$nbis 2025 q2 er

$NBIS just reported 2025 Q2 earning:

  • Revenue was $105.1 million, up 322% YoY and 90% QoQ. This was extraordinarily good, maybe the best growth story I can find in the entire market. This quarter’s revenue number was even after excluding Toloka’s contribution because it was just sold in the quarter.
  • Raised full year guidance of ending monthly ARR from $750 million ~ $1 billion to $900 million ~ $1.1 billion.
  • Clickhouse is a stellar startup and is currently valued over $6 billion. $NBIS owns 28% stake of it.
  • Stills has $1.68 billion cash on balance sheet to burn.
  • Aggressively accelerating capacity expansion. Target 200 MW of power by end of 2025 and >1GW of power in 2026.
  • Announced that Cloudflare and Shopify are their customers.

I’m long at 8% + some call options and I will add more to my position. This company is still significantly undervalued in my opinion.

Link to shareholder letter

Cheers,
Luffy

50 Likes

Hi Luffy,

Thanks for posting. I have, after the pop, a 3% position (I increased it a bit before earnings). In my POV, with these results, they have earned themselves a higher share of my portfolio, and will reinvest some of my HIMS profits into this; the question remains of the entry point.

I don’t think I am comfortable making it a 10-15% position yet, not without another quarter of solid results to extract a clearer trend, but I believe I will make it a 5-8% pos.

Cheers,
Ys

14 Likes

I’m very very happy with these results indeed. I’ve spent the morning (we’re on holiday in France) going through the ER of Nbis, Alab and APP. Wow, on all fronts.

I was hoping Nebius would raise their ARR, and they did.The future looks very bright for them indeed.

Nebius is now 36% of my entire portfolio after yesterday’s 19% pop. (APP is 18% and ALAB is 12%). I can hear Saul very wisely saying not to fall in love with a company, and that he never liked to let positions grow above 20% (I think in 2020 he had Zoom at 30% which was unusual for him), but I still think that Nebius is very undervalued as I’ve been saying all year, so I can’t bring myself to sell any just yet.

YTD return for me is now 67%, significantly more than when I posted my end of July review just 8 days ago.

I’m really enjoying this earnings season.

Best,

Jonathan

35 Likes

I believe the core fundamentals are strong and earnings have reflected that. I believe there is additional value in the portfolio of IP they own. I am long from here and hope to see signs of divesting as that could unlock immediate shareholder value and also demonstrate commitment to the main growth engine.

13 Likes

Their expected growth in revenue is absolutely incredible. However, expected earnings in 2025, 2026, and 2027 are -366m, -486m, and - 414m. There corresponding expected free cash flows are -2245m, - 2029m, and - 3056m. I can understand being on the verge of being profitable, but in this case, 2.5 years down the road, there are still expected to be losing a lot of money. Is the idea that they could be profitable in a second if they wanted to, but they are so interested in growing that they are spending more than they make?

8 Likes

At this early hyper hyper growth stage, the figure to focus on is ARR. They will continue to burn through capex, and will require more funds soon, but as they are able to continue to attract more and more customers (such as hyperscalers, Sovereign AI, as well as companies like Cloudflare and Shopify) it is all about growth right now. Maybe in 3 years we could be looking at 5B+ ARR.

They have other businesses they could use to obtain more funds. And there may be more dilution in the future, but they have so far managed this well and I trust they will continue to do so.

Jonathan

17 Likes

@flyingelephant1 That’s a legit concern about the net income profitability of $NBIS. The reason why they could not be profitable in short-term is because they have to invest in GPUs first at the current stage.

I once found the following rough estimation about Nebius’s potential ROI on GPU investments. It’s based on assumptions from public data, so take it with a grain of salt.

Let’s assume we build a data center in the US with 10,000 H200 GPUs. Cap expense of the hardwares will be $435 million, $43,497 per GPU:

  • GPU: unit price is $30,000, total cost → $300 million
  • servers: OEM HGX H200 (excluding costs of GPU), unit price is $70,000, quantity is 1250, total cost → $87.5 million
  • Switches: Unit price is $35,000, quantity is 156, total cost → $5.5 million
  • Optical Transceivers / Cables: Unit price is $600, quantity is 20,000, total cost → $12 million
  • Storage related cost: $30 million

Let’s assume we also construct the data center rather than lease, the estimated one-time cost will be $146.6 million including 10% on land purchase, 25% on construction costs, 45% on power infrastructure, 20% on cooling system.

So the total capex from data center will be $582 million, or $58,159 per GPU.

Now, let’s turn to Operation Expense of running the data center:

  • We assume Power Usage Effectiveness (PUE) is 1.2 (though Nebius reports 1.13, the lower the better), the power required per server is 10.2kW, total power consumed will be 10.2kW x 1025 x 1.2 = 15,300MW. Assuming a relative low power price of $0.05 / kWh, then the power cost per year is about $6.7 million.
  • Assume cost of IT operations staff salary is $5.2 million per year.
  • Assume a maintenance cost of 2% of the hardware costs, which is $453million * 2% = $8.7 million per year.
  • Total OpEx is $20.6 million per year. OpEx per GPU per hour is $0.24.

Nebius has different pricing tiers, so let’s just assume a rough average price of $2.50 per GPU hour. Then margin per GPU hour is $2.26. And let’s assume the utilization rate of the capacity is 70%. The margin per year will be about $139 million, payback period is about 4.2 year. (I don’t know how long can a GPU’s service life be, but I’m guessing a longer period than this as CoreWeave is using a 6-year schedule to depreciate their GPU assets, though Nebius is using a more conservative 4-year schedule.)

I think Nebius’s payback period will turn out to be much lower than this number because of the following:

  • The above assumptions were mostly based on industry data and did not consider Nebius’s cost saving from optimizations like self-manufactured servers and so on, Nebius’s CapEx and OpEx could be considerably lower than industry standards.
  • Nebius is developing end2end software solutions which will have much higher margin than leasing bare GPU hardwares.

To conclude, I think it’s legit to have negative earnings for the initial few years as that’s the period when Nebius deploys significant amount of capital on building the data centers. I believe as the business scales up and the payback period of the GPU investments is reached, the company will become very profitable.

Cheers,
Luffy

32 Likes