Seekr is set to leverage Oracle Cloud Infrastructure's (OCI) AI capabilities and AMD GPUs to develop advanced vision-language foundation models and enhance multi-node training capabilities.

Chris Gandolfo from Oracle emphasized the efficiency and cost-effectiveness of OCI's infrastructure for handling demanding AI workloads.

Rob Clark, president of Seekr, noted that the partnership with OCI is crucial for accessing the extensive GPU compute power needed for analyzing large volumes of imagery and sensor data.

Oracle plans to deploy a zettascale AI cluster featuring up to 131,072 MI355X GPUs, promising over twice the price-performance for AI training and inference compared to previous models.

These new GPUs are equipped with 288GB of high-bandwidth memory (HBM3) and can achieve memory bandwidths of up to 8TB/s, significantly enhancing training and inference speeds.

Customers can anticipate substantial performance improvements, with throughput increases of up to 2.8 times for AI deployments, leading to faster results and reduced latency.

However, the high power consumption of 1,400 watts per GPU may increase operational costs and necessitate specialized cooling systems, complicating deployment.

Crusoe Cloud, featuring AMD GPUs, is expected to launch in Fall 2025, with early access options for interested customers.

This collaboration is anticipated to generate significant revenue for Oracle, particularly in the premium cloud services market, which offers higher margins than standard compute services.

The partnership includes a joint go-to-market strategy aimed at enhancing customer growth, especially within the federal government sector, with OCI providing dedicated engineering support.

Crusoe is committed to environmentally friendly AI infrastructure solutions, focusing on clean energy and optimized data centers.

This strategic agreement positions Oracle as a formidable competitor in the AI infrastructure market, challenging major players like AWS, Azure, and Google Cloud Platform.