The OpenAI graveyard: All the deals and products that haven't happened

Original Article ↗ Hacker News Discussion ↗

OpenAI’s string of shelved products and deals is prompting doubts about its strategy and sky‑high valuation, with many seeing it as emblematic of a broader AI bubble built more on hype than sustainable business models. Commenters question whether ad‑supported LLMs or enterprise sales can ever cover the enormous compute and training costs, drawing parallels to past tech and housing bubbles where investors waited too long for a “killer app” to justify spending. Others argue that aggressive experimentation is normal for a fast‑growing company, but worry that repeated high‑profile misfires, opaque finances and investor‑driven pressure could damage trust in both OpenAI and the wider AI sector.

OpenAI’s Business Health and “Graveyard” Pattern

  • Many see a pattern of splashy announcements (products, partnerships, megaprojects) that quietly die or are reversed once PR value is captured.
  • Commenters describe OpenAI as “financially zombie-like,” heavily dependent on investor cash with no clear path to sustainable profits at its reported valuation.
  • Some argue frequent product shutdowns and reversals erode trust and make OpenAI feel like Google at its worst, but without a proven ads cash cow.

Monetization, Ads, and Economics

  • One camp believes LLM-based interfaces will eventually replace much of search and generate “hundreds of billions” in ad revenue, similar to Google/Meta.
  • Skeptics counter: every attempt at LLM ads so far has been pulled back; users resist overt ads in work tools; and inference costs may dwarf ad income.
  • Debate over whether consumers will tolerate ads if all major models adopt them; some argue users will flee to the least-annoying option, others say they’ll have no real choice.
  • There is worry that making each “search” via LLM far more expensive may break the classic high-margin ad model.

Anthropic, Google, and Competitive Landscape

  • Anthropic is widely perceived as more focused and “healthier,” with an enterprise-leaning strategy and fewer leadership dramas, but still very expensive and vulnerable if the bubble pops.
  • Several argue Google/Gemini has structural advantages: ad machine, cloud, custom hardware, data, and existing enterprise relationships, making it a likely low-cost provider.

AI Bubble and Historical Analogies

  • Many liken the current AI boom to past bubbles (dot-com, housing, railways, NFTs): transformative tech can still be massively overvalued.
  • Comparisons to early PCs and the web: killer apps like spreadsheets and e-commerce productized and threw off cash faster than LLMs have so far.
  • Some say markets can stay irrational for a long time; others think the hype/reality gap is now too large to sustain.

Technical Reality and User Experience

  • Mixed experiences: some strongly prefer Claude for coding; others find it unreliable and favor GPT; some like Gemini’s value bundle.
  • Many note LLMs still hallucinate and are poor at some factual or niche queries; they can be slower and less dependable than traditional search.
  • Concerns that LLM-generated answers siphon traffic from websites, undermining the very content future models need—framed as a “tragedy of the commons.”

Experimentation vs. Lack of Focus

  • One view: lots of failed launches are normal, even healthy, for a company still seeking product–market fit.
  • Counterview: at an ~$800B+ valuation and enormous burn (with Sora’s daily loss figures disputed), this looks less like scrappy iteration and more like undisciplined “spaghetti on the wall.”
  • Some see OpenAI as excellent at hype, fundraising, and politics, but much less proven at building durable, focused, profitable products.