Groq stopped serving Kimi K2 (1T params) when they got aquihired by NVIDIA, so I guess NVIDIA took most of the hardware in addition to the employees. The largest model they serve now is the relatively minuscule gpt-oss-120b.
I don't really get the value proposition of groq as a user, the performance is really poor for the token price. Data centres on the other hand are becoming a commodity, and I don't see any reason a priori to invest in groq specifically for something like that.
Groq is considerably faster and better at inference, they have a totally superior product to Nvidia for inference based tasks, which will be the dominant concern in the future.
Plausibly, take all the Nvidia hype and multiply that by a factor and that's what 'Groq' could be worth.
And there is no real commodification - there's Nvidia, Cerebras, Groq ... not many otheres.
I love Groq, but across every single line break in your post there is a glaring issue that is easy to refute with in 15 seconds, even without 300t/s of throughput.
You wasted all of your commentary on snark and sadly unfunny humour, and yet still managed to add nothing.
Groq is more performant for the growing categories of inference-based tasks, wherein Nvidia's advantage in inference depends bulk/batch processing which will make up a smaller category over time, in relative terms.
The future of AI Silicon is inference, and the cost structure of AI data centres is constrained around the current necessity to have 'high GPU utilization' otherwise, the cost / amortization of the chips doesn't work out.
That cost structure is a limitation of Nvidia architecture.
Groq serves a lot faster, and without the limiting batching requirement, which opens hosting arrangements common in most classical hosting scenarios aka without necessarily the high utilization requirements.
Groq has bespoke hardware, lack of CUDA, much lower memory desnsity obviously and they don't have the deep distribution networks and leverage over TSMC that Nvidia has - but pound for pound, were we to be able to 'fire up a server' for our inference needs, it would be Groq, not Nvidia that we'd turn to.
Were they not a later market entrant and didn't have those barriers to entry, they'd be gigantic.
Haha. Groq is trash. It can’t be used for anything where reliable work is required.
Groq lasted how long in the social contract experiment? Libertarians always forget that rules and government aren’t tyrannical they are the mechanism to ensure bullies don’t destroy everyone’s freedom.
I don’t understand one part of the licensing here: if it was just a license, can’t they relicense the software and hardware of LPU3 to AMD? Or hire new software and hardware people?
The new designs were their main asset besides the amazing talent that went to NVIDIA, not the remaining DCs.
My company had a really terrible experience trying to use Groq, and I would NOT recommend anyone use their service if you need reliability. So many random errors, so many silly quirks.
The most bizarre thing here is the reporters. They intentionally misrepresented what happened. All the news stories from last year claimed Nvidia "acquired" Groq - and in the same story, quoted what actually happened, and pretended it didn't. It's like the journalists had some kind of group psychosis, pretending it was an acquisition. It wasn't. It was a really white-glove product rental.
From the actual Groq PR release:
"Groq announced that it has entered into a non-exclusive licensing agreement with Nvidia" - "As part of this agreement, Jonathan Ross, Groq’s Founder, Sunny Madra, Groq’s President, and other members of the Groq team will join Nvidia to help advance and scale the licensed technology" - "Groq will continue to operate as an independent company with Simon Edwards stepping into the role of Chief Executive Officer" (https://groq.com/newsroom/groq-and-nvidia-enter-non-exclusiv...)
Nothing about an acquisition there. It says Nvidia is licensing it, and that others can too. The execs work for Nvidia to integrate it into Nvidia's... something. And Groq the company remains the same as before.
There's also no official source for the amount Nvidia paid for the tech, or two unofficial ones. Journalistically speaking, this is some bullshit.
Why's the deal like this? No idea. Does it make sense? No idea. But it's not odd that Groq is continuing to raise more money, because they never stopped being a normally operating company.
If you want an explanation for why Nvidia would do this deal, my best offer is here (https://openrouter.ai/rankings): Of the top 10 fastest AI models, Groq is the provider of 4 of them. And of the price of those top 10 fastest AI models, Groq is #1, #2, #3, and #5. And you wonder why someone's giving them a measly half billion dollars? They're the fastest cheapest thing on the market. If you don't understand the value of that, you really don't understand AI.
