Supermicro at AMD Advancing AI 2026: H15 Servers, $60B Orders, and a Gigawatt SpaceX Deal

Supermicro also showed the densest EPYC 9006 rack at the event: 96 EPYC 9006 CPUs in a 42U rack using FlexTwin, and a separate 42U rack with 168 CPUs (43,008 "Zen 6" cores) . The portfolio is optimized for agentic AI, enterprise, HPC, cloud, and storage workloads, designed to work alongside AMD Instinct GPUs and AMD Pensando networking .

Alongside the H15 launch, Supermicro showcased the Supermicro AMD Helios Platform, a 72-GPU liquid-cooled rack-scale system with AMD Instinct MI455X GPUs and Venice CPUs, priced around $5.25M per rack .

The $60B Q4 Preliminary Update

Two days before the summit, on July 21, 2026, Supermicro released a preliminary Q4 fiscal 2026 business update that reset market expectations on profitability .

Analysts at Mizuho cautioned that fulfilling the $60B backlog may require significant debt financing or capital raises given Supermicro's $1.3 billion cash position, but the margin expansion was widely seen as a structural inflection toward higher profitability .

SpaceX Gigawatt-Scale AI Data Center

Supermicro CEO Charles Liang confirmed that the company will help SpaceXAI (Elon Musk's merged SpaceX and xAI) build a new gigawatt-scale AI data center, with a target operational date in 2027 . Liang stated on X that it would be "the fastest build they have ever done" . The data center is planned for Texas, at a scale comparable to or exceeding SpaceXAI's existing Memphis facility .

Supermicro's role signals its turnkey integration capability for hyperscale AI builds, extending beyond chip-vendor partnerships to full data center deployment . This partnership was cited by multiple outlets as a contributing factor to the $60B order surge, along with broader AI infrastructure demand from hyperscalers .

Competitive Picture vs. Intel Xeon

Supermicro's H15 portfolio directly competes with Intel's Xeon 6+ (Granite Rapids) based platforms, which Supermicro also supports in its X15 lineup . The AMD-based H15 portfolio offers several structural advantages:

• First-mover on 2nm: AMD's EPYC Venice is the first x86 server chip on TSMC 2nm, giving Supermicro's H15 systems a power-efficiency and density advantage over Intel's Xeon 6+ (built on Intel 3) .
• 256 cores per socket vs. Intel's current maximum of ~128 cores per socket (Xeon 6900P series), offering 2x core density for virtualized and AI inference workloads .
• 1.7x generational uplift from EPYC Turin to Venice, while Intel's Xeon 6+ generational gains are more modest (~1.2–1.4x depending on workload) .
• AMD Instinct GPU integration gives Supermicro a unified AMD ecosystem story (Venice CPU + MI455X GPU + Pensando networking + ROCm) for large-scale AI training .
• Open standards: AMD's support for UALink and OCP DC-MHS specifications gives Supermicro's H15 customers an alternative to Nvidia's proprietary NVLink and Intel's CXL ecosystem .

That said, Intel still holds an installed-base advantage in enterprise and legacy HPC, and Supermicro continues to serve both camps — its X15 (Intel Xeon 6+) and N15 (Nvidia CPU-GPU) platforms remain active product lines .

The Big Picture

The three announcements form a coherent growth narrative: the H15 portfolio extends Supermicro's product leadership with AMD's 2nm Venice chips; the $60B order backlog validates surging AI infrastructure demand; and the SpaceX gigawatt data center partnership signals Supermicro's ability to capture hyperscale turnkey deployments. Together, they drove SMCI's sharpest rally in months and reinforced Supermicro's position as a leading multi-vendor AI server integrator.