Summary: Super Micro Computer (NASDAQ: SMCI) said on September 23, 2026, that it is now shipping Nvidia Vera Rubin NVL72 racks with its own direct liquid cooling and deployment services, and a SupermiSummary: Super Micro Computer (NASDAQ: SMCI) said on September 23, 2026, that it is now shipping Nvidia Vera Rubin NVL72 racks with its own direct liquid cooling and deployment services, and a Supermi

Supermicro Begins Shipping Nvidia Vera Rubin NVL72 Racks: What It Means for AI Infrastructure

Summary: Super Micro Computer (NASDAQ: SMCI) said on September 23, 2026, that it is now shipping Nvidia Vera Rubin NVL72 racks with its own direct liquid cooling and deployment services, and a Supermicro post on X said the first systems went to SpaceXAI. Each rack holds 72 Rubin GPUs and 36 Vera CPUs. Dell shipped production Vera Rubin racks first, and SMCI stock closed roughly flat on the news.

Key Takeaways

  • Supermicro announced on September 23, 2026, that it is shipping Vera Rubin NVL72 racks integrated with its DCBBS building blocks and DLC-2 liquid cooling.
  • The first systems went to SpaceXAI from Supermicro's expanded San Jose manufacturing, according to a September 21 company post on X.
  • One NVL72 rack carries 72 Rubin GPUs, 36 Vera CPUs and 20.7 TB of HBM4. One Scalable Unit links 16 racks with 1,152 GPUs and 331 TB of HBM4.
  • Dell delivered the first production Vera Rubin NVL72 racks to CoreWeave in early September, so Supermicro's news confirms a multi-vendor ramp rather than a first.
  • SMCI closed about 0.2% lower on September 23. Supermicro disclosed no volumes, prices or delivery schedules.
Supermicro said on Wednesday, September 23, 2026, that Nvidia's next-generation Vera Rubin NVL72 racks are now leaving its factories. The news is less a scoop than a checkpoint.
Vera Rubin has moved from Nvidia's production lines onto the loading docks of several server makers. What Supermicro adds is a named first customer and a full liquid-cooling package around the rack.

What Did Supermicro Announce About Nvidia Vera Rubin?

Supermicro announced on September 23, 2026, that it is shipping Nvidia Vera Rubin NVL72 racks integrated with its Data Center Building Block Solutions (DCBBS) and DLC-2 direct liquid cooling stack. Customers can order a complete Scalable Unit and receive production-ready systems, with one Supermicro team covering site survey, design, integration, testing, delivery and deployment.
Supermicro says every rack is tested and validated with the full liquid-cooling stack before it ships. Chief executive Charles Liang said the company spent years building its liquid-cooling stack, manufacturing capacity and deployment teams "for exactly this moment."

Who Received the First Supermicro Vera Rubin Racks?

SpaceXAI received the first systems, according to a Supermicro post on X dated September 21, 2026. Supermicro said those racks shipped from its expanded U.S. manufacturing in San Jose, California. The September 23 press release itself named no customers and gave no volumes, prices or delivery schedules.
SpaceXAI runs the data centers where Grok's AI agents operate, and SpaceX has tied its AI plans closely to Nvidia. On SpaceX's second-quarter earnings call in August, Elon Musk said the company had decided to build exclusively on Nvidia, and he later set a fourth-quarter 2027 target for launching a space-optimized Vera Rubin NVL72 into orbit.
For readers following SpaceX (NASDAQ: SPCX), the shipment is one more data point on the company's AI build-out.

What Is Inside an Nvidia Vera Rubin NVL72 Rack?

An Nvidia Vera Rubin NVL72 rack combines 72 Rubin GPUs and 36 Vera CPUs in one liquid-cooled system that operates as a single machine. Eighteen compute trays connect through nine sixth-generation NVLink switch trays for 216 TB/s of scale-up bandwidth, backed by 20.7 TB of HBM4 and up to 54 TB of LPDDR5X memory.
This is the configuration Supermicro is shipping, based on its September 23 announcement:
  • GPUs and CPUs: 72 Nvidia Rubin GPUs and 36 Nvidia Vera CPUs per rack.
  • Compute trays: 18 1U trays, each with four Rubin GPUs and two Vera CPUs.
  • Interconnect: nine sixth-generation NVLink switch trays delivering 216 TB/s of scale-up bandwidth.
  • Memory: 20.7 TB of HBM4 and up to 54 TB of LPDDR5X per rack.
  • Scalable Unit: 16 compute racks with 1,152 Rubin GPUs and 331 TB of HBM4.
  • Power envelope: DCBBS blueprints scale from 5 MW to gigawatt-class sites.
Nvidia's own performance claims go further. The company projects up to 10 times more tokens per megawatt than the GB200 NVL72 and one-tenth the cost per million tokens, but those are Nvidia's projections rather than measured customer results.
Memory is the headline spec, which is one reason AI memory demand has become a market story of its own, as Crypto Pulse covered in SanDisk's price-driven earnings jump.

Is Supermicro the First to Ship Nvidia Vera Rubin?

No. Dell said in early September 2026 that it had delivered the world's first production Vera Rubin NVL72 racks to CoreWeave. The NVL72 is Nvidia's rack design, and several server makers build it. Supermicro's announcement confirms that the Vera Rubin ramp now spans multiple vendors, with Supermicro competing on liquid cooling and deployment rather than on being first.
The timing matches Nvidia's schedule. On May 31, 2026, Nvidia said Vera Rubin was ramping into full production, with production shipments set to begin in the fall. Dell, HPE, Lenovo and Supermicro were among the system makers named in that announcement.
Where Supermicro tries to stand apart is the equipment around the rack. Its package runs from cold plates, manifolds, hose kits and rack power shelves to in-row coolant distribution units rated at 1.8 MW each with N+1 redundancy, in-rack and liquid-to-air sidecar units, rear-door heat exchangers and facility cooling towers.

