The chief executives of Nvidia and SoftBank have issued parallel dismissals of fears that artificial intelligence investment is inflating a bubble, even as Nvidia’s own internal allocation struggles expose the extreme scarcity of the chips powering the boom. Speaking in Tokyo, Nvidia’s Jensen Huang said the sector remains “very far from an AI bubble” and described current demand as “extraordinarily strong”. SoftBank’s Masayoshi Son, addressing the company’s annual event, called the bubble question “foolish” and “absurd”. SoftBank shares rose by double digits on the Tokyo exchange following its quarterly earnings, which showed profits surging nearly five-fold to ¥5 trillion, driven by AI investments.

Nvidia’s automotive chief, Xinzhou Wu, disclosed that teams across the company compete weekly for limited GPU compute, with Huang occasionally stepping in to settle disputes. The competition extends beyond silicon to semiconductor fabrication capacity, Wu said, as the company balances near-term revenue against long-term bets such as autonomous driving—what Huang calls “the zero trillion dollar business”. The internal rationing illustrates a supply-demand imbalance that has made Nvidia the world’s most valuable company and left even its own divisions scrambling for resources.

Son forecast that annual global AI investment will reach $5 trillion and that AI-related industries will account for 20 per cent of world GDP by 2040. He predicted a shift to an “agent-centric world” with 100 trillion autonomous AI agents, and said data centres would require 3 terawatts of power generation capacity—1.8 times current global electricity consumption—initially fuelled by gas before fusion takes over. SoftBank, which sold its entire Nvidia stake for $5.83 billion to fund AI bets, has ploughed $34.6 billion into OpenAI and is pursuing chip manufacturing and energy projects, including a battery business in Japan and talks on a $1 trillion AI manufacturing hub in Arizona.

In Tokyo, Huang highlighted Japan’s advantage in “physical AI”—autonomous robots and hardware—citing its precision manufacturing and chronic labour shortages. He stressed the concept of sovereign AI, arguing that nations must develop their own intelligence domestically, and said Nvidia would make significant sovereign AI announcements this week. The twin interventions from the heads of the world’s most valuable chipmaker and one of its most aggressive AI investors frame the current moment not as speculative excess but as the early phase of a supply-constrained, multi-decade infrastructure cycle.