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

About 42% of corporate AI initiatives are ultimately abandoned, according to S&P Global Market Intelligence, as companies cite funding, technical, scaling, competition, and demand challenges. The article reviews shutdowns and setbacks including Relay, OpenAI products, Notion Mail, Humane AI Pin, Huxe, and Yupp; Humane raised $230 million before HP acquired most of its assets for $116 million. It also describes continuing efforts such as Rabbit’s R1 updates and Apple’s delayed Siri rollout, framing the cases as lessons for the industry rather than a scorecard of failures.

Analysis

The economic signal is less “AI is failing” than “stand-alone AI features are hard to monetize when incumbents can bundle them into products with existing distribution.” That tilts value toward platforms that can improve retention or pricing across a broad installed base, while pressuring narrow apps whose differentiation is a model interface rather than proprietary data, workflow integration, or a durable community. It does not establish that bundling is profitable: inference costs, churn, and cannibalization can absorb the benefit.

For AAPL, the key issue is execution credibility: delayed Siri functionality can weaken the upgrade-cycle narrative if users do not perceive a meaningful device-level benefit. For MSFT, Recall is a different risk—privacy and security concerns could constrain adoption or add remediation costs, though this article alone does not imply a material earnings hit. GOOG’s distribution is an advantage, not immunity from the same product-market-fit test. SPOT’s AI audio may support engagement, but feature parity could limit pricing power; monitor engagement and paid conversion rather than launch activity. HPQ’s asset acquisition is not evidence of a material AI revenue stream absent disclosure on integration or sales.

Near term, this roundup is backward-looking and unlikely to justify a broad AI-sector repricing. Over 1–3 months, watch product usage, monetization disclosures, and security incidents; over 6–18 months, the differentiator is whether AI lifts retention, paid conversion, or hardware replacement. Contrarian point: product shutdowns may be portfolio discipline, not proof that enterprise or consumer AI demand is collapsing. The thesis weakens if bundled features show measurable incremental revenue or retention without margin deterioration; it strengthens if usage fails to convert or privacy events force rollbacks.

Key Decisions for Investors

• No standalone directional trade from this retrospective article. Avoid treating product closures as a clean short signal on AAPL, GOOG, MSFT, HPQ, or SPOT; the disclosed facts do not quantify earnings exposure.
• Watch AAPL for evidence that the new Siri improves device engagement or supports upgrade demand. Reassess the execution-risk view if adoption and usage are disclosed as strong; it is reinforced by further delays or a weak iPhone upgrade outlook.
• For MSFT, treat Recall as a conditional security catalyst, not a current earnings thesis. Escalate risk if a verified exploit exposes sensitive data or Microsoft limits deployment; de-risk that concern if independent testing supports the redesigned safeguards and adoption proceeds.
• Track SPOT’s AI audio through listening engagement, paid conversion, and content costs. Feature launches without improvement in those measures would argue against assuming durable monetization; sustained gains would falsify the commoditization concern.
• For GOOG and other scaled platforms, favor evidence of monetization and retention over product-announcement counts. Revisit the platform-bundling advantage if AI usage raises costs without improving customer retention, paid conversion, or pricing.