Interview, Fireside Chat, Conference Presentation, Panel
David Bitton, David Dubrovsky, Andrew Dieckman & Rachel Metz | Fireside Chat with 5C & AMD
- The company aims to deliver large-scale industrial infrastructure at a pace matching technological evolution, planning data centers to operate for decades with multiple hardware generations while future-proofing for current efficiency needs.
- A "shovel to token" strategy focuses on harmonizing compute, power, cooling, networking, storage, chips, and servers to optimize output, supported by long-lead project timelines and early engineering collaboration on specific hardware like the MI355X.
- Site selection strategies integrate power availability, political stability, cost basis, network connectivity, and community acceptance, recognizing that end-user preferences are dictated by local laws and regulations.
- To mitigate risks related to power entitlement and community disruption, the company intends to engage immediately with local city councils and populations to address concerns regarding noise, water consumption, and electricity, ensuring data centers coexist sustainably with their surroundings.
- The organization plans to align its roadmap with the top four or five customers who control approximately 50% of global consumption, acknowledging market consolidation and the necessity of partnering with major model builders like OpenAI for multi-generational planning.
- Internal development is accelerating through the deployment of AI agents for software, chip work, and validation, allowing kernel optimization engineers to achieve 100 to 1,000 times greater effectiveness and enabling the management of multi-vendor coexistence via a full internal AI agentic system.
- Power density is expected to increase yearly, driving predictions of massive gigawatt-scale liquid-cooled data centers alongside decentralized AI deployment in laptops, robotics, and homes, though current grid capacity is estimated to support only half of global AI demand.
- The company anticipates increased reliance on strategic ecosystem partnerships over the next five to ten years, noting that software and hardware teams are accelerating solution delivery through recent AI tools while managing the constant complexity of product market entry.