AMD Completes Acquisition of Silo AI

[btarunr]

AMD today announced the completion of its acquisition of Silo AI, the largest private AI lab in Europe. The all-cash transaction valued at approximately $665 million furthers the company's commitment to deliver end-to-end AI solutions based on open standards and in strong partnership with the global AI ecosystem. Silo AI brings a team of world-class AI scientists and engineers to AMD experienced in developing cutting-edge AI models, platforms and solutions for large enterprise customers including Allianz, Philips, Rolls-Royce and Unilever. Their expertise spans diverse markets and they have created state-of-the-art open source multilingual Large Language Models (LLMs) including Poro and Viking on AMD platforms. The Silo AI team will join the AMD Artificial Intelligence Group (AIG), led by AMD Senior Vice President Vamsi Boppana.

"AI is our number one strategic priority, and we continue to invest in both the talent and software capabilities to support our growing customer deployments and roadmaps," said Vamsi Boppana, AMD senior vice president, AIG. "The Silo AI team has developed state-of-the-art language models that have been trained at scale on AMD Instinct accelerators and they have broad experience developing and integrating AI models to solve critical problems for end customers. We expect their expertise and software capabilities will directly improve the experience for customers in delivering the best performing AI solutions on AMD platforms."



View at TechPowerUp Main Site

 

[Quicks]

Waste of money, should have used it for research and development of their current lineup. Which is very lacking of late.

 

[DeathtoGnomes]

with all these AI acquisitions all over the place, you gotta wonder why you havent heard of any of those companies. it seems that these "startups" are put together just to be sold for profit.

 

[ZoneDymo]

with all these AI acquisitions all over the place, you gotta wonder why you havent heard of any of those companies. it seems that these "startups" are put together just to be sold for profit.

I mean yeah? that has been the case for just about any of them, hell that is what a start up is for I feel more then anything.

 

[Daven]

with all these AI acquisitions all over the place, you gotta wonder why you havent heard of any of those companies. it seems that these "startups" are put together just to be sold for profit.

I think that’s where the word startup comes from. The prep work required to get someone to come in and buy the company. Very few companies organically grow from a garage to 100,000 employees without outside investment or being purchased. Both routes are basically the same. Someone richer takes over.

By the way there is no expiration on being a startup. A company can last in such a state for a long time.

 

[Rares]

Waste of money, should have used it for research and development of their current lineup. Which is very lacking of late.

This is not "a" waste of money, this is the future !

 

[Vya Domus]

Yikes, buying up companies at the height of this AI tech bubble is not good at all.

with all these AI acquisitions all over the place, you gotta wonder why you havent heard of any of those companies. it seems that these "startups" are put together just to be sold for profit.

Wonder no more, just look up the dot com bubble, this happened before, you had all these companies pop up from nowhere like weeds with zero products/revenue and they'd get acquired for millions just because they were claiming to do something vaguely related to the internet.

 

[Quicks]

Yikes, buying up companies at the height of this AI tech bubble is not good at all.


Wonder no more, just look up the dot com bubble, this happened before, you had all these companies pop up from nowhere like weeds with zero products/revenue and they'd get acquired for millions just because they were claiming to do something vaguely related to the internet.

Makes you think this is how big companies move money around for some dodge dealings or to avoid Tax on profits or something that need to be done under the table.

 

[evernessince]

Waste of money, should have used it for research and development of their current lineup. Which is very lacking of late.

You don't seem to realize that investing in AI is putting the money towards R&D. AI is used to help develop modern chips.

Mind you, AMD could also use more expertise in the field to push it's hardware and software AI stack. Excitement around AI will eventually reduce but the technology will remain huge in many ways and represents an entirely new market of significant size.

Now maybe you can make the argument that investing in this particular AI company was a "Waste of money", I honestly don't follow AI investing and know nothing about them.

 

[R0H1T]

Wait till you find out who's coming for yer jobs next

Waste of money, should have used it for research and development of their current lineup. Which is very lacking of late.

Basically everything we do can be broken down into maths & equations! It's just a matter of when not if

 

[JWNoctis]

Optimists point out that most people who think AI is not "for real" might have at most tried ChatGPT, and not a true "4-level AI." That is, full-sized GPT-4. You can run almost-GPT-4 level LLMs on your own computer right now, if you have enough RAM.

Pessimists point out that even GPT-4 seemed to have little commercial utility, in the sense of starting to replace significant percentages of all human labour.

Who knows? This acquisition might do AMD good in the end.

 

[thesmokingman]

Never fails how some posters here act like they know better.

 

[Vya Domus]

GPT-4 seemed to have little commercial utility

This is the real problem, money. Billions upon billions are spent on this but these things are simply not generating enough money to sustain themselves, OpenAI has survived because they got literal billions in investments and not because their product is a money printing machine.

 

[Steevo]

I can't wait for the AI bubble to burst, it truly is the .com bubble.all over again.

 

[LabRat 891]

Optimists point out that most people who think AI is not "for real" might have at most tried ChatGPT, and not a true "4-level AI." That is, full-sized GPT-4. You can run almost-GPT-4 level LLMs on your own computer right now, if you have enough RAM.

Pessimists point out that even GPT-4 seemed to have little commercial utility, in the sense of starting to replace significant percentages of all human labour.

Who knows? This acquisition might do AMD good in the end.

How much RAM is enough? Splurged on 64GB as part of my AM4 build's 'final upgrade'.
The biggest issue I have with 'AI' is it being mostly off-site/cloud.

 

[maxfly]

Hopefully for AMDs sake they hit on something big in the industrial or commercial arena because consumer wise, people are already sick of AI.

 

[Wirko]

I can't wait for the AI bubble to burst, it truly is the .com bubble.all over again.

