A Quarter of Two Very Different Bets
SoftBank Group's first-quarter results for fiscal year 2026, covering April through June and reported on 6 August 2026, split cleanly into a winner and a non-event. Net profit came to 347.3 billion yen (about $2.2 billion), down roughly 18 percent from a year earlier but ahead of analyst estimates, and the gap between the two headline investments explains why. Intel, a position SoftBank built for about $2 billion in August 2025 for a stake near 2 percent, delivered a 1.3 trillion yen ($8.2 billion) gain this quarter alone as the stock climbed close to 400 percent over twelve months. OpenAI, the far larger and far more closely watched position, contributed nothing: no gain, no loss, a flat mark on a stake now worth $55 billion of invested capital.
Both numbers are real. The contrast is the story. A single-digit-billion legacy chipmaker bet handed SoftBank more quarterly profit than its flagship, multi-tens-of-billions AI investment, and that is only possible because OpenAI is private and Intel is not.
Why the OpenAI Mark Did Not Move
Intel trades every second on Nasdaq, so its stake reprices constantly and mechanically. OpenAI does not trade anywhere; its valuation exists only when a funding round, tender offer or comparable transaction sets a new reference price, and SoftBank marks its holding to the most recent one of those events. No such event landed inside this quarter, so the position stayed exactly where it was: book value of $44.6 billion, fair value of $89.6 billion, both flat against the prior quarter-end. SoftBank's own language was that no new information justified changing the valuation.
That is a defensible accounting position and also an unusually strong claim for a quarter in which SoftBank itself deepened the relationship, having invested $55 billion of a commitment set to exceed $60 billion, rising toward roughly $64 to $65 billion once a further tranche lands around October 2026 for close to a 13 percent stake. A company can be pouring in tens of billions more while simultaneously telling shareholders nothing about the price has changed. Both statements are true. Neither one is informative about what OpenAI would actually fetch in a forced sale, because no forced sale has happened.
Borrowing Against the Thing It Just Marked Flat
The same earnings cycle disclosed a separate transaction: a two-year, $10 billion margin loan signed with a lending group of Goldman Sachs Bank USA, JPMorgan Chase Bank, Mizuho Securities USA, Apollo Global Funding and Sumitomo Mitsui Banking Corporation, collateralized by SoftBank's OpenAI preferred shares. The proceeds are earmarked for general corporate purposes across the group and its Vision Fund II-2 vehicle, and the facility will be drawn down this month.
The loan terms matter more than the headline figure. They include provisions requiring SoftBank to post additional cash or repay early if the value of the OpenAI preferred shares declines substantially. That single clause does something the earnings statement does not: it puts a number, held by five outside banks rather than by SoftBank itself, on what a real decline in OpenAI's private valuation would trigger. A flat internal mark carries no consequence if it turns out to be wrong. A margin loan covenant does.
What EU and UK Investors Watching SoftBank Should Track Instead
Servola's read is that the quarter's most useful signal was not the flat OpenAI mark itself but SoftBank's decision to collateralize against it anyway. A company confident that $89.6 billion undersold its stake would have little reason to accept a covenant that forces action on a decline; a company simply managing liquidity across an AI-investment program committed to well over $60 billion would sign exactly this kind of loan regardless of its private view. The two explanations are not mutually exclusive, but only one of them is testable, and it is the loan, not the earnings call.
For European institutional investors carrying SoftBank exposure directly, or indirect exposure through Arm, which SoftBank still majority-owns and which sits inside a large share of European technology and index products, the flat OpenAI mark should not be read as reassurance. It should be read as untested. The figure worth watching going forward is not next quarter's valuation commentary; it is whether the margin loan's cash-top-up trigger is ever pulled, because that is the one mechanism in this entire structure that forces OpenAI's private valuation to meet a real market test.
Read next: OpenAI's Agent Platform Has One Cloud Reseller | Intel's Foundry Grew 31%, and You Still Depend on TSMC



