谷歌 DeepMind 首席执行官德米斯·哈萨比斯警告，人工智能领域部分投资已呈现“泡沫化”特征，资本规模与商业现实出现脱节。他指出，尚未拥有产品或成熟技术的新创公司却获得数十亿美元级别的种子轮融资，显示出不可持续性，并可能引发市场修正。该表态发表在达沃斯世界经济论坛期间，与英伟达和微软高管对过度投资的否认形成对比，反映出行业内部对风险评估的分化。
近期案例强化了这种担忧。前 OpenAI 高管创立的 Thinking Machine Lab 在成立仅六个月后估值即达 100 亿美元，但披露的技术细节有限，且已流失多名核心员工。同时，围绕 AI 基础设施的竞赛涉及数十亿美元债务驱动交易，其可行性高度依赖技术使用率持续增长。尽管如此，哈萨比斯强调，若泡沫破裂，大型科技公司凭借既有业务和规模优势仍具韧性。
在业绩与竞争层面，Alphabet 市值已突破 4 万亿美元，仅次于英伟达，反映市场对 AI 商业化的信心。哈萨比斯认为，美国公司在前沿模型上仍领先中国约六个月，但也承认中国在“开放”模型和近端应用上进展迅速。中国 DeepSeek 以显著低于美国同行的成本推出免费模型，引发美国市场短期波动。与此同时，关于 AI 风险的讨论升温，涉及法律诉讼与内容滥用问题，促使哈萨比斯强调安全、医学和科学应用等可量化社会收益。
Google DeepMind chief Demis Hassabis warned that parts of the artificial intelligence sector show “bubble-like” behavior, with investment levels increasingly detached from commercial reality. He cited multibillion-dollar seed rounds for start-ups lacking products or mature technology as evidence of unsustainability, likely leading to market corrections. Speaking at the World Economic Forum in Davos, his view contrasted with Nvidia and Microsoft leaders who dismissed overinvestment concerns, highlighting diverging risk assessments within the industry.
Recent cases reinforce these concerns. Thinking Machine Lab, founded by a former OpenAI executive, reached a $10bn valuation just six months after launch despite limited disclosure of its technology and the loss of several key staff. At the same time, the race to build AI infrastructure has involved multibillion-dollar, debt-fuelled deals whose viability depends on continuously rising usage of AI systems. Hassabis argued that even if a bubble bursts, Big Tech firms remain resilient due to scale and the ability to integrate AI into established businesses.
On performance and competition, Alphabet’s valuation has surpassed $4tn, ranking second globally after Nvidia, reflecting confidence in AI-driven growth. Hassabis said US firms still lead China by about six months at the frontier, while acknowledging China’s rapid progress in open models and near-term applications. China’s DeepSeek shocked markets with a free model built at a fraction of US costs, triggering short-term volatility. Parallel debates on AI risks, including lawsuits and misuse, have intensified, reinforcing Hassabis’s emphasis on safety and measurable benefits in science and medicine.