A company that wants a model to answer questions about its own files still has to decide where those files live. A privacy review or a compliance exam is easier to complete when the server sits in a room the company already operates. The delay has been the graphics server. Teams have been waiting months for a custom order, and staff have kept using a public chatbot while that order was open.
Lenovo announced a shorter path on September 30. AI Express is a buying menu for the Lenovo Hybrid AI Factory with NVIDIA. Ashley Gorakhpurwalla, president of the Infrastructure Solutions Group, said customers need a configuration that fits current work and can grow, with less risk of buying too much or waiting through a long delay.
Lenovo is applying the method it already uses for PCs and ordinary servers. It limits the menu, tests those combinations in its lab, and uses its supply chain to plan the parts. The company says that supply chain is what makes the shorter ship time possible. I have watched this group add services around the servers it sells, and this release continues that work. Software and support travel with the hardware, so a buyer can start the workload on company data from the same order.
Three sizes, with a ship time on each
Lenovo says an eligible order can ship in 15 business days. The three configurations are listed from 15 days, from 20 days, and from 25 days. The product page calls the range a 15 to 25 business day goal. The clock starts after order validation, payment or credit clearance, end-user certification, and any export approval. It covers selected parts in markets that are open to these shipments. User counts come from Lenovo's sizing tool and vary by model, workload, and software.
- Small, from 15 days. ThinkSystem SR650a V4 with two NVIDIA RTX 6000 PRO Blackwell Server Edition graphics processors. Lenovo sizes it for tens of users, more than 30 tokens per second, and models of 7 billion to 70 billion parameters. A team can place this server in a rack it already has.
- Medium, from 20 days. ThinkSystem SR675 V3 with eight of those graphics processors. Lenovo sizes it for hundreds of users and models of 70 billion to 400 billion parameters, including agent work.
- Large, from 25 days. ThinkSystem SR680a V4 with NVIDIA HGX B300. Lenovo sizes it for thousands of users and models up to 1 trillion parameters. Ask facilities about power and cooling before this order.
Each configuration can use an AMD or Intel processor. A buyer can add Red Hat AI Factory with NVIDIA or NVIDIA AI Enterprise, and can add Veeam Kasten to protect the applications, data, models, and pipelines. Services cover a use-case review, a return check, a proof of concept, and tuning of the graphics processors. Lenovo says Premier Support Plus for Servers identifies problems earlier. Channel partners sell the same menu through Lenovo 360.
Lenovo says its 2026 CIO Playbook found that organizations moving from a trial to wider use expect $2.79 back for every $1, and that 93 percent of enterprise respondents expect a positive return. The ship times are how Lenovo wants a company to move from that expectation to a server in its own room.
The first server goes in the room you already run
The small configuration is a rack server. The medium configuration is a denser graphics server for the same hall. The large configuration is sized for the heaviest of the three workloads. On September 16 Lenovo announced servers you can put under the applications you already run. That post is Lenovo Adds New Servers Without Moving Your Business Apps. AI Express is the next offer from the same group, for model work on the files those applications already hold.
Dell, HPE, and Supermicro sell factory systems too
- Dell sells the Dell AI Factory with NVIDIA. A November 17, 2025 Dell post says the PowerEdge XE7740 and XE7745 use NVIDIA RTX PRO 6000 Blackwell Server Edition graphics processors. Dell also sells a deskside system.
- Hewlett Packard Enterprise sells HPE Private Cloud AI. The offer includes an air-gapped configuration, NVIDIA RTX PRO 6000 Blackwell Server Edition graphics processors, and expansion racks up to 128 graphics processors. StorageReview reported those racks for July 2026.
- Supermicro said on September 23 that it is shipping NVIDIA Vera Rubin NVL72 racks with liquid cooling. One unit covers 16 compute racks and 1,152 graphics processors, for a company building a larger hall.
Ask two questions before you order
Works Cited
Hartman, Robert. "Scale and Automate AI Success with Dell and NVIDIA." Dell Technologies, 17 Nov. 2025, www.dell.com/en-us/blog/scale-and-automate-ai-success-with-dell-and-nvidia/.
Hewlett Packard Enterprise. "HPE Private Cloud AI." Hewlett Packard Enterprise, www.hpe.com/us/en/private-cloud-ai.html. Accessed 4 Oct. 2026.
Lenovo. "Lenovo AI Express." Lenovo, www.lenovo.com/us/en/ai-express/. Accessed 4 Oct. 2026.
Lenovo. "Lenovo AI Express with NVIDIA Accelerates Path to Hybrid AI Factory in Weeks." Lenovo StoryHub, 30 Sept. 2026, news.lenovo.com/pressroom/press-releases/lenovo-ai-express-with-nvidia-accelerates-path-to-hybrid-ai-factory-in-weeks/.
NVIDIA. "Validated Design for AI Factories." NVIDIA, www.nvidia.com/en-us/solutions/ai-factories/validated-design/. Accessed 4 Oct. 2026.
StorageReview. "HPE Expands NVIDIA AI Computing Portfolio with Scalable Private Cloud AI and Blackwell GPU Integration." StorageReview, 18 Mar. 2026, www.storagereview.com/news/hpe-expands-nvidia-ai-computing-portfolio-with-scalable-private-cloud-ai-and-blackwell-gpu-integration.
Super Micro Computer. "Supermicro Now Shipping NVIDIA Vera Rubin NVL72 Racks." Super Micro Computer Investor Relations, 23 Sept. 2026, ir.supermicro.com/news/news-details/2026/Supermicro-Now-Shipping-NVIDIA-Vera-Rubin-NVL72-Racks/default.aspx.
