SuperX Announced a $38.8 Million Nvidia Server Order. Its Last Reported Period Showed $2.83 Million of Revenue.

by Sofia Lindqvist · August 25, 2026

Markets Writer, The SmallCap Signal

SuperX AI Technology said Tuesday that it has secured an order worth approximately $38.8 million for Nvidia B300 GPU-based server clusters from WOODMAN Inc., a Japanese technology company, with the systems destined for a deployment in Yokohama. According to the Aug. 25 announcement, roughly 20 percent of the value was received as an advance payment by Aug. 20, the balance is due before shipment, and delivery is scheduled for mid-November 2026. The company described it as its first commercial collaboration with Woodman.

Taken on its own, that is a straightforward contract win for a company building AI infrastructure in Japan. Taken against SuperX's own reported financials, it is something stranger: a single order larger than every dollar of revenue the company has ever reported, combined.

The reporting history is worth laying out, because it is not standard. SuperX closed a fiscal year on June 30, 2025, and reported on Oct. 31, 2025 that the period produced $3.6 million of revenue, most of it from a legacy interior design business, with roughly $1 million coming from AI server and IT equipment sales in the final month. The net loss was $21.2 million and cash stood at $17.2 million at year end. The company then changed its year end, and on June 22, 2026 reported an unaudited transition period covering the six months to Dec. 31, 2025: revenue of $2.83 million, up 283.8 percent from the comparable prior period, against a net loss of $76.1 million and operating expenses of $49.57 million, driven in large part by a sharp increase in share-based compensation linked to new management hires and transformation initiatives.

In that transition period, by the company's own description, the AI segment remained in capacity-building mode with no associated revenue. The legacy interior design subsidiary was divested in May 2026. So the most recent reported window predates the business that the company's press releases are now about, it was published nearly six months after it closed, and it was unaudited.

Meanwhile the announcements have been large and frequent, and they are company-reported figures rather than anything that has passed through a filing. On Jan. 30, 2026 SuperX said it had begun production at its first global supply center, in Tsu City, Mie Prefecture, initially capable of manufacturing up to 20,000 AI servers annually, and said it had secured customer purchase orders worth an estimated $910 million in January 2026, alongside non-binding memoranda of understanding for an additional 5,000 AI servers with a potential value of up to $2.1 billion over the following twelve months. The distinction between a purchase order and an MOU is doing a great deal of work in that sentence, and only the company knows how much of either has converted.

Some of it clearly has, on the company's account. On Aug. 19 SuperX said it had shipped approximately $31 million of Pro6000-based servers to Digital Dynamic Inc., a Japanese AI infrastructure operator, across four purchase orders between January and August 2026, and that it had booked roughly $20 million in new orders with about $28 million of projects in phased production and sequential delivery. It said it expects cumulative shipments to Digital Dynamic to reach approximately $38 million by the end of August.

Add the Woodman order and the picture is of two named Japanese customers accounting for the overwhelming majority of disclosed commercial activity. That is customer concentration in its most literal form. It is common at this stage of a hardware company's life, and it is also the reason a single deployment slipping a quarter can rewrite a year.

SuperX's own safe-harbor language acknowledges as much. In both the Aug. 19 and Aug. 25 releases the company cautions that actual delivery schedules and the value of AI servers delivered may vary based on customer data center readiness and supply chain conditions. Server clusters are recognized when they ship and are accepted, and data center readiness is not something the assembler controls.

The financing side carries its own dilution question. SuperX said in October 2025 that it had secured over $70 million in investment from long-term investors since March 2025 and had recently entered into agreements for over $170 million in additional investment from institutional investors. A company with a $17.2 million cash balance at its last reported fiscal year end, building working-capital-intensive GPU server inventory against orders that require payment before shipment, is a company that has been funding itself from the equity market rather than from operations. The share-based compensation that drove the transition-period loss adds to the same count.

The macro backdrop for this kind of story is about to be tested. Nvidia is scheduled to report second-quarter fiscal 2027 results and hold a conference call on Wednesday, Aug. 26 at 2 p.m. Pacific time, after the close, according to its investor relations event listing. Nvidia's own commentary on Blackwell-generation supply and on demand outside the largest hyperscalers is the closest thing to a read-through on whether mid-sized regional integrators like SuperX are getting allocation and at what margin.

Monday's session offered no particular support. The Nasdaq composite closed at 25,980.19, down 200.26 points or 0.8 percent, and the Russell 2000 closed at 2,995.08, down 22.79 points or 0.8 percent, according to the Associated Press tally of major index performance. Nvidia itself closed at 208.48, down 6.24 points or 2.91 percent, in the same tally, going into its own print.

None of this makes the Woodman order less real. A 20 percent advance already received is a harder fact than most order announcements contain, and mid-November is close enough to verify. But the test for SuperX is not whether it can announce orders. It is whether the next set of reported financials, whenever they arrive, shows revenue that resembles the numbers in the press releases, at a gross margin that survives the cost of the Nvidia silicon inside the box.