulties related to operating expenses and retaining users.

These developments demonstrate how even well-funded companies must continuously reassess whether standalone products offer sufficient value to justify their continued development.

Character.AI: From Startup Success to Bankruptcy

The story of Character.AI highlights the challenges facing consumer-focused artificial intelligence businesses.

Established in 2021 by former Google researchers Noam Shazeer and Daniel De Freitas, the platform allowed users to interact with customizable AI personalities. Its popularity grew rapidly, attracting substantial investment and reaching a valuation of approximately $1 billion in 2023 after securing $150 million in funding.

However, the company's trajectory changed following Google's $2.7 billion agreement in 2024 to license its technology, alongside the return of its founders to the search giant.

Character.AI subsequently faced mounting legal and regulatory difficulties, particularly concerning allegations involving interactions between its chatbots and underage users. Disputes over intellectual property also added to its challenges.

The company filed for bankruptcy in October 2026, citing operational, legal and regulatory pressures. Disney emerged as an opening bidder with a $20 million proposal for certain intellectual property rights. The platform has continued operating during the sale process, although its longer-term future remains uncertain.

Smaller AI Startups Face an Increasingly Competitive Market

The difficulties experienced by smaller companies further illustrate how quickly competitive advantages can disappear in the artificial intelligence sector.

Relay, an AI-based workflow automation platform, ceased operations in August 2026. Initially positioned as an alternative to established automation services, it struggled to maintain its independent market position as larger technology companies introduced comparable capabilities within their existing products.

Ollie, a digital assistant designed to help families organize household responsibilities, also announced its closure. Launched in June 2026, the service connected information from calendars, emails, school communications and task lists. Despite its privacy-focused positioning, the company struggled to attract sufficient users, distinguish itself from competitors and secure the financing needed to continue.

Another example is Huxe, an audio application developed by former Google NotebookLM engineers. Its technology transformed written information into conversational, podcast-style content. However, the expansion of similar features across larger platforms, including Spotify, reduced its ability to compete independently. The company shut down in May 2026.

Yupp, which offered users the opportunity to compare responses generated by hundreds of AI models, also closed in March 2026 after failing to achieve the level of product-market fit required to sustain its operations. At its peak, the platform provided access to more than 800 models.

Figgs AI, meanwhile, operated between 2023 and 2024, offering customizable AI characters for interactive storytelling and role-playing. Although it reportedly attracted more than one million users, the financial burden of maintaining a free service ultimately proved unsustainable.

AI Hardware: Ambition Meets Consumer Expectations

The difficulties extend beyond software and online services. Several companies have attempted to introduce dedicated AI devices, hoping to create new ways for consumers to interact with digital assistants. However, the gap between ambitious marketing campaigns and everyday performance has proved difficult to overcome.

The Humane AI Pin became one of the most prominent examples. The wearable device was designed to provide artificial intelligence services without relying on a traditional smartphone screen. Despite attracting approximately $230 million in investment and considerable media attention, it struggled with performance and reliability.

Concerns intensified after the company warned customers against using its charging case because of a potential battery fire hazard. Humane subsequently closed its AI Pin business in February 2025, with HP acquiring most of its assets for $116 million.

The Rabbit R1 experienced a different but equally instructive challenge. Introduced at CES in January 2024, the compact AI device generated strong initial interest, with its manufacturer reporting 100,000 units sold shortly after launch. Nevertheless, early reviews highlighted unfinished features, inconsistent performance and limited integration with other services.

Rabbit has continued developing the device, shifting its positioning towards computer control and AI-driven tasks. The company has also announced another hardware initiative, suggesting that it has not abandoned its ambitions in the market.

Microsoft Recall Raises Questions About Privacy and Trust

Not every AI setback is primarily financial or commercial. Microsoft's Recall feature demonstrates how privacy concerns can undermine even technically ambitious projects.

Introduced in 2024, Recall was designed to function as a searchable digital memory for Windows users. The system periodically captured screenshots of computer activity, allowing individuals to retrieve information about previously viewed content and actions.

However, the concept immediately raised concerns about the potential exposure of sensitive personal information. Critics warned that storing searchable records of computer activity could create significant cybersecurity risks, particularly if confidential messages, financial information or passwords became accessible to unauthorized parties.

Microsoft postponed the rollout and subsequently redesigned aspects of the feature to strengthen its security protections. Nevertheless, questions surrounding the effectiveness of these measures have continued to emerge, keeping privacy at the centre of the debate.

Notion Mail Highlights the Limits of Standalone AI Applications

The evolution of Notion Mail offers another perspective on the changing direction of AI development.

Launched in April 2025, the email service sought to use artificial intelligence to improve inbox organization and automate routine tasks. However, user behaviour indicated that many customers preferred employing separate AI agents to manage their communications.

Notion consequently decided to discontinue the product, with the shutdown scheduled for September 22.

The decision reflects a broader shift in the technology market: consumers increasingly expect AI capabilities to work seamlessly within platforms they already use rather than requiring them to adopt additional standalone applications.

A Necessary Correction to the AI Investment Narrative

The growing number of discontinued projects should not necessarily be interpreted as evidence that artificial intelligence is losing momentum. Instead, it highlights the distinction between technological experimentation and the development of commercially viable products.

The sector remains highly dynamic, but companies are facing growing pressure to demonstrate tangible benefits, maintain user confidence and establish reliable revenue streams. Access to advanced models and substantial investment may help accelerate development, but neither guarantees long-term success.

The experiences of OpenAI, Character.AI, Humane, Microsoft and smaller AI developers point to several recurring challenges: building products that genuinely address customer needs, protecting personal information, managing infrastructure expenses and differentiating services in an increasingly crowded market.

Ultimately, the next phase of artificial intelligence may be shaped less by the number of products launched than by the ability of companies to create services that people consistently use and are willing to pay for. In an industry where innovation moves exceptionally quickly, knowing when to discontinue an unsuccessful project may become just as important as knowing when to launch one.