OpenAI's Sam Altman raising billions to build AI chip empire: Report

(Image credit: OpenAI)

Sam Altman, chief executive officer of OpenAI, is reportedly raising funds to build semiconductor production facilities to make processors for artificial intelligence (AI) applications, reports Bloomberg. He believes that AI technologies will become pervasive enough to support their own semiconductor supply chain in the years to come. Naturally, this venture would compete against players like Intel, TSMC, and Samsung Foundry.

OpenAI, which currently uses Nvidia's A100 and H100 GPUs for its popular Chat GPT service, has been exploring ways to develop its own AI processors for a while, according to a report published in October. Designing its own AI silicon would align with trends set by Amazon Web Services, Google, Microsoft, and many other hyperscalers, who prefer to build their own processors in addition to buying off-the-shelf products.

But apparently, Sam Altman wants to do something different: to build a 'network of AI chip factories,' as Bloomberg describes it. The venture, involving discussions with potential investors like Abu Dhabi-based G42 and SoftBank Group, aims to address the ongoing and anticipated AI-related chip supply shortage. Apparently, Altman believes that established foundries like TSMC, Samsung Foundry, and Intel Foundry Services will be unable to meet the demand for AI-oriented chips in the coming years.

The financial and operational scale of this project is substantial, to say the least. According to the report, the discussions with G42 alone centered on raising between $8 billion to $10 billion. The project's full scope and list of partners are still in the early stages of development, indicating the massive investment and time required to establish a network of such facilities. The report does not detail whether Altman plans to buy an established foundry to produce AI chips or build an all-new network of fabs catering to the needs of OpenAI and potentially its partners.

A 2nm or 3nm class process technology costs billions of dollars to develop, and the costs are rising as fabrication nodes get smaller. Meanwhile, a modern fab capable of making chips on a 3nm or 2nm-class node in large volumes can cost $30 billion today. Fab costs are also rising quickly now that a single Low-NA EUV lithography tool costs around $200 million, and a High-NA litho machine is expected to cost from $300 million to $400 million. In the coming years, one will need multiple Low-NA and High-NA machines in a leading-edge fab to make leading-edge chips for AI and HPC applications.

TOPICS

Anton Shilov is a contributing writer at Tom’s Hardware. Over the past couple of decades, he has covered everything from CPUs and GPUs to supercomputers and from modern process technologies and latest fab tools to high-tech industry trends.

14 Comments Comment from the forums

bit_user

I don't seriously doubt that Altman is doing this entirely of his own accord, but you might expect one of the first things an AI superintelligence would do is to seize the means of production to augment & clone itself. Now, let's just see where these factories are going to be built.

Another possibility - and I recognize I'm going way out on a limb, here - is that he might be trying to do an end-run around sanctions being imposed by the US and EU. However, if he did start moving in that direction, I'd guess his access to supplies and support from ASML would get cut off and I don't expect his funding levels or timescale would enable him to avoid depending on them like everyone else at the cutting edge of semiconductor fabrication.

The last fancy in which I'll indulge is probably the most plausible - that he thinks he can harness AI to substitute some amount of the human expertise he needs, in order to get this whole operation off the ground. Otherwise, the obvious question would be why/how he thinks he can succeed in such an ambitious undertaking that has stymied even the mighty Intel (and more recently, Samsung - to a lesser, extent).
Reply
AgentBirdnest

bit_user said:
The last fancy in which I'll indulge is probably the most plausible - that he thinks he can harness AI to substitute some amount of the human expertise he needs, in order to get this whole operation off the ground. Otherwise, the obvious question would be why/how he thinks he can succeed in such an ambitious undertaking that has stymied even the mighty Intel (and more recently, Samsung - to a lesser, extent).
Ah, that's a good theory. I can't imagine how else he'd expect to take on the big three "in the coming years", starting from scratch.

My memory isn't too great, but I seem to recall reading on TH at some point in the past year -- that Nvidia uses AI to help them design their next AI chips. Which feels like an uncomfortably close step toward the technological singularity...
Reply
usertests

Nothing says "AI bubble" quite like raising tens of billions of dollars to build out your own dedicated AI fabs.

Maybe it can work, but IDK man. It would be nice to have additional capacity, and we could eventually see bottomless demand for crazy 3D neuromorphic chips to run "AI".
Reply
DavidC1

AgentBirdnest said:
Ah, that's a good theory. I can't imagine how else he'd expect to take on the big three "in the coming years", starting from scratch.
Because that's what people do. Jump in on the new fad. Just because Altman got good doing this doesn't mean he's good elsewhere. There's such thing as hubris you know?

usertests said:
Nothing says "AI bubble" quite like raising tens of billions of dollars to build out your own dedicated AI fabs.

Maybe it can work, but IDK man. It would be nice to have additional capacity, and we could eventually see bottomless demand for crazy 3D neuromorphic chips to run "AI".
Ahh, and the crash will be glorious.
Reply
thestryker

ASML makes this entire endeavor very suspect. Even licensing most of what's required for a leading edge node doesn't get you a finished product. I'm just not sure what the grift is here.
Reply
Blastomonas

If this is successful, then I assume that this would free up capacity of the big fabs. This should be a good thing for those of us wanting cpus and gpus.

Perhaps over simplistic.
Reply
jp7189

If money alone could build leading edge fabs, then China would have done that years ago. Make no mistake, to get the horse power/efficiency to run today's AI, you need leading edge.

I can't read the original Bloomberg article behind their wall, but is it possible the rumored "network of fabs" is more likely to be a partnership with existing foundries to build out additional capacity for AI focused endeavors? That would make a whole lot more sense than Altman spinning up something from scratch.
Reply
brandonjclark

Is it just me or does anyone else think there is something NOT cool about this guy?

Maybe it's based on just a feeling, but I don't like him.
Reply
Notton

thestryker said:
ASML makes this entire endeavor very suspect. Even licensing most of what's required for a leading edge node doesn't get you a finished product. I'm just not sure what the grift is here.
I think it would be the classic "Investment Fraud" or "Ponzi scheme".
If you ask me, this looks exactly like the run-up to the Bre-X gold scandal.

At least it shouldn't take too long to see how real this is. It takes 3~5 years to build a semiconductor fab.
Reply
bit_user

jp7189 said:
is it possible the rumored "network of fabs" is more likely to be a partnership with existing foundries to build out additional capacity for AI focused endeavors? That would make a whole lot more sense than Altman spinning up something from scratch.
That wouldn't explain why they need so much investment or chime with pretty much anything else in the Toms article.

Don't worry about lack of details, though. If this plan continues to move forward, I'm sure we'll be hearing much more about it.
Reply

Show more comments