SoftBank's Son Seeks To Build a $100 Billion AI Chip Venture

An anonymous reader quotes a report from Reuters: SoftBank Group Chief Executive Officer Masayoshi Son is looking to raise up to $100 billion for a chip venture that will rival Nvidia, Bloomberg News reported on Friday, citing people with knowledge of the matter. The project, code named Izanagi, will supply semiconductors essential for artificial intelligence (AI), the report added. The company would inject $30 billion in the project, with an additional $70 billion potentially coming from Middle Eastern institutions, according to the report.

The Japanese group already holds about a 90% stake in British chip designer Arm, per LSEG. SoftBank is known for its tech investments with high conviction bets on startups at an unheard of scale. But it had adopted a defensive strategy after being hit by plummeting valuations in the aftermath of the pandemic, when higher interest rates eroded investor appetite for risk. It returned to profit for the first time in five quarters earlier this month, as the Japanese tech investment firm was buoyed by an upturn in portfolio companies.

Re:

[Iamthecheese]

This is entering right at the beginning of the AI game. Late enough to not make NVidia's mistakes (we don't know what most of those are yet) and early enough to catch the rising tide. This has the potential to earn big.

>Trump

Oh, I get it, you're retarded. I'm sorry.

Re:

[iAmWaySmarterThanYou]

Who says nvidia made mistakes? Often being first yields an unstoppable advantage even if there were some glitches along the way.

Re:

[iAmWaySmarterThanYou]

Woof, puppy!

My local waitress has a better life than you ever will.

Re:

[Iamthecheese]

Responding because of what AC wrote. Someone did something similar to me before, posting a bigoted rant in response to someone's response to my post thereby looking like that was me ranting. So it was My post->Response->AC bigoted rant. So my post was discounted and the overall quality of the discussion became collateral damage. I wish there was a better way to respond to this tactic.

Re:

[iAmWaySmarterThanYou]

Because the admins here don't admin.

The mod system is broken. They don't pull mod points from people who abuse the mod system. They don't ban obvious trolls who abuse AC. They don't show who votes which mod option anymore. They killed meta-moderation.

I don't think the site makes enough money for them to bother. One day it will 404.

I've toyed with the idea of buying it and fixing the anti-features which encourage social malfunction but ultimately don't really care that much.

Take it for what it is and en

Sibling rivalry

[Going_Digital]

What about Softbank’s daughter, what is she building :)

Re:

[SeaFox]

I've heard "corporations are people" but I didn't know they could actually procreate.

Re: Sibling rivalry

[carvalhao]

What do you think a spinoff is?

Japaneto's Sound Good

[zenlessyank]

Hopefully they make them out of that Panko.

Great track record with WeWork

[iAmWaySmarterThanYou]

Isn't this the same guy who was responsible for blowing about $10 billion USD on WeWork?

And now he wants to do what Silicon Valley VC's always do: follow the crowd like a sheep.

I wonder how much he lost in NFT apes, quatum computing, AI, bitcoin, and whatever other fotw sheep lovers bullshit over the years.

My dog could pick as well as this guy. Just put him in an enclosed area, mark off squares with fotw tech words and drop a few billion into whatever he pees on.

Re:

[cheesybagel]

Yeah. This guy is a disaster of an investor.

I have a name for his company all ready to go

[Morromist]

Son Microsystems

Re:

[RUs1729]

Son Microsystems

Well, I'll be a sun of a bitch.

Yeah, tech ventures like WeWork.

[thesjaakspoiler]

We all know the story about Adam and Rebekaaaaaaah.

What is needed?

[larryjoe]

The big question is what resources are needed to build a competitive AI processor: money, people, or time or something else.

The other even bigger question is why Softbank will be successful when most other companies struggle.

AI processing boiled down seems straightforward and simple. Software frameworks already exist that at the lowest level simply need to do a bunch of matrix multiplies and other basic options. Slapping together a bunch of multipliers on a chip is trivial, so why isn't it easy for both la