David Green, head of technology for Asia and Japan at AWS (Amazon Web Service), said, "The number of artificial intelligence (AI) robots is expected to increase to 20 million units by 2030," and noted, "To fully benefit from the resulting expansion in efficiency, we need preemptive infrastructure investment, continuous data training, and stable facility management."
Green stated accordingly in a lecture on "Global industry and technology trends in physical AI" at the "2025 Advanced Industry Standards Leadership Forum" held at Hanyang Institute of Technology at Hanyang University in Seoul on the 12th.
Green said, "As industries recognize the role of automation for improving productivity and ensuring business continuity, the use of robots is accelerating," and predicted, "The global AI agent market is expected to surge from $5 billion last year to more than $200 billion by 2032."
Green emphasized the efficiency improvements that the adoption of AI robots will have on industry. He said, "AI robots figure out for themselves what they need to do to perform their assigned tasks," and added, "They will be able to operate autonomously for longer periods in a human-like way."
Green introduced the results Amazon has achieved by adopting robots. He said, "Since 2012, Amazon has deployed more than 750,000 robots across its global network, and in July 2025 set a historic milestone by completing the deployment of its 1 millionth robot."
He continued, "More than 500,000 robots operate daily across the fulfillment center network, supporting about 75% of orders worldwide," and explained, "As a result of optimizing paths and collaboration with human workers through Machine Learning, the number of delivery packages per employee improved 20-fold, from 175 in 2015 to 3,870 last year."
Green advised that to fully reap this expansion in efficiency, infrastructure must be well built. He said, "Without spending years on building infrastructure, we must build a complete system from day one (of AI robot adoption)," and added, "We must continuously train the data and stably manage facilities in 'edge locations,' where remote and offline consolidation can be lost."