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

The rapid expansion of artificial intelligence has produced a growing record of abandoned products, delayed releases and startups eclipsed before they could establish a durable market. The list spans small companies and major technology groups alike, illustrating how quickly a promising concept can become difficult to finance, build, scale or sell to users. Some projects have disappeared entirely, while others have been redirected or have fallen far short of the expectations surrounding their launches.

Relay, a startup overtaken by larger platforms, reflects one of the pressures facing young AI businesses. New entrants often begin with a focused product or technical idea, only to encounter companies with broader distribution, deeper resources and established customer relationships. In a market where leading platforms can add similar capabilities to existing services, a startup’s early advantage can erode quickly.

The pattern is not limited to independent companies. Internal AI programs can also be shelved by the organizations that created them. S&P Global Market Intelligence has estimated that roughly 42% of corporate AI initiatives are eventually discontinued by their parent companies. That figure suggests that the industry’s highly visible launches represent only part of a larger process of experimentation, retrenchment and cancellation behind the scenes.

The causes are varied. A company may run out of money before a product reaches a sustainable business model, or discover that the underlying technology does not perform reliably enough for the intended use. Competition can change the economics of a project, while the cost and complexity of expanding a service can make a limited pilot difficult to turn into a widely available product. In other cases, potential customers simply may not show enough interest.

Apple’s repeatedly postponed plans for a more advanced Siri are among the most prominent examples of an AI effort that has struggled to meet public expectations. Delays do not necessarily mean a product has been abandoned, but they can still reshape how a company is judged. When an anticipated feature is announced or widely expected, each missed timetable raises questions about technical readiness, product integration and whether the original promise can be delivered at the required quality.

OpenAI’s troubled introduction of a so-called super app has also drawn attention to the risks attached to ambitious AI product strategies. Expanding beyond a core service can create pressure to combine functions, manage a broader audience and demonstrate why a new format is useful. A difficult rollout can become especially significant when it follows intense interest in a company’s technology and when competitors are moving quickly with their own products.

These setbacks offer a different view of an industry often presented through funding rounds, model releases and bold forecasts. Failed or delayed efforts can reveal where technical assumptions were too optimistic, where a market was already crowded, or where a product did not solve a problem compelling enough to change behavior. They also show that access to capital or a recognizable brand does not automatically resolve the practical obstacles between a demonstration and a dependable service.

Still, a project that misses its initial target is not always a simple failure. A shutdown, pivot or delay can preserve resources, redirect teams toward more viable work, or expose lessons that influence later products. The value of tracking these outcomes is less about assigning blame than understanding the conditions under which AI initiatives struggle. The same pressures that end one effort may shape how the next version is designed, funded and introduced.

The next phase of the AI market will likely be defined not only by new announcements but by whether companies can convert experiments into products that users want and organizations can maintain. Investors, customers and developers will be watching for clearer evidence of demand, manageable costs and technology that works beyond limited demonstrations. As more companies test the boundaries of the market, the number of visible successes may grow alongside a less celebrated record of efforts that did not endure.

Source: TechCrunch