Everything is going wrong for SoftBank Masayoshi Son’s all-in bet on AI is looking shaky Nov 12th 2026 At the start of the summer, the outlook was sunny for SoftBank and its swashbuckling boss, Masayoshi Son. The giant investment firm briefly toppled Toyota as Japan’s most valuable company. At the end of June the net value of its assets reached ¥72.3trn ($445bn), an increase of 80% in just three months thanks to gains in its artificial-intelligence holdings. Mr Son crowed that investors had once again underestimated SoftBank (and himself), likening his creation to a goose laying golden eggs: “What matters is not the eggs. It is the goose itself.” The period since has not been a fertile one. By early August the net value of SoftBank’s assets had dropped to ¥58.3trn, as the share prices of many of its big investments tumbled. From its peak in June, SoftBank’s market capitalisation has fallen by more than half (see chart); it now ranks only

ninth in the Japanese league. Analysts had expected the firm to report a modest profit in the most recent quarter, but on November 10th the company disclosed a $7bn loss, much of it the result of write-downs linked to its American data-centre investments. Mr Son’s eggs look less golden than cracked. To understand what has gone wrong, consider the four elements of SoftBank’s bet on AI: semiconductors, models, robotics and infrastructure. One of SoftBank’s biggest holdings is its stake of about 90% in Arm, a British company that made its name creating blueprints for chips in smartphones and other gadgets. Now Arm is pushing into selling its own silicon for use in AI data centres. Investors are worried that royalties from its original business will slow faster than new profits can replace them. Such concerns have sent Arm’s stock down by more than 50% since mid-June. In the same period shares in Intel, an American chipmaker and another of SoftBank’s investments, have also fallen by more than half. Turn next to Mr Son’s $65bn bet on OpenAI. The model-maker has delayed its plans for a public share offering, allowing its arch-rival Anthropic, which set its price range on November 10th, to beat it to the stock market. That will slow any payout for SoftBank, which borrowed vast sums to invest in OpenAI. Analysts estimate that the ratio of SoftBank’s debts to the value of

its assets has risen from 16% at the end of September to somewhere between 20% and 25%, SoftBank’s self-imposed notional maximum. Some expect it to hit 28% by March. That has made lenders twitchy. This month spreads on SoftBank’s five-year credit-default swaps hit a four-year high, implying a 31% probability of it being unable to service its borrowings. Another of Mr Son’s debt-fuelled bids, to buy the robotics arm of ABB, a European machinery-maker, is also in peril. This month regulators in China, the division’s biggest market, blocked the sale of the local business to SoftBank. They are said to be uncomfortable with the prospect of Mr Son harvesting data from ABB’s machines to help OpenAI and others train models. SoftBank and ABB are exploring whether a carve-out of the Chinese business would salvage the deal, though doing so would make the purchase less worthwhile. Last, and the latest to run into trouble, is SoftBank’s investment in AI infrastructure. This includes what will, if built, be one of the world’s largest data centres, a ten-gigawatt project in Ohio for which OpenAI is the primary tenant. Yet having rushed to nail down contracts for its data centres, SoftBank has cancelled reservations for power-generation equipment, taking a $3.7bn writedown. The company has given few details about the written-down kit, other than to deny that it was related to the project in Ohio. But the situation suggests that SoftBank may be rethinking its broader ambitions. Part of what lies behind the recent market turmoil surrounding AI is a change in the political wind. Several victorious Democratic candidates in America’s elections this month, including the next governor of Ohio, campaigned aggressively against data- centre development. Mr Son is not alone in having staked his firm and his reputation on AI. He may yet be vindicated, if Arm’s data-centre business takes off or OpenAI’s listing is a roaring success. He and his firm have suffered setbacks—and staged comebacks—before. As a new chapter in the AI fable unfolds, will the goose produce once again? ■ This article was downloaded by zlibrary from https://www.economist.com/business/2026/11/12/everything-is-going-wrong-for- softbank

