On Super Bowl Sunday 2026, roughly 130 million Americans were told to visit a two-letter website: ai.com. The thirty-second advertisement, wedged between beer commercials and celebrity cameos, promised an autonomous AI agent that could trade stocks, automate workflows, and — in a detail that may one day define this era — update your online dating profile. What those 130 million visitors found when they arrived was a loading spinner, a crashed website, and the most honest metaphor the artificial intelligence industry has yet produced.
The story behind ai.com, as told by the engineer who built it, is a parable about an industry that has confused naming things with building them, and discovered that the market doesn’t know the difference. Peter Girnus, who identifies himself as the lead engineer at ai.com, laid out the economics in a post on X that has since ricocheted across the tech world: of the $78 million budget, $70 million went to the domain name, $8 million went to the Super Bowl ad, and the engineer got $500 and access to Cloudflare’s free tier. The ratio — 156,000 to 1, marketing to engineering — is not, Girnus argues, a bug. It is the business model.
When the Most Expensive Asset Is the Name on the Door
The domain ai.com has a storied provenance. Its previous owner was OpenAI, the Sam Altman-led company that used the two-letter address as a simple redirect to ChatGPT — a product built by thousands of engineers, running on billions of dollars of compute infrastructure, serving hundreds of millions of users. OpenAI’s decision to sell the domain, and the identity of the buyer, tells a story about where value resides in the current AI gold rush. The buyer was Kris Marszalek, the CEO of Crypto.com, a man whose previous claim to marketing fame was spending $700 million to rename a Los Angeles basketball arena and hiring Matt Damon to tell America that “fortune favors the brave” — approximately six months before cryptocurrency markets lost 70% of their value.
Marszalek paid for the domain in cryptocurrency. Girnus noted he was told this was “tax efficient” and that he had “learned not to ask follow-up questions about things that are ‘tax efficient.’” The acquisition, first reported by Reuters, was part of a broader pivot — or, in the preferred industry terminology, a “convergence” — from crypto to artificial intelligence. As Girnus defined it: “Convergence means the last bubble popped so you inflate the next one using the same PowerPoint deck with different nouns.”
A Super Bowl Crash That Wrote Its Own Review
The Super Bowl ad aired during the fourth quarter of the game. At $266,666 per second of airtime, each tick of the clock cost more than the entire engineering budget behind the website it promoted. Second fourteen of the commercial featured the logo — a planet with a ring around it that, as Girnus noted, bears an unfortunate resemblance to the logo of Saturn, the General Motors brand that went bankrupt in 2010. “The logo was free and our design budget went to the domain, so here we are, orbiting a dead brand at $70 million per revolution,” Girnus wrote.
When those 130 million viewers did what the ad told them to do and visited ai.com, the site collapsed under the traffic. The entire operation was hosted on Cloudflare’s basic tier — infrastructure designed, as Girnus put it, “for food blogs and wedding photographers, not for absorbing the combined curiosity of a nation told to visit a two-letter domain during the biggest television event on earth.” Marszalek responded on X with a message that read, “Prepared for scale, but not for THIS,” accompanied by three fire emojis. Girnus’s assessment was drier: “The fire emojis were load-bearing. They were doing more work than our infrastructure.”
The Product That Doesn’t Exist Yet — and Nobody Noticed
What ai.com actually offers, once the servers recover, is an invitation to pick a username for an AI agent. The company’s marketing materials describe the product as an “autonomous AI agent” that “organizes work, sends messages, and executes actions across apps.” Which actions? Which apps? At what cost? These are what Girnus calls “implementation details,” and implementation details, he says, are beneath a “vision company.” The vision cost $70 million. The implementation cost $500. “The gap between the two,” he wrote, “is where shareholder value lives.”
The site offers a free tier and a paid tier at $20 per month. The free tier provides access to a product that does not exist. The paid tier provides access to the same nonexistent product but with more input tokens. “No one has asked ‘input tokens for what,’” Girnus observed. “This is the kind of question that delays launches.” The marketing claims users can create an AI agent in 60 seconds — type a username, click “generate,” and receive a loading spinner. What you do not receive, Girnus clarified, is an AI agent. “But the experience of waiting for one is, I’m told, ‘the product.’”
The $300 Billion Loading Spinner
Girnus’s account, delivered with the deadpan precision of an engineer who has seen too much, resonated because it described not just ai.com but an entire industry operating on similar principles. The numbers across the sector are staggering. OpenAI has raised more than $40 billion in funding, according to The Wall Street Journal, while its product reportedly loses money on every user. Anthropic, the company founded by former OpenAI researchers, has raised approximately $15 billion to build what its own leadership has described as technology that could pose existential risks to humanity — and investors are competing to write larger checks. Microsoft has committed $80 billion to AI infrastructure spending in the current fiscal year, as reported by CNBC, while its flagship Copilot product has generated headlines primarily for telling users to put glue on pizza.
