The AI Burn Rate: Where the Venture Capital Went While the Product Vanished

As a tech columnist, I spend my days staring at pitch decks that promise the moon and P&L statements that suggest the company is currently orbiting a black hole. Right now, the Silicon Valley narrative is dominated by the 'AI pivot.' Every legacy SaaS tool and every ambitious hardware startup is rebranding as an AI-first entity to capture the current wave of venture capital. But if you look past the hype, you don't see a revolution; you see a graveyard.

In my view, we are witnessing a massive misallocation of capital. Investors are funding 'solutions' for problems that don't exist, or worse, funding products that are rendered obsolete the moment a Big Tech player decides to integrate a similar feature for free. The result is a staggering burn rate for products that never actually materialize into sustainable businesses.

Take the case of Relay. As first reported by TechCrunch, this five-year-old startup attempted to carve out a space as an AI-powered workflow automation alternative to Zapier. Relay allowed users to automate tasks and emails via AI agents, but it ultimately shut down entirely this past Monday. Why? Because the 'moat' Relay tried to build was non-existent. When OpenAI and Google integrated similar automation directly into their existing ecosystems, Relay lost its reason to exist as a stand-alone product. This is the fundamental flaw in the current AI startup model: you aren't just competing with other startups; you are competing with the platforms that provide the very intelligence your product relies on.

Then there is the hardware sector, where the disconnect between investor enthusiasm and product reality is most glaring. TechCrunch notes that the Humane AI Pin raised $230 million from investors. For a device intended to replace the smartphone with a wearable AI interface, that is a colossal sum. Yet, the product was plagued by performance issues and a terrifying safety flaw—Humane had to warn customers to stop using its charging case due to potential battery fire risks. By February 2025, the AI Pin business was shut down, and HP acquired most of the company's assets for $116 million. When you compare the $230 million raised to the $116 million acquisition price, the math reveals a stark reality: a significant portion of that venture capital simply evaporated into a product that failed to function.

Similarly, the Rabbit R1 arrived with immense hype at CES in January 2024. Rabbit claimed to have sold 100,000 units shortly after launch, but as TechCrunch reports, early reviews painted a picture of an unfinished device with unreliable performance and limited integrations. While Rabbit continues to push updates and has announced a new project called "Project Cyberdeck," the R1 serves as a cautionary tale of shipping 'beta' hardware to paying customers in hopes that the software will eventually catch up to the marketing.

Even the titans of the industry are bleeding resources on AI bets that lead nowhere. OpenAI, despite its valuation, has a history of casualties. TechCrunch reports that OpenAI launched ChatGPT Atlas, an AI-powered web browser, only to discontinue it on August 9 after less than a year. They did the same with Operator, an AI agent for web browsing. Most telling is the fate of Sora, OpenAI's video-sharing platform, which shut down in March 2026 after struggling with user-retention challenges and high operating costs. Even the 'super app' redesign of ChatGPT on July 9—which attempted to merge 'Chat,' 'Codex,' and 'Work' modes—was so poorly received by users that OpenAI had to roll it back quickly.

Corporate giants are not faring any better. Microsoft's Recall feature, pitched at the 2024 Build conference as a 'photographic memory' for PCs, became a privacy nightmare. Despite a year of delays and a redesign to address security concerns, a cybersecurity researcher recently demonstrated a tool capable of extracting Recall's captured data, proving that the fundamental privacy risks remain. Meanwhile, Apple's attempt to revamp Siri with 'Apple Intelligence' resulted in repeated delays due to bugs and engineering issues, which TechCrunch reports contributed to a $250 million settlement regarding the marketing of the iPhone 16's AI capabilities. Siri AI didn't even hit the iOS 27 beta until this past July.

Perhaps the most damning evidence of this trend is the sheer volume of failure. S&P Global Market Intelligence, as cited by TechCrunch, estimates that approximately 42% of AI initiatives are abandoned by their corporate parents. Whether it is due to technical hurdles, insufficient funding, or simply weak user demand, nearly half of these expensive bets are ending in failure.

We see this even in the software-as-a-service (SaaS) world. Notion Mail launched in April 2025 to help users automate their inboxes, but the company discovered a fatal truth: users preferred using separate AI agents to manage their emails rather than using Notion's dedicated tool. Because of this, Notion Mail is set to close its doors on September 22.

From my perspective, the 'AI graveyard' isn't just a collection of unlucky startups; it is a symptom of a venture capital ecosystem that is prioritizing 'AI' as a buzzword over actual product-market fit. When you spend hundreds of millions of dollars on a wearable that might catch fire or a browser that lasts six months, you aren't innovating—you're gambling. The industry needs to stop asking 'How can we add AI to this?' and start asking 'Does this product actually solve a problem that a free update from OpenAI or Google won't solve tomorrow?' Until then, the burn rate will continue to climb, and the graveyard will continue to grow.