Business · Business | From “Backrooms” to living rooms

YouTube’s dominance has Hollywood running scared Studios and streamers alike are struggling to respond to its success Nov 12th 2026 IN THE MOVIE industry, there are two competing storylines. The darker one —call it old Hollywood—prevailed on November 9th when Skydance, the newly formed combination of Paramount and Warner Bros, laid out plans to reduce its $87bn debt. Like other media giants, such as Netflix, its near future is one of attrition and job cuts, glossed over by promises of big hits to come. The other storyline—call it new Hollywood—is less well known and more upbeat. It is made on YouTube. To glimpse it, your correspondent travelled to an unmarked building in the Hollywood hinterland of Burbank, home to the new 32,000 square-feet studio of Smosh, a YouTube comedy channel as old as the video platform itself. The precise whereabouts are kept secret to deter over-zealous fans. Suffice to say, it is not glitzy. Similarly, Smosh’s comedy has less polish than prestige TV. Yet it is a YouTube sensation. Smosh has more than 45m followers across its channels, revenue that has grown fivefold in the past three years and a headcount that has tripled to 100, says Alessandra Catanese, its boss. “I’ve been making this joke lately, but I’m totally serious. Paramount, Disney and Netflix: their layoffs are our draft,” she says. Burbank is home to much of this YouTube demi-monde. Its scruffy creators are not big enough to offset the malaise of old Hollywood. But they have started to rattle the incumbents, which have been forced to battle YouTube for screen time in part by imitating its ways. Whether what some call “YouTube panic” is working is debatable. Yet it may be unavoidable. In the words of Mark Fischbach (aka Markiplier), who earlier this year mobilised his 40m-odd YouTube followers to turn his self-financed horror film “Iron Lung” into a box-office hit, “YouTube is the thousand-pound monster lurking over everyone’s shoulder.”

YouTube’s effect on the media zeitgeist has been palpable this year. After “Iron Lung”, two young directors who honed their craft on YouTube released “Backrooms” and “Obsession”, which became blockbuster films. The platform’s disruptive force has become increasingly evident on TV. Nielsen, a data gatherer, measured YouTube’s share of America’s TV watching at 15.8% in August, almost as much as Netflix and the streaming services of Disney, Paramount and Warner Bros combined (see chart). That is double its level in early 2023, when YouTube overtook Netflix as the biggest streamer. When it comes to attention, “YouTube is by far the biggest winner,” says Ernesto Pagano of BCG, a consultancy. The platform’s economic heft is becoming more apparent, too. For the first time this year it disclosed annual revenue, reporting $60bn from ads and subscriptions in 2025. That was more than Netflix’s $45bn and far more than other streaming platforms. KPMG, an accounting firm, estimates that it easily outspent its peers on content last year, shelling out $28bn, compared with, say, $18bn at Netflix. The more content YouTube streams, the more data it has to improve its algorithm, further boosting viewer engagement— and revenue. YouTube’s megabucks enable it to throw its weight around in ways that challenge Hollywood’s incumbents. It is investing in more traditional fare

such as sports. Next year it will offer the ad-supported tier of Peacock, the streaming service of NBCUniversal, at no charge for its Premium subscribers. Its creators, as they earn more money, are pouring it back into making longer, more professional shows. YouTube benefits from a business model that has less risk than that of its rivals. Whereas studios and streamers pay upfront to commission content, and own or license the resulting work, YouTube usually pays nothing in advance. Instead it takes a 45-55% cut of ad revenue and channel subscriptions. This pay-for-performance approach makes YouTube less of a hostage to fortune than its peers, who can waste huge sums on films that bomb. The platform’s ownership by Google, a pioneer in artificial intelligence, has also helped it to develop new technology to support creators. AI tools offered by YouTube are used to simplify editing, autodub content into other languages and automatically tag products that appear in clips. Many in Hollywood fret about leaning too heavily on AI, because surveys suggest it turns audiences off. But while other streamers such as Netflix use the technology to trim production costs, YouTube leads the way in using it to boost revenue—for its creators and itself. YouTube’s growing clout has stirred a belated fightback. Several streaming services, including Netflix and Disney Plus, have introduced swipeable short-form vertical videos similar to YouTube Shorts. Last year Ted Sarandos, Netflix’s co-chief, dismissed the creator community as a place for “killing time”. Now he is eating his words. Netflix has pledged $1bn over three years for deals with creators, according to LightShed Partners, a media-research firm. In some cases it is paying lavish sums upfront for new shows. Next year, for instance, it plans to stream a kids’ competition by Mark Rober, a YouTube science geek, that an insider at the company describes as “Pop Idol for scientists”. With Netflix’s support, the budget is said to be equal to what Mr Rober normally spends on content in a year. The poaching extends beyond creators to YouTube executives. In October Sony Pictures, one of Hollywood’s biggest studios, named Fede Goldenberg, who at YouTube helped bring “Iron Lung” and other hits to the big screen,