SoftBank's 40% Slide Raises New Questions About the AI Capital Cycle

"Lee and I both agreed that few investors fully appreciate how central SoftBank has become to the global AI capital stack. The company owns stakes in OpenAI, has backed emerging infrastructure projects like the Trump-aligned Stargate initiative, and effectively repositioned itself as the Berkshire Hathaway of AI by reallocating its Nvidia profits into higher-risk frontier plays. Yet unlike traditional cash-rich tech giants, SoftBank must borrow to fund its commitments. That leverage profile is now front and center."

"Lee pointed out that the real shock wasn't just SoftBank's liquidation of its Nvidia stake, but the fact that respected insiders like Peter Thiel also dumped every share they held. As I told him, this wasn't a typical multi-quarter trimming that shows up gradually on 13Fs. These were full-scale exits, executed quickly and all at once. When prominent AI insiders remove exposure simultaneously, it raises the question of whether the long AI rush has started to hit valuation limits."

"I highlighted that Nvidia's stock weakness goes beyond profit-taking. Alphabet recently introduced a high-end chip already deployed internally, and Meta has reportedly purchased large volumes. For the first time, cloud operators may have credible alternatives to Nvidia's 90 percent data-center GPU share. If AMD, Amazon, Microsoft, and others introduce similarly competitive hardware, what was once a near-monopoly could turn into a multi-player dogfight."

SoftBank experienced a sudden 40% decline as it shifted Nvidia profits into higher-risk AI frontier investments while leveraging to fund commitments. The company holds stakes in OpenAI and has backed infrastructure initiatives like the Trump-aligned Stargate project. Rapid, full exits from Nvidia positions by prominent insiders, including Peter Thiel, occurred rather than gradual trims, raising valuation concerns across AI assets. Competitive dynamics are shifting as Alphabet deployed a high-end in-house chip and Meta purchased significant capacity, creating credible alternatives to Nvidia's roughly 90% data-center GPU share and pressuring incumbent hardware and model leaders.

Read at 24/7 Wall St.