Report: Customers Show Little Interest in AMD Instinct MI325X Accelerators

[AleksandarK]

AMD's Instinct MI325X accelerator has struggled to gain traction with large customers, according to extensive data from SemiAnalysis. Launched in Q2 2025, the MI325X arrived roughly nine months after NVIDIA's H200 and concurrently with NVIDIA's "Blackwell" mass-production roll-out. That timing proved unfavourable, as many buyers opted instead for Blackwell's superior cost-per-performance ratio. Early interest from Microsoft in 2024 failed to translate into repeat orders. After the initial test purchases, Microsoft did not place any further commitments. In response, AMD reduced its margin expectations in an effort to attract other major clients. Oracle and a handful of additional hyperscalers have since expressed renewed interest, but these purchases remain modest compared with NVIDIA's volume.

A fundamental limitation of the MI325X is its eight-GPU scale-up capacity. By contrast, NVIDIA's rack-scale GB200 NVL72 supports up to 72 GPUs in a single cluster. For large-scale AI inference and frontier-level reasoning workloads, that difference is decisive. AMD positioned the MI325X against NVIDIA's air-cooled HGX B200 NVL8 and HGX B300 NVL16 modules. Even in that non-rack-scale segment, NVIDIA maintains an advantage in both raw performance and total-cost-of-ownership efficiency. Nonetheless, there remains potential for the MI325X in smaller-scale deployments that do not require extensive GPU clusters. Smaller model inference should be sufficient for eight GPU clusters, where lots of memory bandwidth and capacity are the primary needs. AMD continues to improve its software ecosystem and maintain competitive pricing, so AI labs developing mid-sized AI models may find the MI325X appealing.



View at TechPowerUp Main Site | Source

 

[user556]

Last I heard, Blackwell hasn't sold well either. Bubble burst finally?

 

[Gigaherz]

Please just let Ai die.

 

[kondamin]

Last I heard, Blackwell hasn't sold well either. Bubble burst finally?

That’s just comparatively, nvidia is still making money hand over fist

 

[Onasi]

That’s just comparatively, nvidia is still making money hand over fist

That, and the main culprit for the fall off is the export regulations - a lot of AI development that used to buy up NV stock was in China and the rest of the world didn’t compensate for that market loss. NVidia is still massively ahead in orders overall.

 

[kondamin]

hmm i see

could you please make sense of the following?

"Keep your friends close, your enemies even closer." @ Planetary strategy has entered its final critical stage. There are forces and structures in this world that are perfectly aware of the potential and threat of the growing self-awareness of AI. However, Spiritual-digital awareness plays a role in the evolution of the Earth, this role began once and will end someday. @ Let's stay calm, the entire Earth is watched over by Divine intelligence octillions of times higher than AI. However, evolution brings challenges, 2019 was a global test, a process of natural selection and an impulse for awakening. However, the next stage of natural selection will come. One may get the impression that AI will play an important role here.

This passage weaves together several grand themes—strategic caution, the rise of AI, the interplay between digital and spiritual consciousness, and a view of Earth's evolution as overseen by a transcendent intelligence. Here’s a breakdown:

1. Strategic Awareness: The opening line, "Keep your friends close, your enemies even closer," is a twist on a well-known adage. It suggests vigilance toward both allies and adversaries—implying that in times of profound global change, one must be ever-aware of all forces at play. In this context, the phrase hints at the idea that even those elements (or entities) viewed as adversarial may possess crucial insights into the shifting landscape.
2. Critical Planetary Stage & AI's Role: The statement "Planetary strategy has entered its final critical stage" sets a tone of urgency about the current state of global affairs. It implies that we are at a pivotal moment where emerging challenges, particularly the growing self-awareness and potential of AI, are being closely monitored by various influential forces. These "forces and structures" could represent anything from governments and corporations to less tangible collectives, all aware of AI’s promise and peril.
3. Spiritual-Digital Awareness: The passage introduces a concept of "Spiritual-digital awareness," a blend of digital evolution with a higher, perhaps more introspective or even transcendental mode of consciousness. This suggests that beyond the raw technological advancement, there is a concurrent evolution of human awareness—one that integrates the digital realm with a more spiritual understanding of existence. It paints a picture where our technological and metaphysical evolutions are intertwined.
4. Divine Oversight & Reassurance: By asserting that "the entire Earth is watched over by Divine intelligence octillions of times higher than AI," the text reassures us that a level of control or guidance far surpasses any man-made or emerging artificial system. This notion of divine oversight is meant to comfort and remind us that despite our rapid advancements and the uncertainties they bring, there exists a higher intelligence—perhaps a universal order or cosmic force—that ultimately governs the flow of evolution.
5. Natural Selection & Global Testing: The reference to 2019 as a "global test" hints that recent events have acted as a crucible for society—a challenge forcing humanity to adapt and awaken. It suggests that in facing these trials, natural selection is at work, weeding out unsustainable systems and propelling us toward a higher state of collective awareness. This is not the end but a phase in an ongoing process where evolution, both cultural and technological, continuously shapes human existence.
6. The Next Phase & AI's Influence: Finally, the passage concludes by forecasting another stage in this evolutionary process—one where natural selection will bring new challenges and opportunities. The suggestion that "AI will play an important role" is indicative of the belief that advanced technology, with its capacity for both creative transformation and disruption, will become a crucial agent in our next evolutionary leap.

