Ed Zitron published a piece this week that HN ate alive: 191 points, 238 comments. The title says it all: “OpenAI Is Just Another Boring, Desperate AI Startup”.
I do not agree with everything Zitron writes. But his inventory of dead OpenAI products and collapsed deals is worth reading because the list is longer than most people realize.
The graveyard
Sora — launched September 2025, killed March 2026. Six months. Burning an estimated $1 million per day in inference costs against $2.1 million in total lifetime revenue. Each 10-second clip cost OpenAI roughly $1.30 to generate. We covered this already. It is still the most spectacular unit economics failure in recent AI history.
Disney $1 Billion Deal — announced with fanfare. $1B equity investment, 200+ Disney characters available in Sora, Disney+ integration. Disney was notified less than one hour before Sora’s shutdown was made public. The deal is dead.
Windsurf $3B Acquisition — signed in May 2025, collapsed in July. Microsoft, OpenAI’s largest investor, had IP rights that would have given it access to Windsurf’s codebase. The CEO refused. Google swooped in within 72 hours, hired the CEO, co-founder, and 40 senior engineers for $2.4 billion.
Stargate $500B Data Center Initiative — announced January 2025 with SoftBank and Oracle. Described as the largest AI infrastructure investment in history. Reality: no staff hired, no data centers developed under the joint venture. Oracle has pushed delivery from 2027 to 2028. At least one specific Stargate data center was reportedly canceled over contract disputes.
AMD GPU Partnership — “definitive agreement” announced October 2025. First gigawatt of capacity due by end of 2025. AMD has reported zero revenue from OpenAI and did not change guidance.
ChatGPT Instant Checkout — launched September 2025 with Shopify, Etsy, Stripe. Promised 1 million+ merchants. Actual integrations at shutdown: roughly a dozen. Near-zero sales. Being quietly removed.
GPT-4o — deprecated nine months after release, with two weeks notice. 800,000 users who had formed bonds with the model lost access.
Operator Standalone — launched January 2025 as a breakthrough autonomous browser agent. Deprecated August 2025, folded into ChatGPT as “agent mode”.
o3 Standalone Model — planned, then canceled in favor of bundling into GPT-5.
The pattern
Announcement. Headlines. Investor confidence. Then quietly: killed, collapsed, stalled, deprecated, or folded into something else.
Zitron’s framing is harsh but hard to argue with: “using the press as a tool to pump its bags”. Every announcement serves the narrative of inevitability. The narrative sustains investor confidence. Investor confidence sustains the next funding round. The funding round sustains the burn rate.
Whether the products actually work, whether the deals actually close, whether the partnerships actually generate revenue, those are secondary to the announcement itself.
The financial picture
OpenAI projects $40 billion in losses by 2027. Profitability is not expected until 2030. Revenue is growing, $2 billion per month by their own numbers, but the cost structure grows faster. Every new model is more expensive to train and serve.
The company raised $122 billion at an $852 billion valuation. To justify that valuation, it needs to become one of the most profitable companies on Earth within a few years. The graveyard of failed products and collapsed deals is the evidence that the path from here to there is not as clear as the press releases suggest.
What I take from this
I am not anti-OpenAI. ChatGPT is useful. GPT-5.4 is impressive. The technology is real.
But there is a difference between a technology company and a narrative company. A technology company ships products that work and generate revenue. A narrative company ships announcements that generate confidence and funding.
OpenAI does both. The question is which one is driving the business. The graveyard suggests the narrative is doing more heavy lifting than the products.
And that should matter to anyone building on top of their platform, because the products you depend on today might be in next quarter’s graveyard.
Sources: Ed Zitron, HN Discussion (191pts), TechCrunch, Tom’s Hardware