Thursday, November 13th 2025
NVIDIA Moves to In-House Server Production, Leaving OEMs with a Secondary Role
NVIDIA plans to reshape the industry's approach to building high-performance AI servers with its "Vera Rubin" product stack, which will enter volume production as the VR200 line in late 2026. Rather than supplying CPUs and GPUs to OEMs, NVIDIA plans to ship fully finished L10 compute trays that come pre-populated with "Vera" CPUs, "Rubin" accelerators, memory, 800G NICs, 110kW power delivery, and liquid-cooling infrastructure, all of which are validated before they leave the factory. This, however, will leave many system vendors and OEMs shifting away from designing core server electronics and toward tasks such as rack-level assembly, power configuration, installation of rack cooling sidecars, and final certification.
Rubin silicon is rumored to have a TDP of up to 2.3 kW for highest-end models, leading to rack-level power consumption exceeding 250 kW and making custom cooling designs economically unviable. NVIDIA's strategy reportedly involves standardizing cooling and power across two performance tiers while increasing network I/O density. These choices decrease the feasibility and cost-effectiveness of third-party replication. By centralizing production with select partners, development cycles can be shortened and manufacturing yields improved. However, this approach also limits the options OEMs and hyperscalers have traditionally used to differentiate their hardware.The entire supply chain value of the lucrative AI server market will be different, as value and margins are likely to migrate toward certified component makers and high-volume integrators, leaving branded OEMs with narrower opportunities for hardware differentiation and tighter margins. Cooling and power subsystem vendors that secure qualification are set to benefit, while smaller suppliers may find it harder to compete.
Source:
@Jukanlosreve on X
Rubin silicon is rumored to have a TDP of up to 2.3 kW for highest-end models, leading to rack-level power consumption exceeding 250 kW and making custom cooling designs economically unviable. NVIDIA's strategy reportedly involves standardizing cooling and power across two performance tiers while increasing network I/O density. These choices decrease the feasibility and cost-effectiveness of third-party replication. By centralizing production with select partners, development cycles can be shortened and manufacturing yields improved. However, this approach also limits the options OEMs and hyperscalers have traditionally used to differentiate their hardware.The entire supply chain value of the lucrative AI server market will be different, as value and margins are likely to migrate toward certified component makers and high-volume integrators, leaving branded OEMs with narrower opportunities for hardware differentiation and tighter margins. Cooling and power subsystem vendors that secure qualification are set to benefit, while smaller suppliers may find it harder to compete.


30 Comments on NVIDIA Moves to In-House Server Production, Leaving OEMs with a Secondary Role
www.techpowerup.com/342291/amd-sells-zt-systems-data-center-manufacturing-to-sanmina
At this point, Nvidia has everything except the chip manufacturing in-house, while AMD can't decide which parts of the product stack they want to have in-house vs outsource to streamline.
AMD selling off their server production business isn't the same as this news, what Nvidia is doing is vertical monopoly tactics, not good for anyone but their own profits.
They've already sold all the ML hardware needed for the next 10 or 20 years.
Then they could really struggle over the next decade.
It's like tiger outsourcing his work to his fleas.
They wanted to be greedy and wanted to be the only ones selling their GPUs.
the difference is, 3DFX didnt had an inflated 4.5 trillion capitalization.
Which helps a lot with their bribed influencers that make sure to keep the hype going.
In this case though, Nvidia's "cost efficiency" goal is in the same way standard oil vertically integrating offered cost efficiency, by increasing control over a market it has a monopoly over and decreasing competition.
Rest assured, this will not bode well for Nvidia's customers, be they business or individuals.
Again, market cap has no relevance and is NOT in any way connected to companies free cash flow or reserves. In the words you can understand - it doesn’t mean that NV suddenly has more money to “bribe influencers”. This is not how it works. I agree that they are overvalued, but this is just how the stock market works - it’s pure speculation all the way down. Let’s say that tomorrow, for whatever reason, NV stock drops 10%. This absolutely does not mean that they are suddenly 450 million poorer.
The more a company tries to do, the less focus it has on it's core products. I don't think it's a good move to take over the building part, because companies like Dell only have an 18% margin and they have decades of experience in that market. It's really not worth it from a resource perspective, it reminds me of Intel when it had it's monopoly over the CPU market. The only angle it makes sense from is Nvidia wanting to exert more control. They clearly aren't doing it to help it's partners or for the no so fat margins.
