🧠 Food for thought

Implications, context, and why it matters.

AMD seeks more chips as large deals support a multi-billion-dollar target

• AMD’s AI chip demand is linked to reported partnerships with OpenAI, the maker of ChatGPT, and Oracle, the enterprise software and cloud computing company. Those deals could bring in more than US$100 billion over time 1.
• AMD is aiming for tens of billions of dollars in AI data center revenue by 2027. It is also pursuing 80% compound annual growth in data center AI revenue over the next three to five years 1.
• That plan leans on AMD’s MI300 series chips and an annual product roadmap built to challenge Nvidia’s H100 chips, which lead this market 2.
• Nvidia’s supply limits have opened a window for AMD to win orders from customers that cannot get enough Nvidia chips 2.

Advanced packaging stands in the way of AMD’s AI plans

• Raising output is harder than it looks because the industry lacks enough advanced chip packaging capacity, especially TSMC’s CoWoS, a method for combining multiple chips in one package to improve performance 3.
• Demand for AI chips is running more than 30% above this packaging capacity. The gap is expected to last through at least 2026 because the needed equipment takes time to order and install 3.
• AMD is fighting for the same packaging supply as Nvidia and cloud computing providers building custom AI chips 3.
• Access to packaging capacity now plays a large role in market share and in how competition in AI hardware develops 3.