Amazon Closes Nick Saraev’s Account After Nvidia DGX Spark Orders

Nick Saraev, who teaches AI automation to a large online audience, said he placed orders for Nvidia DGX Spark desktop AI computers on Amazon on October 31, 2026. The orders were cancelled. Shortly after, Amazon closed his account.

On November 1 he posted on X: “Amazon just closed my account.” He added that “buying even modest amounts of compute immediately puts you on a watchlist.” The post spread quickly among AI builders, and by November 3 “Amazon closes Nick Saraev account” was a rising Google search in the US.

Amazon has not said why. According to the reports, the closure email gave no specific reason but mentioned “previous attempts to verify your account.” TweakTown and Wccftech both said Amazon had not responded to requests for comment.

What is known is the general picture:

• Amazon’s Conditions of Use let it refuse service, close accounts or cancel orders at its sole discretion.
• Scarce, high-value hardware attracts resellers and fraud, so retailers watch for several units bought quickly by one account.
• Automated systems make many of these decisions first, and genuine buyers can be caught by rules meant for scalpers.

Without Amazon’s side, the exact trigger remains unconfirmed. The case struck a nerve because many small AI teams buy hardware the same way.

The DGX Spark is Nvidia’s desktop-sized AI computer. It is built around the GB10 Grace Blackwell chip with 128 GB of unified memory, enough to run and fine-tune large AI models locally instead of renting cloud GPUs. Nvidia launched it in October 2025, and partners including ASUS, Dell, HP, Lenovo, MSI, Acer and Gigabyte sell their own GB10 versions.

For a solo developer or small team, a few of these boxes can replace a recurring cloud bill. That is exactly why demand is strong and supply tight, and why buying several at once can look unusual to a retailer.

The Tim Sweeney case shows it is not only about large orders: the Epic Games CEO said his order for two Nvidia GPUs was also cancelled.

If you need more than one or two AI machines, a consumer shopping account is the wrong tool. Practical options:

• Use a business account. Amazon Business, Dell, HP, Lenovo and other vendors have business channels built for multi-unit orders, with invoicing and verified company details.
• Buy from Nvidia’s partners or authorised resellers. They can quote multiple units, confirm supply and avoid the consumer fraud filters.
• Ask for a quote first. For anything beyond a handful of units, a quote locks in price and delivery and creates a paper trail.
• Keep your main account separate. Do not risk the account you use for everything else on a hardware purchase that might be flagged.

Once a team outgrows desktop boxes, the next step is rack servers, and that brings power, cooling and space into the budget.

Saraev’s post captured a real shift: AI developers want their own compute rather than renting it. Seven desktop units are a workstation cluster. Seventy GPUs is a room. Beyond that you are planning a small data center, with an electrical service, cooling and a power contract.

The economics change at each step. Cloud rental is easy to start and expensive to keep. Owned hardware is cheaper over time but needs somewhere to run, and in Canada and the northern US cheap hydro power and cool air make that much more affordable.

The AI Build Configurator shows what an owned AI facility would cost to build and run for any Canadian province or US state, including power cost and payback. Try the free demo before you commit to the next hardware order.