It's foolish to bet against a company doing something for profit because that's the reason for the existence. Even more so when you are talking about Nvidia, a company that has publically been called out by it's partners multiple times and has even driven some of them out of business. You could not pick a worse company to imply their goal it to help partners. Helping partners would be the implied benefit but it is 1000% not Nvidia's goal. You'd have to be born yesterday to think Nvidia's of all companies is aiming for that.
Nvidia's strength is offering everything through one vendor, which again, if you know anything about this industry, is an incredible advantage. Instead of writing your own code for your GPUs, you can just use CUDA. Instead of wrangling with Intel or Mediatek networking, you use Nvidia. Instead of dealing with a seperate CPU vendor, you just use Nvidia. And now, instead of having to reach out to an OEM for help with your server, you just ask your Nvidia rep.You don't understand basic economics? If something is better for Nvidia's customers, Nvidia makes more money. That's how a free market works. Exchanging money for goods and services means both parties end up better off.
We don't have a free market, we and most other countries have a mixed market. There's a reason for that, pure free markets don't work. You cannot create a system where supply and demand are the sole determiners of prices with no undue influence (as anyone who's lived in the last 60 years of capitalism would know). That kind of market is equivalent to anarchy, wherein the most powerful players will always gain and control the most influence due to the complete lack of any moderating forces. Government influence is prohibited in a pure free market.
Mind you, it's extra funny that you would imply that Nvidia of all companies is trying to encourage a free and fair market, particularly with this move that furthers their grip over the market. There are dozens of papers written on the impact of vertical integration with America being the top case study of the negative effects hitting every part of the economy: rooseveltinstitute.org/wp-content/uploads/2020/07/RI-Vertical-Integration-and-Market-Power-Crisis-Issue-brief-201904.pdf#:~:text=While the increase in horizontal concentration across,effects. If anything, the reverse is true.
You certainly are trying your best to side-step the fact that Nvidia, like all publically traded companies, exists to make money, they have a long history of being anti-competitive, and they are in a more dominant position than most past monopolies at their height. They also happen to be in an extremely lax regulatory environment that's taken basically no meaningful action in the last several decades to stop mass mergers, acquisitions, and vertical integration that have fixed power in an increasingly smaller amount of hands. No amount of insults will change those facts.Again, I'm not talking about the industry now, but how it got here. None of that vertical integration is relevant if it wasn't for the GPU hardware they built the company on. You are putting the cart before the horse and saying the cart is what got us here when that's not how it works.
There's a reason you see companies increasingly vertically integrate as they have grown. Walmart, Amazon, Google, Meta. They've all done it. It's not something a company does until it has a successful business, usually. It very frequently doesn't work in the customer's favor either.
Things can turn around just like that.I disagree. Verticals have been a thing in corporate businesses for a while now and what I see is that companies are in fact created to dominate an entire value chain. Screw the monopoly, you can escape that reality by just getting the entire chain in such a tight control that you offer the best product, in ways no competitor can offer it.
And if you look at Nvidia as a company doing their Geforce line up, the same thing has occurred. Their FE is the result of that same line of thought, the exit of EVGA is another. Everyone else still selling these products are just shoving boxes for a minor margin. It has nothing to do with their 'current success'. Its a defined business approach that has gained popularity.
Mind you, the idea that it's every company's prerogative to dominate their value chain would be an effect of every company's goal to make as much money as possible, which I what I posited earlier. You are in essence agreeing with me here but stating the symptom and not the root cause for that desire. The root of which is much more important, as it's not in every company's best interest to dominate the value chain. It depends on the industry but more important than all, it requires a massive amount of capital. It's a warped modern perception to think every company MUST dominate their value chain, only in a world full of mega-corps does such a perception make sense. For small to medium companies it almost never makes sense. Even for large companies, they should be weighing the pros and cons. Most of all, the negative effects of mass vertical integration across the broader markets are well noted in the paper I linked earlier in addition to several other's readily available on the internet. One need merely look at the US economy and loss of competition across the board to see the impacts. Customer choice is robbed, competition stifled, and innovation greatly hampered among other negative effects.