Oracle does not have any proprietary IP that OpenAI depends on. That means Oracle is providing commodity services to OpenAI.<p>AI hardware is advancing rapidly. If NVIDIA's next generations deliver radically more tokens per watt and per dollar, an Oracle facility designed, financed, and contracted around an older GPU generation can be overtaken before it is fully online or fully monetized.<p>That's a big risk.<p>And to finance this they are making deep cuts to the ERP business where Oracle does have proprietary IP.
> If NVIDIA's next generations deliver radically more tokens per watt and per dollar,<p>> an Oracle facility designed, financed, and contracted around an older GPU<p>> generation can be overtaken before it is fully online or fully monetized.<p>Every user of Nvidia's gear has thought of this, and Nvidia provides a solution: Newer components are (usually) compatible with older racks, and Nvidia will buy back parts you replace with newer hardware.<p>For example, let's say you have an NV72 with Blackwell-ultra. Rubin comes out, and you decide you want to upgrade. There is a SKU of Rubin designed specifically to work in the racks shipped with Blackwell. Use it, and you only need to replace the compute board (not the power system, cooling, networking, etc.)<p>Nvidia will buy back the Blackwell boards you remove (and sell them to someone else, with a warrantee).<p>If you want to do an upgrade that increases power use, you may have to get a new rack. All of the parts are modular, and all parts can be sold back to Nvidia.<p>Nvidia does this because they have to. Without it, many customers would consider other vendors when upgrading, or delay purchases in case something better comes along later.
I would be very very curious in some source of these Nvidia buybacks
What does Nvidia do with the parts they buy back? Resell to others or grind to dust?
Forward them to their, partner Microsoft…
resell them? there is a decent market for people fine with less efficient, already used gear.
I doubt NVidia wants thriving second hand market, so I would guess grind to dust?
I'm not really sure they have thought this through. Or at least their solution has been to essentially insulate the company and even the physical building from the debt used to buy the GPUs. These SPVs typically lease the building and secure the debt to the GPUs themselves so if it all goes pear-shaped, it'll be the debtholders who are left holding the bag, priamrily.<p>But I have questions:<p>1. How often do GPUs fail? I've seen figures that this is as high as 9% annual failure rate for DC GPUs;<p>2. How easy is it to get replacement stock? At what cost?<p>3. If you trade in, what value do you get for your GPUs?<p>4. What is the availability like? How does NVidia decide who gets the next wave of GPUs? That's NVidia's big power here. It's a bit like how Boeing and Airbus decide on customer order and if you upset them, you may lose other orders. This happened to Qatar Airways over the A380 [1].<p>5. How long are the current form factors and interfaces going to last? How long before the power requirements get larger than what DCs are currently built for? Blackwell GPUs use a lot more power than, say, 2 or even 1 generation ago did.<p>GPUs are a huge cost factor, allegedly more than 50% of the cost of a data center [2]. If you take a bath on the trade in and are forced to do it because you can't compete with the current generation's performance-per-watt, that's not a good situation to be in.<p>[1]: <a href="https://www.arabnews.com/business/qatar-airways-in-mystery-airbus-row-1867761" rel="nofollow">https://www.arabnews.com/business/qatar-airways-in-mystery-a...</a><p>[2]: <a href="https://orennia.com/insights/what-it-actually-costs-to-build-a-1-gw-data-center" rel="nofollow">https://orennia.com/insights/what-it-actually-costs-to-build...</a>
Sometimes I wonder if the people committing and spending this much money is really thinking about this or not?<p>My feeling is there is some kind of upgrade packaging done here. Like if the chips advance by X factor it can be offered at a massive discount or something? Otherwise it seems so silly to buy so much hardware which is evolving so fast.
I'm no fan of Oracle, but you could make the same claim about land developers, or datacenter operators, or REITs. Or a random conglomerate like Unileaver or GE or Siemens.<p>Ultimately, someone has to provide "commodities", and if you think your ship is the right vessel to do it betting hard on this is some kind of strategic direction.<p>Whether they are right, and to what extent debtors and the market will put up with cyclic and performance issues, etc.. I certainly won't hold the stock, but I wouldn't have earlier either.
This sounds a lot like the play that SpaceX made for their IPO.
ERPs are done. Data center permits, space, contracts to buy scarse GPUs and energy are the moats today. The chips can provide more tokers per watt but the world needs exponentially more tokens each day
Curious to know why you think ERPs are done. I actually do somewhat agree with you in that their best days are behind them.<p>It is going to be incredibly difficult to bill outrageous amounts for armies of consultant to implement or customise the product when agents can do the same for much less, or at-least a 5-10x reduction in headcount for similar complexity of changeset.<p>If anyone is in the industry, please feel free to chime in as to what you are seeing on the ground.
>Oracle does not have any proprietary IP that OpenAI depends on. That means Oracle is providing commodity services to OpenAI.<p>Huh? Is Oracle's plan with this funding to provide compute services for the big AI specialist companies? I assumed it was a buildout for their own AI services, as you're right, the former wouldn't make sense. (At risk of making a trollish, low-effort comment, OpenAI/Anthropic are savvy compute users and would have to be insane to risk getting entangled with Oracle.)<p>I read the article and I couldn't tell from it which of these is the case.
On strictly business terms, you're absolutely correct. Considering that OpenAI and Oracle both eat from Trump's hand though, I think that Larry Ellison has a few tricks up his sleeve to remain a "necessary" contractor. This is probably a shorter-term investment to pad out the coffers 5-10 years from now.<p>If rampant despotism is the end-goal, Oracle has already aligned themselves with the despots.