Disclosure: I am not a professional. I am a random internet stranger piecing together data. Do your own research; this is not financial advice.

    Bubble talk is increasingly silenced

    The revenue era has arrived. Nebius has continuously raised projections, and AI adoption is taking off faster than anyone thought. Just 6 months ago the market was worried there was a bubble. However, in the past 6 months, Anthropics ARR has gone from 9 billion to above 44 billion. OpenAI is also above 30 billion ARR. The revenue is now there to justify the AI market, and this revenue is ahead of even the optimistic projections.

    Demand & Depreciation

    This means that demand has increased beyond what anyone thought. Anthroipic just signed a 5 billion a year deal with xai for 300 mw of Hopper and Blackwell mix. This is roughly 16 billion per GW of older chips. A year ago the forums were saying that AI infra plays were doomed due to depreciation. But Nebius recently raised Hopper prices by 30%. Older chips so far are selling for more per GW than when they were brand new. Depreciation seems to be overblown. Demand is insatiable.

    Enterprise Adoption

    And we are just getting started. The enterprise software market is massive. The top 4 companies in this area are: Microsoft, Oracle, Salesforce, and SAP. And all 4 are going all in on AI. And at the beginning of their journey. For example, SAP announced just this week their vision for the "autonomous enterprise", showing off workflows where AI does much of the work with a human in the loop approving certain things, and AI working autonomously on the rest. Just about every fortunre 500 company uses SAP. Of course this sort of thing takes time to implement, and further to adopt and rollout to every customer. But this is how GDP gets bumped up. These massive companies, like Coca Cola, are dependent on enterprise software companies to implement autonomous AI workflows in order to see real change. And its just now starting. I cannot understate how big this is. Enterprise software vendors serve hundreds of thousands of customers. You are talking about tens of millions of agents deployed by releasing a new version of software

    We are at the cusp. And demand will only explode further. Whats more: we are now seeing the second exponential take off. The first being training, the second being RL. Inference compute will very likely be the vast majority of overall AI compute in the coming years.

    "Connected" vs. "Monetized" Power, and next gen chips

    Analysts, and many in this sub, are being highly pessimistic on per GW pricing. Nebius connected power does not mean monetizied power. So when we see 9 billion ARR and 1GW of connected power, this does not mean 1GW is equal to 9 billion of revenue. Here is a quote from nebius in an SEC filing regarding their 2025 connected power:

    We expect to have 220 MW of connected power, which includes both active power and power that is fully provisioned in our data centers and can be activated immediately upon GPU installation

    Hence, connected includes power even without GPUs installed.

    The next gen coming, vera rubin, will be much more compute dense, and likely be 2x-4x per GW.

    Nebius is more than infra

    Nebius has a sizeable stake in clickhouse and avride. Clickhouse is used by anthropic, tesla, openai, and many more. They landed a 15 billion valuation, and since then, the ARR of OpenAI and Anthropic have taken off. I would not be surprised to see clickhouses valuation take off as well.

    Nebius has many software services. Including Tavily, an absolutely essential service for AI agents. AI is trained on data that is often 2 years old or older. To get up to date information they need web search. Its essential. You would not use chatgpt without web search. This is what Tavily is. It give the agent a web search tool, returning results in a AI friendly way. And Tavily is the best in class in this area. Used by many just for this purpose.

    My Bullish estimate

    So, with that said, here is my 2030 projection. I am likely not including enough dillution here. We saw SAAS companies with much lower growth routinely get 20x price-to-sales multiples, and if AI really is changing the world economies, then I think 10x price to ARR will be the base case, especially as Nebius has more revenue from their higher-margin services. It could get a much higher multiple here once we start to see a shift and GDP increase:

    Year Est. Monetized Capacity Est. Annual Revenue Projected ARR Multiple (P/ARR) Projected Market Cap Est. Share Count Projected Stock Price
    2026 ~0.7 GW $3.2B $8.5B 10x ~$85B ~265M ~$320
    2027 ~1.6 GW $15.0B $28.5B 10x ~$285B ~280M ~$1,015
    2028 ~3.0 GW $40.0B $63.0B 10x ~$630B ~295M ~$2,135
    2029 ~4.0 GW $80.0B $106.0B 10x ~$1.06T ~305M ~$3,475
    2030 5.0+ GW $125.0B $150.0B+ 10x $1.50T+ ~315M ~$4,760+

    AI is just getting started, and so is Nebius
    byu/Longjumping_Kale3013 instocks



    Posted by Longjumping_Kale3013

    4 Comments

    1. Aint_EZ_bein_AZ on

      Im with you man. These people on reddit are missing out on huge gains. Wonder why they are even on this sub

    2. I understand your excitement my bullishness is hitting ATHs as well but I find your forecast way to optimistic. After 2026 i wouldn’t use ARR multiple anymore. After that only revenue, EBIDTA, EBIT, earnings, etc.

      I do believe there will be way more dilution. I assume will have between 280m and 330m outstanding shares YE2027.

      After 2027 is really hard to predict anything because of quickly changing variables. The prices per GW could vary between $10 – ∞ after 2027. The only thing I know for certain is that Nebius will try their best.

    Leave A Reply