Amazon Closed an AI Creator's Account Over Seven Nvidia DGX Sparks

Amazon closed AI creator Nick Saraev's account after he tried to buy seven Nvidia DGX Spark mini-PCs, roughly $59,500 to $66,500 at current street prices. The closure email cited a verification process. It gave no reason.

Saraev, who says he helps entrepreneurs implement AI and claims 500,000 YouTube followers and 700,000 on Instagram, posted on X that Amazon simply shut him down mid-purchase. He was, in his words, "4 orders deep" when Amazon stopped allowing him to buy DGX Sparks at all.

"Amazon just closed my account," Saraev wrote. "Buying even modest amounts of compute immediately puts you on a watchlist." The claim is his. Amazon has not confirmed it.

What do the screenshots show?

The screenshots Saraev shared indicate seven canceled orders in total. Each carries the same message: "Your order was canceled. You have not been charged for this order." He does not state a unit count himself.

Every order targeted the same listing: a DGX Spark with 128GB of memory and 4TB of storage. The 128GB DGX Spark carries an MSRP of $6,950. Two Amazon listings showed $8,499.99 and $9,499.99 when checked today — 22% to 37% above MSRP. Both listings offered Prime delivery the next day. The hardware remains on sale; the buyer did not.

Amazon's support account responded to Saraev's earlier post about three hours before his closure post, offering a review. The final email said Amazon had "made the final decision" to close the account. It referenced "previous attempts to verify your account" but offered no specific reason. Saraev says he completed the verification process. Amazon's Conditions of Use reserve the right to "refuse service, terminate accounts" or "cancel orders in its sole discretion" — language that makes the decision effectively unchallengeable.

Is Saraev alone?

Epic Games CEO Tim Sweeney reports a different but similar pattern. "Amazon didn't cancel my account but continued accepting all my regular orders and canceling them just prior to arrival dates, after I tried to buy two NVIDIA GPUs," Sweeney wrote. "Rather bizarre." Two GPUs is a modest order. It still triggered cancellations, according to his account.

Saraev replied that everyone would "appreciate a bit more transparency." None has come.

In September, a user named Dracones89 on the Nvidia developer forum reported "suspicious activity locks" on an Amazon account while trying to buy two Asus GX10s. That buyer ended up purchasing two DGX Sparks from Micro Center instead. Another X user, @PurzBeats, wrote: "They canceled the order when I tried to buy my 5090, twice."

This marks a shift from earlier shortage eras, such as the Mac mini rush during the OpenClaw days, when buyers faced long waits rather than canceled orders or closed accounts.

Why might Amazon act this way?

Export control pressure offers one partial explanation. Micro Center required one buyer to sign an Advanced Computing Product Purchaser Declaration — with photo ID, end user, and install location — just for a workstation with an RTX 5090. The form forbade exporting or reselling the hardware. Amazon's conditions likewise require customers to "comply with all U.S. or other export and re-export restrictions." China and the United States remain locked in a dispute over AI hardware exports.

Commercial risk compounds the regulatory one. Hardware prices have climbed on rising memory costs, and local AI inference has grown with open-weight models. High ticket prices plus uncertainty about model safety — open-source AI agent tools have reportedly appeared in multiple recent cybersecurity incidents — may make retailers uneasy about bulk buyers.

Amazon has not explained itself, and the pattern may be narrower than it appears: seven DGX Sparks is hardly a typical purchase. Until Amazon responds in detail, buyers of significant AI hardware face a practical warning — traditional online retailers may cancel orders without explanation, and brick-and-mortar chains like Micro Center may tighten declarations further to block illegal export workarounds.