The 2026 OCP Global Summit opens on Monday, October 12, 2026 at the San Jose McEnery Convention Center and runs through October 15, according to the Open Compute Project Foundation’s event page. The opening day carries two keynotes from AI chipmakers: AMD chair and CEO Lisa Su at 4:15 p.m. PDT, and NVIDIA’s Ian Buck later in the evening.
The Open Compute Project develops open hardware designs for data centers, from servers and racks to power, cooling and networking. AMD describes the four-day summit as a chance to “explore advances in open infrastructure for artificial intelligence and data centers,” which is why AI chip vendors use it to talk about full racks and power systems rather than single accelerators.
On this page
The opening-day keynotes
| Speaker | Company | Time (PDT, Oct 12) | Topic, as the company describes it |
|---|---|---|---|
| Lisa Su, chair and CEO | AMD | 4:15 p.m. | “Open, full-stack infrastructure for agentic AI” from enterprise systems to gigawatt-scale deployments |
| Ian Buck, VP of Hyperscale and HPC Computing | NVIDIA | 5:45 p.m. | “AI Factories at Scale: Advancing Open Infrastructure for Agentic AI” |
AMD. In a media alert published September 29, AMD said Su “will highlight the portfolio of open, full-stack infrastructure for agentic AI that AMD delivers at every scale – from enterprise systems to gigawatt-scale deployments,” and that AMD and its partners will show platforms, demos and customer use cases throughout the event. AMD has not said in advance which products the keynote will cover. Its current data-center GPU line-up is covered in our AMD Instinct MI355X explainer and the MI400 Series launch report.
NVIDIA. NVIDIA’s OCP event page lists Buck’s keynote for Monday, October 12 at 5:45 p.m. in the South Hall, followed by NVIDIA sessions from October 13 to 15. Several deal with power: “800 VDC AI Factories: From Architecture Alignment to Industry Deployment” (October 13), “Designing Safe 800 VDC Systems” and “Grid Interconnection Requirements for Large Loads” (October 14). Others cover liquid-cooled racks, GPU management standards and the MGX power rack reference design. The page also features an AI Factory MGX showcase; NVIDIA says more than 200 ecosystem partners have adopted MGX designs. Background on the platform NVIDIA is promoting is in our Vera Rubin explainer.
The program
The foundation says the summit has 22 technical tracks. The AI-related ones include AI Clusters (scale-up, scale-out and scale-across networking, 800G/1.6T interconnects, liquid cooling), AI Open Data Center: Power Distribution (low-voltage DC architectures up to 1500VDC/800VDC), Rack & Power (800V DC, 1 MW+ racks, Open Rack Wide), Networking (ESUN 1.0, UALink-SAI, 102.4T liquid-cooled switching), Photonics (co-packaged and near-packaged optics) and Open Chiplet Economy (custom AI silicon, HBM integration).
Registration is closed: the foundation says the 2026 summit is sold out, onsite registration will not be offered and it does not plan to reopen registration. It also says all sessions will be recorded, with video and slides posted on the OCP website in the days after the show.
Next year: San Francisco
The summit is leaving San Jose. In a September 1 announcement, the foundation said the 2027 OCP Global Summit will run October 4–7, 2027 at the Moscone Center in San Francisco, with later editions there on September 25–28, 2028, October 15–18, 2029 and September 30–October 3, 2030. OCP CEO George Tchaparian said the larger venue “provides the scale and the flexibility we need for the future.”
We will report product announcements from the keynotes once the companies publish them. Accelerator specifications from every vendor are in the AI chips catalog, and announced AI campuses are tracked in AI data centers.
