高盛认为NBIS因强劲交易而具备50%的上涨空间，但需警惕存款驱动现金流表象及定价权可持续性方面的风险。

That's great. If you can sign deals like that regularly and have a sub-two-year payback for all the hardware, really, really good. That's anti-AI bubble narrative. That's pro chip build-out. That's pro Nvidia. That's like just good. That's the easiest way to put it. It's good.

So, Nebius signing contracts above 20 billion is pretty good. But, it does also reiterate that in the long term this 30 to 50 might not be mega sustainable, but we'll keep an eye on it because there is talk that Nevius per Goldman is also quote-unquote negotiating shorter-term agreements in the 40 to 50 billion dollar or million per megawatt range, which should serve as a meaningful top-line contributor in 2027 beyond.

Nevius obviously a much smaller company than SpaceX or you know, even Google. Google, for example, raised 86 billion dollars in the last few months. SpaceX raised 80 billion dollars. Nevius has like 8 billion dollars of cash, right? Like and and the size of the company is is less than 80 billion dollars. It's like a 50 billion dollar company. So, much much smaller company and player compared to these big guys, but it's good to see this for what's that what's pricing doing on the edge, right?

And this will help with the price target for Nevius from Goldman Sachs. So, preparations are underway for materially higher deployment in 2027. 2027 capacity, some of it is being deliberately retained rather than contracting it today reflecting management's confidence in future pricing.

I actually don't like that idea. I don't like the idea that they're holding back inventory on purpose to squeeze pricing up. So, it's sort of like artificially holding back supply to squeeze pricing up. I get it for wanting to lock in higher earnings driving contracts, but I think that's a misleading spot price that you're negotiating because you're artificially holding supply that you potentially have. And we're not actually getting a real market clearing of what those spot prices should be or what we should expect those longer term contracts to actually run at. So, I don't love that as much.

Uh in addition to that, they have secured access to hundreds of megawatts of behind the meter power generation. This is really important so you can get faster permits for your data centers.

But anyway, they're expanding uh and adding to their CPU um fleet as well for agentic workloads. This is kind of the range that I think I I wrote this earlier. I think if you get sub eight billion dollars per gigawatt, things get scary. If you start falling below that, signing at 20 is great, especially since I think mostly we're at like 14 to 15 on longer contracts. Uh and if you could get anywhere over 30, that's fantastic. If you could get to 50, scaled up, not shortage pricing, that's ridiculous. But that probably requires AGI and robotics, but it would be such a boon for all of these data center providers, SpaceX, Nebia's, uh you name it.

Uh Goldman Sachs here suggests that their operating cash flow is doing well. I wrote probably because of deposits. It turns out just the very next line Goldman then says that 70% of deals closed in Q2 included upfront payments. I do think that this is a little bit of a risk including $9 billion of upfront funding because when you actually go into the cashflow statements for these companies, they could look really exciting uh because you're like, "Oh my gosh, there's just there's so much cash coming in. This is great. Look at this. We've got $4.5 billion of cash over here. Wow. Yeah, we spent $8 billion on hardware but we brought in 4.5. Yeah, well 4.4 of that was upfront deposits, deferred revenue, right? So we have to be careful that in the future those deposits won't come in so those numbers won't look as good. And so as long as you're aware of that then I think that's okay. That's important.

A lot of these also have pretty high cancellation costs. So you know, it's unlikely that uh companies really want to walk away from the secure uh compute that they've secured just because the cancellation costs are probably over 50% but we're not sure. There's not a lot of color on exactly what all of these cancellation costs are for all these various different companies.

Uh but I do think that Nebius has pricing power and how they can handle their cancellation clauses. They don't want to get left holding the bag although they still might. Right? So it depends. It all depends on that number. What's that megawatt number, right? The millions per megawatts or billions per gigawatts.

Uh selling out compute as it comes online, we see uh pricing remains constructive for Blackwell GPUs. Pricing's actually been moving up uh quite a bit.

So, Goldman Sachs, their price target is $328. So, they see a 50% upside in the stock. I will say if I look at their balance sheet, their balance sheet is actually pretty decent. They've got $8 billion of cash. They've got $1.4 billion of bills and they have long-term debt of about eight. It's yellowish, right? Like it's not super perfect, but it's not like Broadcom or Oracle where you're like, "Oh my gosh, that's a lot of debt." Or worse like UPS, right?

Okay, then if I look at this projected yield by 2030. So, this is sort of like a forward pricing model. I get to a you know, projected future yield of about 13.9 on a PE basis. And that's 2030. That's mostly because we don't expect them to be profitable until 2028. So, you kind of have to use some forward numbers to really price this. Goal again is to be profitable in 2028 and and there's some significant growth expected between now and the end of the decade if they can go profitable.

So, biggest risk factors in my opinion profitability and that the markets get misled by upfront deposits. You know, they get confused by that. And the balance sheet is okay. And I'm most concerned in the sustainability of these negotiations. Even at 20 million per megawatt, that's great, but once we really start scaling, are we going to be able to maintain that pricing? I hope so. I hope this is more sustainable than just a temporary bounce that we're seeing because of a genetic coding and once we get caught up with that, we go back to trend. Because if we go back to trend, all of these prices have to come down 20% at least. Going to take you to about Yeah, probably about 20% reduction in GPU per hour pricing that's going to extend payback periods. These are some of the things to watch for, but that's an overall breakdown on Nevius and why there's about a $330 price target from Goldman for this as well as some of the risk factors to consider when you're looking at the stock.