苏姿丰在 2025 年 12 月 4 日于 WIRED 的活动上表示，从她的角度看「绝对不是」AI 泡沫，并认为相关担忧「有些被夸大」。她主张 AI 仍需要大量晶片与资料中心算力；即使 AMD 的市值约 3530 亿美元，仍远小于 Nvidia 约 4.4 兆美元（约 12.5 倍差距），但 AMD 正在加速追赶。

她也谈到出口与成本：AMD 将为恢复向中国出货的 MI308 晶片支付川普政府设定的 15% 税。公司先前估算，美国对 MI308 的出口限制将使 AMD 付出约 8 亿美元代价；政府曾暂停对中销售，之后在夏季重新开始审查申请。

在资料中心布局上，AMD 与 OpenAI 达成多年合作：OpenAI 将在数年内部署 6 吉瓦的 AMD Instinct GPU。交易还包含以每股 0.01 美元购买 1.6 亿股的选择权，等同取得约 10% 持股；首批 1 吉瓦预计于 2026 年下半年推出。苏姿丰称她更在意的是如何更快创新，并认为 AI 仍在早期、模型会持续变好。

At a WIRED event on December 4, 2025, AMD CEO Lisa Su said the tech industry is “emphatically” not in an AI bubble and that bubble fears are “somewhat overstated.” She argued AI still needs massive chip volumes and data-center compute. Although AMD’s market cap is about $353 billion, it remains far smaller than Nvidia’s roughly $4.4 trillion (about a 12.5× gap), but AMD is gaining momentum.

She also addressed export frictions and costs: AMD will pay a 15% tax on MI308 chips as it plans to resume shipments to China. The company previously said U.S. export restrictions on MI308 would cost it roughly $800 million. Sales to China were halted earlier, then applications began being reviewed again over the summer.

On data-center bets, AMD struck a major multi-year deal with OpenAI to deploy 6 gigawatts of AMD Instinct GPUs over several years. The agreement also lets OpenAI buy 160 million AMD shares for $0.01 each, effectively giving it a 10% stake, with the first 1-gigawatt deployment slated for the second half of 2026. Su said her main worry is moving faster on innovation, and that AI is still early and models will keep improving.