Whatever happened to AI.com, the site that was supposed to make agentic AI popular for the masses following a splashy Super Bowl ad? At GTC 2026, NVIDIA may have helped answer that question.

$70 Million for a Domain

The AI.com story began with a record-breaking bet. Crypto.com’s Kris Marszalek acquired the domain in April 2025 for a reported $70 million, paid entirely in cryptocurrency, the largest publicly disclosed domain sale in history. He then spent millions more on a Super Bowl ad urging viewers to rush to the site and claim their handles. As AI.com crowed in its launch press release: ‘Users simply choose their user handle, AI handle, and generate their agent immediately.’

What they actually got was a waitlist and a request for credit card details. The site crashed under traffic within minutes of the ad airing. As of late March 2026, the product still isn’t live.

The fine print made things worse. Buried in AI.com’s terms of service is language making clear that if your AI agent commits a crime, like spamming contacts, scraping paywalled content, making a trade on nonpublic information, you’re the one holding the bag. The platform builds the autonomy. The user absorbs the liability. People who signed up are left with unanswered questions.

Is the Competition Too Fierce?

Before AI.com’s Super Bowl ad even aired, the competitive landscape had already moved against it.

OpenClaw (free, open-source, and built by Austrian developer Peter Steinberger as a side project) had gone viral in late January 2026, accumulating 200,000 GitHub stars and 1.5 million agents created before Marszalek’s ad ran. Five days after AI.com’s launch-night crash, OpenAI hired Steinberger to drive their next generation of personal agents. Sam Altman described him as ‘a genius with a lot of amazing ideas about the future of very smart agents interacting with each other.’

The competitive pressure intensified in a big way at NVIDIA’s flagship GTC event. On March 16, Jensen Huang announced NemoClaw, a full security and privacy stack built specifically for OpenClaw that installs in a single command on consumer hardware, running locally on RTX PCs, DGX Station, and DGX Spark.

A major knock on OpenClaw had been that it was too complex and insecure for everyday users. NVIDIA took a significant step toward fixing that. When the most important hardware company in AI endorses your open-source competitor and calls it ‘the operating system for personal AI,’ that’s a statement about where the personal AI platform layer will live, and it is not at AI.com.

Meanwhile Anthropic is building personal agent functionality directly into Claude Code and has already moved to restrict third-party tools from using its subscriptions as a cheap backend, a sign that it intends to own the agent layer rather than enable others to build on it. Google has Gemini agents embedded across Workspace. Amazon relaunched Alexa with agentic capabilities as Alexa+.

Each of these companies is building agentic capabilities into their existing platforms, but within defined, narrow environments. AI.com promised something more sweeping: a fully autonomous agent managing your entire digital life, from email and finances to stock trading and social profiles, with no human in the loop. That is an even harder product to build. And weeks after the Super Bowl, there is still nothing to show for it.

The technology giants also something AI.com doesn’t: an existing relationship with hundreds of millions of users, and a distribution channel that requires no Super Bowl ad to reach them.

Is AI.com’s Real Product Going to Be Something Else?

There is a second question about AI.com that merits a closer look.

Marszalek’s pitch is built around a ‘decentralized network’ of agents. Decentralized is a crypto term. His entire career has followed the same playbook: acquire a premium domain, build a user base, and monetize through tokens. He bought Crypto.com for reportedly around $12 million in 2018, built it into a platform with 150 million users, and launched the CRO token into that existing base. Take the handle reservation system, the credit card collection, and the ‘agent marketplace’ roadmap together, and you have infrastructure that maps almost exactly onto what you would build if you were planning a token launch. The handles function like wallet addresses. The waitlist functions like a pre-registration for a network. No token has been announced. If a token launch is the plan, the Super Bowl ad was a user acquisition campaign.

Whether that is the plan from the start or an improvised pivot as the AI agent vision stalls, the architecture points in that direction.

Pivot or Blunder?

On March 19, three days after NVIDIA’s GTC announcement, Marszalek announced he was laying off 12% of Crypto.com’s workforce, citing AI integration as the reason. ‘Companies that do not make this pivot immediately will fail,’ he posted on X. The pivot he is demanding of others is the same one he has yet to deliver at AI.com. Marszalek is simultaneously shrinking the human operation at Crypto.com while building a user acquisition machine at AI.com. That is consistent with a platform play, not a product play. You cut costs on one side while accumulating assets (in this case, user identities and payment credentials) on the other. Whether that makes Marszalek a visionary or an opportunist may depend entirely on whether AI.com ever ships a product.

AI.com is competing with every major AI lab simultaneously absorbing the wrapper layer into their core products. The App Store analogy applies: your feature can vanish into the platform overnight.

The Broader Lesson

AI.com is competing with every major AI lab simultaneously absorbing the wrapper layer into their core products. The App Store analogy applies: your feature can vanish into the platform overnight. And AI.com is the clearest current example of a company caught in the most exposed position in the AI food chain.

In the AI industry, competitive advantage belong to three kinds of companies. The first builds the underlying models. Think OpenAI, Anthropic, and Google, among others. The second builds the hardware those models run on, with NVIDIA above all others. The third goes deep into a specific industry or business process, solving a problem so precisely that a generic platform cannot easily replace it. AI.com is none of those things. It is a consumer brand sitting in the middle of the stack, with no model, no hardware, and no specialized workflow to defend.

The wrapper layer, or the consumer-facing interface sitting between foundation models and end users, is being systematically eaten by the companies that own the models and the silicon. What looked like a distribution advantage six months ago looks like a no-man’s land today.

The pattern is familiar to anyone who lived through previous technology waves. It played out in mobile when Apple made third-party features native to iOS. It played out in search when Google absorbed entire content categories. In AI, the cycle is faster. OpenClaw went from side project to market influencer. NVIDIA announced NemoClaw, a dedicated security and privacy stack for OpenClaw, within weeks of OpenClaw going viral. AI.com went from the most audacious brand launch in recent memory to an open question about whether it has a product at all, and it happened before the Super Bowl confetti had settled.

A brand promise is worthless without performance. AI.com had the domain, the ad, and the moment. What it didn’t have was a product. And in a market moving this fast, the window for becoming that test case for everyday agentic AI may already be closing.