Why Do Power and Liquid Cooling Matter for Vera Rubin?

Power and liquid cooling decide how quickly Vera Rubin capacity can come online. An NVL72 rack is liquid-cooled from the chip outward, and Supermicro's deployment blueprints start at a 5 MW power envelope. For many buyers, the practical constraint is the building, including power delivery, coolant loops and heat rejection, more than the GPUs themselves.
The mechanics are simple to follow. Cold plates sit on the chips and carry heat into a liquid loop. A coolant distribution unit (CDU) transfers that heat from the rack loop into the facility's water system, and cooling towers or heat exchangers reject it outside.
N+1 redundancy means a site runs one more CDU than the load requires. A single unit can fail without forcing racks offline, which matters when one Scalable Unit holds 1,152 GPUs.
Traders have been watching this layer too. In MEXC's June 2026 TradFi data, seven of the top ten stock futures by trading volume were AI infrastructure names spanning GPUs, memory and data-center networking.

What Does Vera Rubin Mean for SMCI Stock?

The shipment news did little for SMCI stock, which closed about 0.2% lower near $41.44 on September 23, 2026. Without disclosed volumes, prices or margins, investors had little new to model. The stock also carries overhangs, including a March 2026 export-control case involving a co-founder, dilution concerns and an unfinished board review.
On March 20, 2026, U.S. prosecutors charged co-founder Yih-Shyan "Wally" Liaw, a sales manager and a contractor with allegedly diverting about $2.5 billion of servers containing Nvidia chips to China. SMCI fell about 33% that day. Supermicro was not named as a defendant; it placed the two employees on leave and ended its relationship with the contractor.
The August earnings report raised new questions. Fiscal fourth-quarter revenue of about $11.12 billion landed near the low end of management's forecast, and a roughly $7 billion equity and equity-linked financing announced in June raised dilution concerns. An independent board review tied to the alleged export-control issues was still under way as of late August.
SMCI's 52-week range runs from $19.48 to $58.78. The next catalysts to watch are disclosed Vera Rubin orders or backlog, the revenue contribution from NVL72 racks and gross margins on liquid-cooled systems.

What Does It Mean for Nvidia and the AI Infrastructure Trade?

For Nvidia, several vendors shipping NVL72 racks shows Vera Rubin moving from launch to delivery on the schedule it set in May. For the wider AI infrastructure trade, attention shifts to the parts of the stack that set the pace of deployment: HBM4 memory supply, liquid-cooling equipment and data-center power.
Eligible MEXC users can follow that trade through MEXC Stock Futures, which list NVDA among their U.S. stock contracts, or hold actual shares through RealStocks, which MEXC says covers more than 7,000 U.S. stocks and ETFs. Availability depends on region.

Frequently Asked Questions About Supermicro and Vera Rubin

Is Supermicro shipping Nvidia Vera Rubin?

Yes. Supermicro announced on September 23, 2026, that it is shipping Vera Rubin NVL72 racks with its DCBBS building blocks and DLC-2 liquid cooling, after first saying so on X on September 21.

Who received the first Supermicro Vera Rubin NVL72 racks?

SpaceXAI, according to Supermicro's September 21 post on X. The systems shipped from Supermicro's expanded manufacturing in San Jose, California.

How many GPUs are in a Vera Rubin NVL72 rack?

Each NVL72 rack contains 72 Nvidia Rubin GPUs and 36 Nvidia Vera CPUs across 18 compute trays. A Supermicro Scalable Unit of 16 racks holds 1,152 GPUs.

How much HBM4 memory does a Vera Rubin NVL72 rack have?

Each rack has 20.7 TB of HBM4 plus up to 54 TB of LPDDR5X. A 16-rack Scalable Unit carries 331 TB of HBM4.

Who shipped the first Vera Rubin NVL72 racks?

Dell said in early September 2026 that it had delivered the world's first production Vera Rubin NVL72 racks to CoreWeave.

Why didn't SMCI stock rise on the Vera Rubin news?

Supermicro disclosed no shipment volumes, prices or margins, and the stock still carries legal, dilution and governance overhangs. SMCI closed about 0.2% lower on September 23, 2026.

What is Supermicro DCBBS?

Data Center Building Block Solutions (DCBBS) is Supermicro's package of computing, liquid cooling, networking, cabling and deployment services, sold as a defined bill of materials for AI data centers from 5 MW to gigawatt scale.

Can I trade SMCI or NVDA on MEXC?

MEXC Stock Futures list NVDA among their U.S. stock contracts, and MEXC says RealStocks covers more than 7,000 U.S.-listed stocks and ETFs for eligible users. Search SMCI in the MEXC app to confirm availability in your region.
Market Opportunity
Gensyn Logo
Gensyn Price(AI)
$0.02012
$0.02012$0.02012
USD

The articles shared on this page are sourced from public platforms and are provided for reference only. They do not represent the position or views of MEXC. All rights belong to James Mitchell. If you believe any content infringes upon the rights of a third party, please contact service@support.mexc.com for prompt removal. MEXC does not guarantee the accuracy, completeness, or timeliness of any content and is not responsible for any actions taken based on the information provided. The content does not constitute financial, legal, or other professional advice, nor should it be interpreted as a recommendation or endorsement by MEXC. For expert insights and in-depth analysis, visit MEXC Learn.