“The entire industry is a $300 billion screensaver with a loading spinner,” Girnus wrote. “We fit right in.” The characterization is hyperbolic, but the underlying tension is real. Across Silicon Valley and beyond, the gap between AI investment and AI revenue remains a chasm that executives prefer to describe as “early-stage market development.” Goldman Sachs analysts have questioned whether the industry’s capital expenditure will ever generate adequate returns, with a widely cited Goldman Sachs research note asking whether the technology is spending too much for too little. The question hangs over every quarterly earnings call, every funding round, and every Super Bowl commercial that directs viewers to a website that cannot stay online.
The Ghost of Pets.com and the Super Bowl Ad Graveyard
Twenty-three percent of Super Bowl ads in 2025 were purchased by AI companies — 15 out of 66 total spots, according to advertising industry tallies compiled by Adweek. The concentration of a single industry in the most expensive advertising real estate in the world inevitably invites comparison to the last time this happened. In 2000, at the peak of the dot-com bubble, fourteen internet companies bought Super Bowl ads. Among them was Pets.com, whose sock puppet mascot became the most enduring artifact of the era — outliving the company, which went bankrupt nine months after its big game debut, by decades.
“I’m not saying history repeats,” Girnus wrote. “I’m saying it rhymes, and the rhyme scheme is expensive.” The parallel is imperfect but instructive. The dot-com Super Bowl advertisers were selling products — pet food delivery, online currency, job searches — that in many cases did eventually become viable businesses, just not for the companies that burned through their venture capital to buy thirty seconds of airtime. The AI Super Bowl advertisers of 2025 are, in several cases, selling something even more abstract: the promise of intelligence itself, packaged as a subscription service, delivered through a chat interface, and valued at multiples that would make a dot-com-era venture capitalist blush.
OpenAI Sold the Name, Then Bought the Ad Next Door
Perhaps the most revealing detail in the ai.com saga is the competitive dynamics of the Super Bowl broadcast itself. OpenAI sold the ai.com domain to Crypto.com’s Marszalek, then purchased its own Super Bowl advertising slot in the same broadcast to promote ChatGPT — the product that ai.com used to redirect to. As The Verge reported, OpenAI’s ad was a sweeping, emotionally charged spot that positioned artificial intelligence as the next great leap in human progress, from the printing press to the personal computer to the chatbot. Meanwhile, a few commercial breaks away, the new owners of the domain OpenAI had just sold were directing the same audience to a crashed website offering usernames for agents that do not yet exist.
“We are now competing with the company that built the thing we may or may not be reselling,” Girnus wrote. “During the same commercial break. On the same channel. For the same audience. The AI industry is a snake eating its own tail, except the tail cost $70 million and the snake can’t stay online.” The circularity is dizzying but characteristic. In the current AI economy, companies routinely build products on top of other companies’ models, which are themselves built on data scraped from the open internet, which increasingly contains content generated by the models themselves. The supply chain is a Möbius strip, and the only thing moving in a straight line is the money — from investors to cloud providers, from cloud providers to chip manufacturers, from chip manufacturers to the sovereign wealth funds that are financing the next round of data centers.
The Crypto-to-AI Pipeline and the Art of the Pivot
Marszalek’s trajectory from crypto evangelist to AI entrepreneur is not unique. Across the technology industry, founders and executives who rode the cryptocurrency wave of 2021-2022 have been repositioning themselves as AI visionaries with remarkable speed. The vocabulary has shifted — “decentralized” has been repurposed, “tokens” now refer to units of language model input rather than units of blockchain currency, and “mining” has given way to “training” — but the underlying dynamics of speculative capital chasing narrative momentum remain largely unchanged.
Ai.com’s press materials describe a “decentralized network of billions of agents.” Girnus noted that the word “decentralized” was included because the CEO comes from crypto, where the term means “we haven’t decided how it works yet.” The company has not, he added, changed the definition. The pivot from crypto to AI is, in this framing, less a strategic evolution than a lexical one: the same ambiguity, the same grandiose promises, the same gap between the press release and the product, repackaged for a market that has moved on from blockchain to neural networks but has not updated its due diligence process.