Bubble? Burst? It was a medium-size correction before the cancer-like growth that hasn't stopped until now, it just ceased a bit in 2023. One more bubble burst like that, and Nvidia + Amazon + Apple + Microsoft + Google + a couple AI newcomers will constitute 80% of the world economy before 2050.

 

[JWNoctis]

How much RAM is enough? Splurged on 64GB as part of my AM4 build's 'final upgrade'.
The biggest issue I have with 'AI' is it being mostly off-site/cloud.

FWIW 64GB is sufficient for quantized 70B versions of Llama 3.1, its finetunes, and similar-sized LLMs. 96GB and it's Mistral Large 2, "almost" GPT-4 level stuff. It is kind of slow, but usable. The main capability bottleneck in current consumer PCs is RAM size, but the speed bottleneck is mostly-sequential RAM bandwidth.

The largest Llama 3.1 models, reasonably quantized without too much damage to capability, need 200+GB just to load, before any allocation of working memory.

Bubble? Burst? It was a medium-size correction before the cancer-like growth that hasn't stopped until now, it just ceased a bit in 2023. One more bubble burst like that, and Nvidia + Amazon + Apple + Microsoft + Google + a couple AI newcomers will constitute 80% of the world economy before 2050.

It won't matter if AI actually works out; AI capable of surpassing human capability in all categories would do better without humanity, and lesser intelligence cannot reliably predict - or control - actions or designs of greater intelligence. It's Scylla of burst soap scum and dashed dreams or Charybdis of AI doom in some form, and there is not much of a middle ground I could see right now.

 

[Minus Infinity]

You don't seem to realize that investing in AI is putting the money towards R&D. AI is used to help develop modern chips.

Mind you, AMD could also use more expertise in the field to push it's hardware and software AI stack. Excitement around AI will eventually reduce but the technology will remain huge in many ways and represents an entirely new market of significant size.

Now maybe you can make the argument that investing in this particular AI company was a "Waste of money", I honestly don't follow AI investing and know nothing about them.

You fail to realise, not one single company wasting billions on mostly Nvidia's AI hardware is getting a cent in revenue from said AI. This will be a massive dotcom style bubble. These companies need to make nearly a trillion a year in earnings to pay for the outlays. Good luck with that.

 

[csendesmark]

Waste of money, should have used it for research and development of their current lineup. Which is very lacking of late.

There is a big "AI" bubble on the stock market,
One day it is going to burst...
Just like the dot-com bubble, it will be ugly

 

[dragontamer5788]

Yikes, buying up companies at the height of this AI tech bubble is not good at all.

$665-million all cash is fine. This is a small company / small purchase in the great scheme of things.

As long as the focus is on "Good GPU programmers", its a fine purchase. AMD is weak on the software front, and the software being written today is AI. I am not bullish on AI at all, but AMD is sorely in need of good programmers.

 

[evernessince]

You fail to realise, not one single company wasting billions on mostly Nvidia's AI hardware is getting a cent in revenue from said AI.

This will be a massive dotcom style bubble.

Saying this is like the dotcom bubble isn't the negative implication you seem to think it is. Online retail rose during the dotcom bubble and became successful because of the funding it provided:

Dot-com bubble - Wikipedia

en.wikipedia.org

"In 2000, the dot-com bubble burst, and many dot-com startups went out of business after burning through their venture capital and failing to become profitable.[5] However, many others, particularly online retailers like eBay and Amazon, blossomed and became highly profitable."

This is the same for any new market, a ton of startups pop up and most of them go out of business finding what does and doesn't work.

If you are implying that AI has no use you'd be dead wrong, just from a medical and engineering perspective it's already indispensable.

 

[dragontamer5788]

Saying this is like the dotcom bubble isn't the negative implication you seem to think it is. Online retail rose during the dotcom bubble and became successful because of the funding it provided:

We're calling it a dotcom bubble because most of us are investors of some kind. The immediate effects of the dot-com bubble were random companies winking out as bankrupt, even well respected ones.

There was basically no way to tell Ask Jeeves vs Google in the 90s. Similarly today, there is absolutely no way you can tell a good AI company from a bad one. And the vast, vast majority of AI companies will be bad, even if there's a couple of good apples in there somewhere.

The only guaranteed winner here seems to be NVidia. However, USA is running out of electricity this year due to the massive increases in training / GPUs getting plugged into new data centers (!!!), the expansion of GPU-data centers will be forcibly stopped in the near future due to lack of power. So even NVidia's growth has an end.

Oh, that’s not good: Energy prices at PJM capacity auction skyrocket 9x

Insufficient future transmission planning, the retirement of fossil-fired generation, long interconnection queues and the implementation of FERC market reforms are all contributing to the price hikes.

www.power-eng.com

The auction produced a price of $269.92/MW-day for most of the PJM footprint, compared to $28.92/MW-day for the 2024/2025 auction. Capacity auction prices fluctuate annually based on the need for investment in generation resources, but a more than 800% increase will have a massive ripple effect across PJM’s 13-state footprint.

 

[Vya Domus]

As long as the focus is on "Good GPU programmers", its a fine purchase.

There must be less expensive options to acquire talent.

 

[dragontamer5788]

There must be less expensive options to acquire talent.

Its usually less expensive to buy talent after they've proven to build a successful product.

Alternatively, you hire them as interns and wait like 5 years, and then maybe 50% of them are reasonable programmers and maybe 10% of them are comparable to a company purchase in terms of skill level. And those 10% will be looking for new jobs as their resumes have improved and there's no guarantee they actually stay at your company.

Some serial entrepreneurs will never join a big company (and small companies like this have plenty of those kind of programmers). But others are tired of the startup-uncertainty and are happy to get taken under the care of a larger company.

Acqui-hires are a trope for a reason.