Softbank Group 13F Portfolio

The biggest incremental dollar move this quarter is the creation of a new Capital One position at 0.31% of the book, worth $55.5M. The sizing is modest, but pairing a scaled U.S. card and digital bank with existing consumer‑finance rails like Nubank, Inter & Co, Klarna, Chime, and Better suggests Softbank wants more direct exposure to the economics of data‑driven lending, not just the app layer.

Life360 is the other new position, a tiny 0.0% weight at about $0.5M. On its own it barely registers, but it fits their pattern of seeding telemetry‑rich, software‑enabled consumer platforms that could eventually plug into payments, insurance, or location‑based commerce.

Conviction, though, is most obvious in what they didn’t top up. Intel, Symbotic, T‑Mobile, and the core fintech cluster (Inter & Co, Nubank, Chime, Better, Ethos, eToro) are all unchanged in share count, despite sharp mark‑to‑market moves on some of them. The absence of adds there says position size is where they want it; the thesis is about time, not more capital.

Even some of the more painful growth names — Klarna, VTEX, Seer, Neumora, ESS Tech — remain untouched. That restraint implies Softbank views them as long‑dated volatility they’re willing to ride rather than situations calling for either capitulation or aggressive averaging down.

The only real source of liquidity this quarter is Taiwan Semiconductor. Softbank cut its TSM stake by 71.5%, freeing up an estimated $678.1M while still leaving a residual 1.48% weight. That’s not a thesis abandonment; it’s profit‑taking after a 136.9% gain against their average buy price and a deliberate narrowing of the semiconductor bet back toward Intel.

The intra‑semis message is blunt: they’d rather own the turnaround and political risk in Intel than the near‑priced‑perfection execution machine in TSM. Intel at 66.81% plus smaller Ambiq and the remaining TSM stub leave them leveraged to both edge and foundry silicon, but clearly skewed toward a contrarian upside narrative in U.S. chips.

The only other trim is Satellogic, down 6.0% in shares and now a de‑minimis 0.08% position. Given the stock sits 33.8% below their average cost, this looks less like a victory lap and more like risk control on a speculative, capital‑intensive satellite imaging play.

Everything else — including underwater bets like Webtoon, VTEX, Klarna, Neumora, Seer, Dingdong, ESS Tech — is left alone. That pattern says exits are opportunistic and price‑driven (harvest where you have big gains), while the losers are treated as out‑of‑the‑money options rather than active capital drains.

Technology still utterly dominates the book at 71.18%, even after the Taiwan Semi sale nudged it down from 72.44%. That tech slab is not a generic growth bet; it is overwhelmingly a single‑name Intel position, flanked by smaller software and semiconductor options like Tempus, VTEX, YMM, Ambiq, Life360, and Satellogic.

Industrials and telecom, at 10.09% and 9.23% respectively, are effectively one‑stock sectors: Symbotic and T‑Mobile. Both saw slight weight increases driven by price rather than new capital, underscoring Softbank’s conviction in automation and bandwidth as the physical complements to its chip exposure.

Finance crept up from 5.79% to 6.29%, helped by the new Capital One stake layered onto Inter & Co, Nubank, Klarna, Chime, Ethos, eToro, and Better. This isn’t a generic “financials” call; it’s a focused bet on consumer credit, digital origination, and embedded insurance, now anchored by a scaled, regulated U.S. bank.

The rest of the book — consumer names like Webtoon, QXO, Dingdong and satellite‑enabled Globalstar; niche health‑care tools and biotech; and a sliver of education‑linked Afya — barely moves the sector bars. Their role is to give Softbank upside to content, logistics, and biology without diluting the headline theme: AI‑driven infrastructure and the financial rails that monetize data.

Softbank’s 2026‑Q2 13F is a referendum on concentration as a strategy. With one name at 66.81% of exposure and the top‑10 at 96.8%, they are explicitly trading tracking error for the possibility of outsized gains from Intel’s restructuring and AI‑foundry ambitions.

The decisive trim in Taiwan Semi and the absence of any Intel selling say they think the market is closer to “fair” on TSM than it is on Intel. In other words, the easy money in the best‑in‑class Asian foundry may be behind us, while the payoff for a successful Intel turnaround remains asymmetric.

At the same time, modest adds in regulated, scaled finance (Capital One) and data‑rich platforms (Life360) show Softbank quietly extending its thesis from pure technology into the profit pools that sit on top of it. The fintech cluster, health‑tech bets like Tempus and Recursion, and industrial bio tools such as PacBio and Seer all rhyme with that idea: compute plus data unlocks new products in money and medicine.

Going forward, unless there is a dramatic strategic shift, this book will live or die on a handful of infrastructure calls: Intel’s ability to capture AI silicon economics, Symbotic’s capacity to automate physical distribution, and T‑Mobile’s role in a denser, more data‑hungry wireless world. The long tail of smaller growth positions is best understood as optionality — they may add spice to returns, but the core risk budget is already spoken for.