Brand-First Development and the Inversion of Silicon Valley Orthodoxy
For decades, the mythology of Silicon Valley has been built on the garage — the idea that great technology companies begin with a product, built by engineers, in conditions of material deprivation, and that the brand follows the breakthrough. Apple started in a garage. Hewlett-Packard started in a garage. Google started in a dorm room. The narrative arc runs from invention to commercialization, from technical achievement to market dominance. Ai.com inverts this entirely. The brand came first, at a cost of $70 million. The product, if it arrives at all, will come later, built on whatever budget remains after the marketing spend — which, at last count, was approximately $500.
Girnus calls this “brand-first development.” “In every other industry,” he added, “it’s called something else.” The something else he leaves unnamed, but the implication is clear. The AI industry in 2025 has developed a remarkable tolerance for companies whose most impressive technical achievement is their domain name. The two-letter .com domain has become a status symbol, a signal of seriousness, a $70 million handshake with the market that says: we belong here, we are real, we are the future. That the website behind the domain was built in a weekend by an AI tool (which Girnus refers to as “OpenClaw, previously Moltbook, previously Clawdbot” — a joke that lands because the naming chaos of AI products has become its own punchline) and hosted on free-tier infrastructure is, in this framework, an implementation detail.
The Engineer’s Lament and What It Reveals
There is a temptation to read Girnus’s account as pure satire — a comedian’s riff on an industry that has lost its mind. But the details are too specific, too verifiable, and too consistent with the broader pattern of AI industry behavior to dismiss as fiction. The domain purchase was real. The Super Bowl ad was real. The website crash was real and witnessed by millions. The gap between the money spent on marketing and the money spent on engineering is, if anything, understated in Girnus’s telling, because it does not account for the ongoing costs of maintaining the fiction that a product exists while the product is still being imagined.
What Girnus’s account captures, with surgical precision, is the moment when an industry’s narrative outpaces its substance by such a wide margin that the narrative itself becomes the product. Users are not paying $20 a month for an AI agent. They are paying $20 a month for the feeling of having an AI agent — for the username, for the loading spinner, for the proximity to a two-letter domain that sounds like the future. This is not new in the history of technology bubbles, but the scale is unprecedented. The dot-com era’s most extravagant failures involved tens of millions of dollars. The AI era’s most extravagant experiments involve tens of billions, and the loading spinners are better designed.
What Happens When 130 Million People Call the Bluff
The crash of ai.com on Super Bowl Sunday was, as Girnus noted, “the most honest thing the AI industry has produced.” A $78 million promise that, when 130 million people showed up to collect, returned an error page and a suggestion to refresh. Every technology bubble has its defining image — the Pets.com sock puppet, the Theranos black turtleneck, the WeWork beer tap. For the AI bubble, if it is indeed a bubble, the defining image may be a loading spinner on a $70 million domain, spinning and spinning while the servers gasp for air on infrastructure designed for a food blog.
The question that Girnus’s account raises, and that the broader market has been reluctant to answer, is how long the gap between promise and product can persist before the market demands a reckoning. In the dot-com era, the answer was roughly eighteen months from the Super Bowl ads to the crash. The AI industry has deeper pockets, more patient capital, and — crucially — underlying technology that does, in many applications, actually work. ChatGPT is a real product. Claude is a real product. The enterprise AI tools being deployed across industries are generating real, if modest, returns. The danger is not that artificial intelligence is worthless; it is that the speculative froth surrounding it has created a class of companies whose primary innovation is in branding, and whose primary product is the expectation of a product.
Hiring Now: Backend Engineers, Budget Negotiable
Girnus closed his post with a hiring notice: “Backend engineers preferred. Budget: whatever’s left.” It reads as a joke, but it also reads as an invitation — to engineers who have watched the industry’s center of gravity shift from the server room to the marketing department, and who recognize that the loading spinner cannot spin forever. At some point, the website has to load. At some point, the agent has to do something. At some point, the $70 million name has to be attached to a product that justifies the price.
Until then, the AI industry has ai.com: a two-letter monument to the proposition that in a market driven by narrative, the story is the product, the domain is the moat, and the engineer is an afterthought with a $500 budget and a free Cloudflare account. It is, depending on your perspective, either the most cynical thing the technology industry has ever produced or the most honest. Perhaps it is both. Perhaps that is the point. The loading spinner keeps spinning. The fire emojis keep doing their load-bearing work. And somewhere, a backend engineer is waiting for a budget that may never come, building the future on infrastructure designed for wedding photographers, one refresh at a time.
The $70 Million Domain, the $500 Website, and the Loading Spinner That Tells the Truth About the AI Industry first appeared on Web and IT News.
