Microsoft Exec: More Memory Chips Won't Solve AI Bottlenecks

• Microsoft's Azure hardware chief Rani Borkar stated that increasing memory chip capacity will not solve AI infrastructure bottlenecks.
• Borkar called for innovation in components to address limitations in the AI supply chain.
• AI chip production is currently a binding constraint on the pace of AI compute buildout.
• Demand for AI compute power is growing exponentially and outpacing chip manufacturers' forecasts.
• Major tech companies like Microsoft, Alphabet, Amazon, and Meta are planning significant capital expenditures for AI infrastructure.

Microsoft's Azure hardware chief Rani Borkar has stated that simply increasing memory chip production will not resolve the supply chain bottlenecks that are hindering the build-out of artificial intelligence infrastructure. Speaking at SEMICON Taiwan 2026, Borkar emphasized that innovation in components is necessary to overcome these limitations.

AI chip manufacturing has emerged as a binding constraint on the pace of AI compute buildout, with demand for computing power to train and deploy advanced AI models growing exponentially and outstripping many chip manufacturers' forecasts. Supply chains for these critical components and their inputs cannot scale rapidly enough to meet this demand, as building additional manufacturing capacity takes years.

This situation contrasts with previous years, where power for data centers was the most common constraint on AI scaling in 2024 and 2025. However, in 2026, the tightest constraint faced by AI companies in procuring additional compute is the production of the AI chips themselves. Major technology firms, including Microsoft, Alphabet, Amazon, and Meta, are planning to spend nearly $700 billion on capital expenditures in 2026, the majority of which is allocated to AI infrastructure.