SoftBank plans record yen bond sale to fund its AI push

Money is pouring into the AI build-out from every direction today — SoftBank is tapping Japanese retail investors for a record bond to keep funding its OpenAI bet, while South Korea just rewrote its privacy law to let AI developers actually use the data they're collecting.

SoftBank plans a record one trillion yen retail bond to fund its AI bets. SoftBank Group said it will issue one trillion yen (roughly $6.6 billion) of seven-year retail bonds in Japan, the largest by any issuer in the country, with an indicative coupon of 4.3–4.9% and pricing expected on September 4. Proceeds will refinance 400 billion yen of retail bonds maturing in September and help close the funding gap for its AI strategy — the company has committed more than $60 billion to AI investments, including its stake in OpenAI, and faces a shortfall of around $20 billion. The move is a deliberate workaround of cautious banks: with limited institutional appetite for long-duration credit risk, SoftBank is going directly to everyday savers for liquidity. It's the clearest sign yet that funding the frontier is no longer a venture or sovereign-wealth affair alone — the retail crowd is now financing OpenAI's runway, and the seven-year horizon says SoftBank expects the payoff to be slow and steady rather than immediate. Alibaba, by contrast, raised $10 billion in Hong Kong's largest-ever secondary listing on the same day for chips and data centers, underscoring just how broad the capital race has become.

South Korea passes a law letting AI developers train on personal data. The National Assembly has unanimously passed amendments to the Personal Information Protection Act that, for the first time, allow AI developers to use already-collected personal data for model training without obtaining per-subject consent — as long as the Personal Information Protection Commission (PIPC) approves the use on a case-by-case basis. The bill sets three hurdles: developers must show anonymization or pseudonymization can't achieve their goal, establish strong security measures, and demonstrate a public-benefit purpose with low risk to individuals. It takes effect six months after cabinet approval. Korea's previous framework — requiring consent for each dataset, with only narrow two-year exemptions for voice-anti-fraud and autonomous-driving uses — had become a wall for the large models the government is now spending heavily to build. The oddity is how the country chose to solve it: instead of predefined exemptions like Japan's, it hands the regulator sole discretion over every application. That "trust the regulator" model settled a debate among privacy lawyers about who watches the watcher, but it also leaves developers unable to predict whether their use case passes until they ask.

South Korea is betting the PIPC gets every AI-data call right — do you think regulator-by-regulator approval will hold up as models scale, or will it become the bottleneck? Tell us in the comments.

AI 101 — How do AI video generators work?

An AI video generator is a model that starts with a grid of random noise and repeatedly refines it into a moving picture that matches your prompt — the same trick image generators use, stretched across time. Nothing is filmed, and nothing is stitched together from clips the model has seen. The pixels in the finished video were invented, frame by frame, all at once. Why it matters right now Kling 4.0, Kuaishou's flagship video model, went fully live on October 15 with up to 30 seconds of vid

Houmo moves compute inside memory — and A20 enters production

Beijing chipmaker Houmo used its partner conference this week to lay out two compute-in-memory routes at once — and to say the first chip built on the newer one is already in production. Houmo's A20, its first chip built on a 3D compute-in-memory architecture, has entered production. The part supports up to 16 chips fully interconnected, claims a 50% improvement in compute utilization over the previous generation, and covers models from 3B to 30B parameters, according to co-founder and product

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Kuaishou's Kling AI has spent two years becoming the video model people actually finish projects with. Its new flagship, Kling 4.0, went fully live on October 15 — and this round is less about prettier frames than about the two things that make a generated shot usable in a real edit: length and control. Kling 4.0 generates up to 30 seconds in a single native pass, outputs 10-bit HDR at 4K, 1080p and 21:9, and takes up to 15 reference assets in one task. That budget splits across ten images, fi

Brain organoids classified Braille at 83% — in a trio, not alone

Two stories today: human brain tissue on a chip decoded a tactile signal, and the data-centre company behind some of the largest AI contracts in Europe turns out to be six months behind on its flagship site. Human forebrain organoids grown on a microelectrode array classified Braille letters with 61 percent accuracy on their own and 83 percent when the responses of three organoids were combined, in a peer-reviewed study from the University of Bristol. The team took event-based tactile readings

Read more

AI 101 — How do AI video generators work?

Houmo moves compute inside memory — and A20 enters production

Kling 4.0 goes live with 30-second native video at 10-bit HDR

Brain organoids classified Braille at 83% — in a trio, not alone