Super Micro Computer, the server maker widely known as Supermicro, has begun shipping its new H15 line of servers powered by AMD's sixth-generation EPYC 9006 processors. Alongside the launch, the company raised its revenue target for fiscal year 2027 to as much as $82 billion, citing steadily increasing demand for AI data center equipment.

The rollout came one day after AMD formally released the new chips on November 10. Codenamed Venice, the processors are based on AMD's new Zen 6 design. Supermicro confirmed that customers have already started receiving the new servers, according to its announcement. Turning a chip launch into customer deliveries within a single day underscores how much weight server makers now place on speed to market for AI data center equipment — the same demand Supermicro cited in lifting its fiscal 2027 target.

More than 20 new systems for AI

The H15 lineup spans more than 20 different systems, covering servers, storage, workstations, and GPU platforms. The machines are designed for cloud companies and businesses that need powerful computers, and they can also run agentic AI systems, which are capable of completing multiple tasks with little human assistance.

Each processor socket is capable of supporting up to 256 cores and 512 threads, helping the servers handle many tasks at the same time. The systems use AMD's SP7 and SP8 sockets and come with either one or two processors. They are available in several models, including Hyper, CloudDC, GrandTwin, FlexTwin, Petascale, and SuperBlade. That parallel-processing capacity is particularly relevant for AI workloads, which often require many requests or agent steps to run at once.

The new servers also feature faster memory and connections to help them move data more quickly. They carry 16-channel DDR5-8000 memory, which can handle roughly 1.6 terabytes of data per second for each processor socket, and PCIe 6.0 connections with up to 96 lanes for moving data between parts of the system.

Because each processor socket can draw up to 600 watts of power and generate considerable heat, Supermicro offers configurations that use either air or liquid cooling to prevent overheating.

Faster servers, thinner margins

According to Supermicro, the H15 servers provide up to 33% more processing cores than older models, move data twice as fast through PCIe connections, and handle memory data 2.6 times faster. When running large AI models, the company says the servers can produce up to 1.8 times as much AI-generated text per second compared with systems built on older AMD chips, while overall performance is roughly 1.7 times better than the previous generation.

Supermicro added that its 2U Hyper and CloudDC servers also broke several performance records in SPEC CPU tests. However, these performance figures are based entirely on Supermicro's own tests against older AMD systems, and the company did not say how many H15 servers it had shipped, who bought them, or how much money it made from them.

The launch comes as Supermicro struggles to make as much profit from its sales. In the quarter ended September 30, the company posted revenue of $19.9 billion, above the $16.3 billion analysts had expected, according to Yahoo Finance. Its gross profit margin nonetheless dropped from 17.5% to 12.7%, which Supermicro attributed to a larger share of its sales coming from large AI data center projects, which earn less profit than sales to other businesses and distributors. That trade-off explains why the path to the $82 billion target will be judged not only on revenue, but on how much profit each incremental dollar of AI data center business brings in.

More AMD-powered systems on the way

Supermicro is also developing two larger AMD-based systems, Helios and Helion. Helios is built to train large AI models, combining AMD EPYC processors, AMD Pensando networking technology, and up to 72 AMD Instinct MI455X GPUs. Helion likewise uses AMD EPYC processors but is designed to link multiple servers together to handle compute-intensive tasks. Neither system is included in the current H15 shipments; Supermicro plans to release both later together with AMD, according to Silicon Report. Their arrival will mark the next stage of the company's AMD-based lineup, moving from individual servers to integrated large-scale AI platforms.

Analysts remain divided

Supermicro shares rose more than 6% in early trading after the company launched its new servers and raised its revenue forecast, Yahoo Finance reported. But the stock was still down about 24% over the past year, trading near $29 compared with its June high of $51.40.

Wall Street analysts remain divided on the stock. On November 10, Susquehanna raised its price target for Supermicro shares from $15 to $18 but kept its Negative rating because it still considered the stock too expensive, according to BigGo Finance. On November 4, Rosenblatt kept its Buy rating and $57 price target, while BofA Securities kept its Underperform rating and a $39 target. Overall, analysts have a Hold rating on Supermicro shares, with an average price target of about $36.64. The spread between the lowest and highest targets — $18 to $57 — captures the unresolved debate over how to weigh Supermicro's revenue growth against its margin pressure and the pending regulatory review.

The company also faces questions over reports that servers using Nvidia chips were rented to China's Moonshot AI despite US export restrictions. Regulators are examining the claims, according to a Yahoo Finance report. Supermicro has said it does not do business with the companies involved. How that review concludes, and when Helios and Helion reach customers, are among the developments that will frame Supermicro's push toward its fiscal 2027 goal.