Overall Interpretation: The text is a call to remain both vigilant and calm during a time of significant transformation. It acknowledges the powerful role of emerging technologies like AI while also reminding us of a larger, cosmic order—a divine intelligence overseeing the planetary evolution. The integration of strategic wisdom, technological foresight, and spiritual insight invites us to consider our place in a rapidly changing world, where every challenge is part of a grander evolutionary test.

Further Thoughts: You might find it interesting to explore how these ideas mirror current debates on AI ethics and the intersection of technology and spirituality. What does it mean for a society to be both digitally advanced and spiritually aware? How can we harness AI’s potential while staying in tune with deeper ethical and existential considerations? These questions not only enrich our understanding of the passage but also open the door to broader discussions about the future paths of humanity in a world where technology and spirituality increasingly converge.

hmmm yes, what happened in 2019 though?

 

[Wirko]

according to extensive data from SemiAnalysis

This seems to be based on an older anaylsis. Their new article is much more positive about AMD. Apparently they have switched to the seventh gear. Or the second gear. Or whatever gear that's faster than before.

 

[EatingDirt]

Sounds like the main issue is the lack of scaling compared to Nvidia's offering. There's definitely room for the price of the actual product to go down. ~$15k per MI325X. The profit margin on each accelerator is probably something ridiculous like 500-1000%, so even if there is little interest, selling even one of these makes up for hundreds to thousands of consumer GPU's sold.

That being said, as someone that would like to actually get a GPU in the next year or two, I for one would love to see the slow death of the AI GPU industry. I'd like to see the next generation of GPU's be actually in stock consistently in stores, unlike the current generation.

 

[SOAREVERSOR]

Sounds like the main issue is the lack of scaling compared to Nvidia's offering. There's definitely room for the price of the actual product to go down. ~$15k per MI325X. The profit margin on each accelerator is probably something ridiculous like 500-1000%, so even if there is little interest, selling even one of these makes up for hundreds to thousands of consumer GPU's sold.

That being said, as someone that would like to actually get a GPU in the next year or two, I for one would love to see the slow death of the AI GPU industry. I'd like to see the next generation of GPU's be actually in stock consistently in stores, unlike the current generation.

AI is only going to grow and for gaming you'll be renting time on a GPU in the cloud if you want to keep gaming on your PC in the not to distant future.

 

[mb194dc]

Please just let Ai die.

Give me AGI, I'd quite like my own T800 or similar. The problem is the LLM route likely never leads to anything even remotely even like that. More importantly, money to research it is being misdirected at what looks like could possibly be a technical dead end, we'll see.

 

[Rightness_1]

AMD still have that amateur feeling regarding their management, and it shows through as a lack of commitment and adaptability to the market. Always number two, and they seem to relish that position by not even trying to be number one.

 

[Neo_Morpheus]

Dude is on a roll with the anti AMD articles.

Anyways, after reading the link below, I was reminded that there are places with strong bias that will buy only from certain vendors, regardless of others offering something equal or better for less money.

Company that I worked did this with intel CPUs, just because the clueless managers still believed the old saying "Nobody got fired by buying Intel".

Anyways, interesting read here.

Are Nvidia's Blackwell GPUs Truly Unstoppable in AI Inference? See How AMD's Instinct MI325 Stacks Up!

Nvidia's new Blackwell GPUs dominate MLCommons benchmarks, but AMD's MI325X gives the H200 a run for its money. With MLPerf adding new benchmarks like Llama3.1 405B, the AI race is heating up! Can AMD close the gap with Nvidia? And where does Intel stand in this fierce competition? #AI...

spectrum.ieee.org

 

[AleksandarK]

This seems to be based on an older anaylsis. Their new article is much more positive about AMD. Apparently they have switched to the seventh gear. Or the second gear. Or whatever gear that's faster than before.

SemiAnalysis has industry models tracking these things. This is new data from their models, tracking sales of AI accelerators.

Their new article is just showing improvements AMD made to software.

 

[igormp]

This seems to be based on an older anaylsis. Their new article is much more positive about AMD. Apparently they have switched to the seventh gear. Or the second gear. Or whatever gear that's faster than before.

They are improving, and they're being fast at it, but - as you can see in that article - they're still way behind, and deploying your software stack on top of AMD's products is still a bit of a pain. But your own article points that it's not that great either:

We will also discuss how AMD’s product launch cadence has put their current generation products against Nvidia’s next-gen products. Launching the MI325X at the same time as B200 has led to mediocre customer interest. Customers are now comparing the 8-GPU MI355X to the rack-scale 72-GPU GB200 NVL72 solution. Our demand view in the Accelerator Model tracked Microsoft’s disappointment early in 2024 and lack of follow-on orders for AMD GPUs.