Are stagnating models really a problem? Because if so, then how is "if we stop training we are profitable" an acceptable answer to the profitability question?
what's the confusion. Groq offers a fast inference solution that is currently unique (why do you think Nvidia paid $8 billion to end-run around the SEC to acquire the technology). This is good news as it ensures that Groq customers can can be assured continuity to use their service.
You pay them money to run models. Their website doesn’t list them as offering any models that were released recently. For a “pay for inference” provider, questions like “do they have Deepseek or Qwen 2.6?” are germane
I’m confused by the confusion. Groq licensed their technology (sold part of their business) to Nvidia for a large amount of money and distributed the spoils to their investors. Seems quite normal? But then the Axios article says…
“Existing shareholders will receive the remaining cash distributions and then have the opportunity to invest into a new company”
New company? But Groq still exists and continued to exist.
“The bottom line: Don't be surprised if this becomes a new transaction template in the AI private markets.”
A transaction template? I don’t follow what was novel about this situation. The Meta not-acquisition-acquisition of Scale seems more novel.
I guess I feel like Zach’s confusion is because of the way Axios has presented what is happening to Groq. Looking at why actually happened with Groq, it seems like Axios are reporting it weird.
Unless Groq really is starting a new company in which case I am equally as confused.
The interesting thing here isn't "how, logistically, is the Groq corporate entity able to raise more money?". That's straightforward.
Rather, the interesting thing and the topic of most of the article is "how, after Nvidia hired most of Groq's team and licensed all their IP, did Groq manage to convince investors to invest in the remaining corporate entity?"
There's nothing normal at all about the Nvidia Groq deal, it's hard to read in terms of what it means. A straight licensing deal would have been easier to ingest.
I could be completely off the mark but I thought the non-exclusive license was necessary because Groq’s datacenter business uses the technology already? Nvidia acquired the assets but Groq needed to retain rights to use the technology for their own product.
Please don't post shallow-indignant comments. The article raises an interesting question. If the discussion terminates in an angry cliché before it even gets started, that's a boring outcome.
Well dang, that's sort of my goal. I think the masses should be angry and should be highly polarized (regardless if its cliche) against the groups of people investing billions in chat bots while millions can't afford food and medicine.
I’ll also agree with you, this is not a topic that needs nuance, it’s as straightforward as fuck the billionaire class.
If you want nuance, the obvious answer to this is that the rules that apply at our level do not apply to them. Raising money is an inevitability and does not require any fundamental basis other than the name behind it.
The community support forum is also getting retired and there haven't been any posts by support employees in forever anyway, so they are probably gone, too. Also, the number of issues have been piling up, suggesting that the developers are gone as well. https://community.groq.com/c/forum/4 (archive link for when it goes down https://web.archive.org/web/20260602064050/https://community...)
To me, it looks they are trying to raise 650M with a few remaining (ancient) LPUs and no employees.
Plausibly, take all the Nvidia hype and multiply that by a factor and that's what 'Groq' could be worth.
And there is no real commodification - there's Nvidia, Cerebras, Groq ... not many otheres.
Was this comment created using quantized llama 3?
I love Groq, but across every single line break in your post there is a glaring issue that is easy to refute with in 15 seconds, even without 300t/s of throughput.
Groq is more performant for the growing categories of inference-based tasks, wherein Nvidia's advantage in inference depends bulk/batch processing which will make up a smaller category over time, in relative terms.
The future of AI Silicon is inference, and the cost structure of AI data centres is constrained around the current necessity to have 'high GPU utilization' otherwise, the cost / amortization of the chips doesn't work out.
That cost structure is a limitation of Nvidia architecture.
Groq serves a lot faster, and without the limiting batching requirement, which opens hosting arrangements common in most classical hosting scenarios aka without necessarily the high utilization requirements.
Groq has bespoke hardware, lack of CUDA, much lower memory desnsity obviously and they don't have the deep distribution networks and leverage over TSMC that Nvidia has - but pound for pound, were we to be able to 'fire up a server' for our inference needs, it would be Groq, not Nvidia that we'd turn to.
