Urgent.News

What's breaking now, across thousands of outlets.

AI

SoftBank seeks $100 billion for AI-refined projects from Middle Eastern investors

SoftBank plans to buy companies that do not use AI or robotics, then apply new technologies to improve their operations and valuation.

SoftBank seeks $100 billion for AI-refined projects from Middle Eastern investors

SoftBank is aiming to raise up to $100 billion from Gulf investors to create a fund focused on acquiring companies and enhancing their operations through AI and robotics, as reported by the Financial Times. This new initiative comes amid SoftBank's struggles with its $65 billion stake in OpenAI and a delayed IPO. The proposed fund differs from previous SoftBank investment vehicles, which mainly supported technology startups.

Instead, it targets established firms to integrate AI and advanced technologies, with the goal of boosting their value—a novel approach to profiting from AI. SoftBank's robotics and physical AI division, Roze, is set to play a crucial role in these transformations, although specifics are still undefined. Son intends to eventually bring Roze public at a high valuation, but the fund's establishment and financing are not yet confirmed.

SoftBank has already invested $65 billion in OpenAI, which remains the highlight of its investment strategy. The Japanese firm has funded investments using internal resources, external capital, and borrowing, partly secured by its Arm holdings. SoftBank's fortunes heavily hinge on OpenAI's valuation, which has not yet turned profitable.

Despite the risks linked to OpenAI's valuation, SoftBank persists in its aggressive AI expansion. Previously, Son successfully attracted significant funding from Middle Eastern sovereign wealth funds, such as Saudi Arabia's Public Investment Fund and Abu Dhabi's Mubadala, which participated in the original $100 billion Vision Fund in 2017.

Since its inception, Vision Fund 1 has earned around $29 billion in investments, while Vision Fund 2, primarily financed by SoftBank, has generated $20.5 billion as of June. Abu Dhabi has also increased its AI investments through MGX and G42. Recently, SoftBank secured over $11 billion through the largest junk bond issuance ever, with interest rates of 9.75%.

As of June, SoftBank's net asset value was ¥72.3 trillion ($456.57 billion), with a loan-to-value ratio of 13%, well below its usual limit of 25%. However, a large chunk of SoftBank's assets is invested in technology firms like Arm and OpenAI, so a drop in their valuations could raise the company's loan-to-value ratio to an uncomfortable level, even without additional borrowing.

Son's investment track record includes Alibaba's success and WeWork's 2023 bankruptcy, leaving the future of his current strategy in SoftBank uncertain.

Written by urgent.news from Tom's Hardware's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at tomshardware.com →

More in AI

What If Our Desktop Had Its Own Little AI Companion? Meet Luna

This is a submission for the MLH x DEV Writing Challenge Meet Luna 🐱 What I Built My husband and I have always liked those little animated desktop pets — the tiny characters that move around your…

  • Luna is an AI desktop companion designed for practical help.
  • Built with Electron, React, TypeScript, and Google's Gemma AI model.
  • Challenges in implementation taught lessons about AI product development.

Logistic Regression

When you start your journey in Machine Learning, the first algorithm you usually learn is Linear Regression. Linear regression draws a straight line using the equation: y = m x + c This straight line…

  • Logistic Regression is foundational in Machine Learning, learned post Linear Regression
  • Predicts binary outcomes like student placement (1) or not (0) based on exam scores
  • Converts raw scores into probabilities between 0 and 1 using Sigmoid function

Noticed: a quiet photo diary with captions from a local AI

This is a submission for the Hacktoberfest Open-Source AI Challenge Week 1: Touch Grass What I Built Noticed is a quiet photo diary that runs entirely on your own computer.

  • Noted is a local AI-powered photo diary
  • Gemma 3 model generates captions locally
  • Photos and captions never leave user's computer

I Measured My RAG Pipeline Honestly. It Was 40x Slower Than I Thought.

A few days ago I published the architecture behind Vicquant’s RAG Vault; a 9-stage retrieval pipeline built to ground financial AI answers in actual source documents, with strict citations, so a user…

  • RAG pipeline measured at 14.81 seconds mean latency, 21.28 seconds p95
  • Routing and concurrency optimizations reduced latency to 6.24 seconds
  • OpenRouter routing inconsistency improved by pinning efficient providers

More from Saturday 10 October →