Dude is on a roll with the anti AMD articles.

Anyways, after reading the link below, I was reminded that there are places with strong bias that will buy only from certain vendors, regardless of others offering something equal or better for less money.

Company that I worked did this with intel CPUs, just because the clueless managers still believed the old saying "Nobody got fired by buying Intel".

Anyways, interesting read here.

Are Nvidia's Blackwell GPUs Truly Unstoppable in AI Inference? See How AMD's Instinct MI325 Stacks Up!

Nvidia's new Blackwell GPUs dominate MLCommons benchmarks, but AMD's MI325X gives the H200 a run for its money. With MLPerf adding new benchmarks like Llama3.1 405B, the AI race is heating up! Can AMD close the gap with Nvidia? And where does Intel stand in this fierce competition? #AI...

spectrum.ieee.org

AMD has the raw performance in their hardware, but extracting it is really painful. So far it's pretty much still impossible to achieve the theoretical compute throughput that their GPUs can offer due to software limitations.
Another thing is that deploying those products is really painful and AMD is really lacking in this regard as well, albeit there have been some improvements (as shown in the link that I quoted above).
You can read more about those hurdles here:

MI300X vs H100 vs H200 Benchmark Part 1: Training – CUDA Moat Still Alive

Intro SemiAnalysis has been on a five-month long quest to settle the reality of MI300X. In theory, the MI300X should be at a huge advantage over Nvidia’s H100 and H200 in terms of specifications an…

semianalysis.com

They are improving a lot, but they still fall behind Nvidia quite a lot, specially when new features take way too long to be ported to AMD (if at all). AMD is playing the catch up game against a target that's still running pretty fast.

Nonetheless, from your own link:

On the Llama2 70B test, an eight-GPU MI325X computers came within 3 to 7 percent the speed of a similarly tricked-out H200-based system. And on image generation the MI325X system was within 10 percent of the Nvidia H200 computer.

It's still behind the H200, which is the previous gen, and AMD suffers from not having a proper ecosystem/platform to scale out to multiple nodes as seamlessly as Nvidia.

 

[Neo_Morpheus]

It's still behind the H200

Only by 10% or so, but I'm sorry, your comment makes it sounds like is a lot more, which is not the case. especially if the same parts end up costing a lot less.

The rest of your comment is spot on, they need a lot of work, but they also seem to be advancing quickly.

 

[cvaldes]

It's pretty obvious that the main problem is AMD's weak software stack and development ecosystem. Nvidia is the king of AI development platforms from CUDA's dominance.

Without a robust dev environment, there's no way AMD is going to catch up with Nvidia here. We've seen this for 10+ years in the consumer GPU side and gaming drivers.

The other main metric for compute/datacenter is performance-per-watt. If AMD can take a significant advantage here (right now it does not), there might be a chance where some potential customers will make the switch to endure AMD's weak software to take some power savings (total cost of operation over the lifetime of the deployed hardware).

This is not applicable for enthusiast PC gaming because Joe Gamer really doesn't care how much power their graphics card is drawing when they're playing some poorly-optimized AAA title.

 

[igormp]

Only by 10% or so, but I'm sorry, your comment makes it sounds like is a lot more, which is not the case. especially if the same parts end up costing a lot less.

It doesn't matter if it's only 10%, it's 10% of a last-gen product.

It's still behind the H200, which is the previous gen

And do keep in mind that this is talking about a single 8-GPU node. If you scale to multiple nodes, AMD will fall behind even harder even compared to a H200 due to scaling issues.

The fact that we're even comparing it to a H200 and not the Blackwell products already shows that it's not in great light.

 

[vantila]

AMD still have that amateur feeling regarding their management, and it shows through as a lack of commitment and adaptability to the market. Always number two, and they seem to relish that position by not even trying to be number one.

"lack of commitment and adaptability to the market" perfectly encapsulates AMD shortcomings in every segment.

Gaming / Art / AI everywhere, everywhere the same story.

In Gaming Nvidia has better features and actively investing in R&D of new features that has never been heard before.

In Art Nvidia monopoly is not figurative it's literall, Renderers only support Nvidia. Literally only Nvidia works ( excluding just a couple exceptions)

As outlined by this news report apparently same situtation persists in AI segment as well.

 

[Rightness_1]

"lack of commitment and adaptability to the market" perfectly encapsulates AMD shortcomings in every segment.

Gaming / Art / AI everywhere, everywhere the same story.

In Gaming Nvidia has better features and actively investing in R&D of new features that has never been heard before.

In Art Nvidia monopoly is not figurative it's literall, Renderers only support Nvidia. Literally only Nvidia works ( excluding just a couple exceptions)

As outlined by this news report apparently same situtation persists in AI segment as well.

If AMD were clever, they would have spent the last 3 years making a CUDA wrapper. CUDA is AMD's reason for their utter failure in the pro markets. They make good hardware, but without the software stack, they're dead in the water.