Were they not a later market entrant and didn't have those barriers to entry, they'd be gigantic.
They're not really competing with Nvidia because 1) Nvidia owns their chips now, and 2) Nvidia is not really an inference provider.
Nvidia doesn't own them or all their IP now, we don't quite know the terms of the deal.
Groq lasted how long in the social contract experiment? Libertarians always forget that rules and government aren’t tyrannical they are the mechanism to ensure bullies don’t destroy everyone’s freedom.
The new designs were their main asset besides the amazing talent that went to NVIDIA, not the remaining DCs.
> Groq, the AI chip company company that was acquired by Nvidia in December of last year, is raising $650M
? Could you provide details?
From the actual Groq PR release:
"Groq announced that it has entered into a non-exclusive licensing agreement with Nvidia" - "As part of this agreement, Jonathan Ross, Groq’s Founder, Sunny Madra, Groq’s President, and other members of the Groq team will join Nvidia to help advance and scale the licensed technology" - "Groq will continue to operate as an independent company with Simon Edwards stepping into the role of Chief Executive Officer" (https://groq.com/newsroom/groq-and-nvidia-enter-non-exclusiv...)
Nothing about an acquisition there. It says Nvidia is licensing it, and that others can too. The execs work for Nvidia to integrate it into Nvidia's... something. And Groq the company remains the same as before.
There's also no official source for the amount Nvidia paid for the tech, or two unofficial ones. Journalistically speaking, this is some bullshit.
Why's the deal like this? No idea. Does it make sense? No idea. But it's not odd that Groq is continuing to raise more money, because they never stopped being a normally operating company.
If you want an explanation for why Nvidia would do this deal, my best offer is here (https://openrouter.ai/rankings): Of the top 10 fastest AI models, Groq is the provider of 4 of them. And of the price of those top 10 fastest AI models, Groq is #1, #2, #3, and #5. And you wonder why someone's giving them a measly half billion dollars? They're the fastest cheapest thing on the market. If you don't understand the value of that, you really don't understand AI.
-- allam-2-7b
-- canopylabs/orpheus-arabic-saudi
-- canopylabs/orpheus-v1-english
-- groq/compound
-- groq/compound-mini
-- llama-3.1-8b-instant
-- llama-3.3-70b-versatile
-- meta-llama/llama-4-scout-17b-16e-instruct
-- meta-llama/llama-prompt-guard-2-22m
-- meta-llama/llama-prompt-guard-2-86m
-- openai/gpt-oss-120b
-- openai/gpt-oss-20b
-- openai/gpt-oss-safeguard-20b
-- qwen/qwen3-32b
-- whisper-large-v3
-- whisper-large-v3-turbo
“Existing shareholders will receive the remaining cash distributions and then have the opportunity to invest into a new company”
New company? But Groq still exists and continued to exist.
“The bottom line: Don't be surprised if this becomes a new transaction template in the AI private markets.”
A transaction template? I don’t follow what was novel about this situation. The Meta not-acquisition-acquisition of Scale seems more novel.
I guess I feel like Zach’s confusion is because of the way Axios has presented what is happening to Groq. Looking at why actually happened with Groq, it seems like Axios are reporting it weird.
Unless Groq really is starting a new company in which case I am equally as confused.
edit: when announced last year it was announced as an asset acquisition https://www.cnbc.com/2025/12/24/nvidia-buying-ai-chip-startu...
Rather, the interesting thing and the topic of most of the article is "how, after Nvidia hired most of Groq's team and licensed all their IP, did Groq manage to convince investors to invest in the remaining corporate entity?"
“One could argue that Groq’s datacenters alone could make them worth billions of dollars.”
Groq is a successful datacenter business with a high-revenue cloud product. That’s a compelling investment in its own right, right?
https://groq.com/newsroom/groq-launches-european-data-center...
That sounds like they are renting racks in a Equinix data centre. Do Groq have 4 data centers worth billions?
These 'we get your executives' type of deals - aka Windsurf - are new, weird thing in M&A.
If you want nuance, the obvious answer to this is that the rules that apply at our level do not apply to them. Raising money is an inevitability and does not require any fundamental basis other than the name behind it.