Research from Water Tower Research, Water Tower Research

Water Tower Research

Research from Water Tower Research

Water Tower Research is a leading equity research firm that provides research on micro, small and mid cap companies.

Water Tower Research was founded by Wall Street veterans after witnessing a fundamental shift in the investment research industry with the result being a declining knowledge base among investors, a lack of awareness and a shrinking pool of fundamental research being conducted on companies across the small-cap and micro-cap spectrum. The same issues are now increasingly occurring in the mid-cap space. This sea change within the investment community is impacting not only the companies themselves but also current shareholders and potential investors. With significantly less active sell-side sponsorship by Wall Street, a substantial information gap has developed between companies and various constituents in the financial community. The net result is fewer channels for insight and information to flow between companies and investors.

Equity Research Firm

Equity Bottom-Up

<ul><li>The (Compute) Price is…Right? It’s hard to argue with Nebius’ 2Q results; the market apparently agreed, with shares closing up 34% on August 12.</li><li>Nebius now has joined the $50M/MW/year club (on certain deals); stepping back, industry analysts like SemiAnalysis have estimated that the frontier labs like OpenAI and Anthropic can generate over $100/MW/year, so it seems like these prices are justifiable at the moment.</li><li>That said, the sudden availability of an open-weight model (Kimi K3) that is competitive with frontier labs at significantly lower list prices sheds some potential long-term doubts on the sustainability of that revenue. </li></ul>

The (Compute) Price is…Right? It’s hard to argue with Nebius’ 2Q results; the market apparently agreed, with shares closing up 34% on August 12.Nebius now has joined the $50M/MW/year club (on certain deals); stepping back, industry analysts like SemiAnalysis have estimated that the frontier labs like OpenAI and Anthropic can generate over $100/MW/year, so it seems like these prices are justifiable at the moment.That said, the sudden availability of an open-weight model (Kimi K3) that is competitive with frontier labs at significantly lower list prices sheds some potential long-term doubts on the sustainability of that revenue. 

NBIS: 2Q26 Results; The (Compute) Price Is Right?

<div><ul><li>Lowering near-term estimates on capacity timing, not demand. </li><li>We are reducing our 2Q26 revenue estimate to $550MM from $600MM and our 2Q26 adjusted EBITDA estimate to $146MM from $185MM. </li><li>This still reflects 2Q revenue growth of 38% Q/Q and 423% Y/Y, although our estimate is about 5% below consensus. </li></ul></div>

Lowering near-term estimates on capacity timing, not demand. We are reducing our 2Q26 revenue estimate to $550MM from $600MM and our 2Q26 adjusted EBITDA estimate to $146MM from $185MM. This still reflects 2Q revenue growth of 38% Q/Q and 423% Y/Y, although our estimate is about 5% below consensus. 

NBIS: Tweaking Estimates; Strong Demand, but Potential Capacity Bottlenecks

Thematic (Sector/Industry)

<div><ul><li>The model tests the economics of the AI buildout. It maps spending from power and capex through clouds to end-market revenue, anchored on FactSet consensus for hyperscaler capex and cloud revenue while our assumptions for AI allocation, utilization, pricing and demand, and modeled intercompany transfers are eliminated.</li><li>Over time, we expect to use this model as a framework for evaluating new data points on capex, AI services revenue, and token pricing/consumption.</li><li>The buildout stays cash flow negative through 2030. AI capex rises from $824B in 2026 to $1.76T in 2030 ($6.7T cumulative) versus revenue of $306B and $1.56T. Cloud FCF is negative $530B and $588B, including assumed internal use benefits; infrastructure vendors generate $258B and $628B.</li></ul></div>

The model tests the economics of the AI buildout. It maps spending from power and capex through clouds to end-market revenue, anchored on FactSet consensus for hyperscaler capex and cloud revenue while our assumptions for AI allocation, utilization, pricing and demand, and modeled intercompany transfers are eliminated.Over time, we expect to use this model as a framework for evaluating new data points on capex, AI services revenue, and token pricing/consumption.The buildout stays cash flow negative through 2030. AI capex rises from $824B in 2026 to $1.76T in 2030 ($6.7T cumulative) versus revenue of $306B and $1.56T. Cloud FCF is negative $530B and $588B, including assumed internal use benefits; infrastructure vendors generate $258B and $628B.

AI Flow Industry Model : Tracking the Capex Flood and Revenue Streams

<div><ul><li>Strong Nvidia beat reinforces strength in demand environment for AI. </li><li>Nvidia’s 3QFY26 results materially exceeded consensus expectations, with revenue of $57.0 billion (+22% Q/Q, +62% Y/Y) versus Street estimates of ~$54.9 billion and adjusted EPS of $1.30 versus Street estimates of $1.25.</li><li>Data center revenue of $51.2 billion handily surpassed consensus of ~$49.0 billion, accounting for 90% of total revenue and underscoring sustained enterprise AI adoption.</li></ul></div>

Strong Nvidia beat reinforces strength in demand environment for AI. Nvidia’s 3QFY26 results materially exceeded consensus expectations, with revenue of $57.0 billion (+22% Q/Q, +62% Y/Y) versus Street estimates of ~$54.9 billion and adjusted EPS of $1.30 versus Street estimates of $1.25.Data center revenue of $51.2 billion handily surpassed consensus of ~$49.0 billion, accounting for 90% of total revenue and underscoring sustained enterprise AI adoption.

Nvidia Earnings Read-Across; Now for Something (Very) Different!

<ul><li>Massive AI infrastructure boom. There is a massive data center investment cycle driven by artificial intelligence (AI). </li><li>The global AI infrastructure market will surge from $88 billion in 2025 to $198 billion by 2030 (18% CAGR), with the independent AI cloud segment increasing significantly from $25 billion to $170 billion (48% CAGR).</li><li> The shift to inference is a significant driver, which, as a share of workloads (versus training), is expected to shift from 20% to 80%, creating a massive opportunity for specialized providers over general-purpose hyperscalers</li></ul>

Massive AI infrastructure boom. There is a massive data center investment cycle driven by artificial intelligence (AI). The global AI infrastructure market will surge from $88 billion in 2025 to $198 billion by 2030 (18% CAGR), with the independent AI cloud segment increasing significantly from $25 billion to $170 billion (48% CAGR). The shift to inference is a significant driver, which, as a share of workloads (versus training), is expected to shift from 20% to 80%, creating a massive opportunity for specialized providers over general-purpose hyperscalers

Netherlands

Nebius Group is a technology company providing infrastructure and services to AI builders globally. The group's core business, Nebius, is an AI-centric cloud platform built for AI workloads. Nebius Group serves customers worldwide.

Price Gauge
52-week Range	0.00 - 0.00
Market Cap (USDm)	0.00
P/E	0.00
Dividend (Yield)	0.00

Nebius Group (NBIS US)

NBIS: 2Q26 Results; The (Compute) Price Is Right?

The (Compute) Price is…Right? It’s hard to argue with Nebius’ 2Q results; the market apparently agreed, with shares closing up 34% on August 12.

Water Tower Research

NBIS: Tweaking Estimates; Strong Demand, but Potential Capacity Bottlenecks

We are reducing our 2Q26 revenue estimate to $550MM from $600MM and our 2Q26 adjusted EBITDA estimate to $146MM from $185MM.

Water Tower Research

AI Flow Industry Model : Tracking the Capex Flood and Revenue Streams

Over time, we expect to use this model as a framework for evaluating new data points on capex, AI services revenue, and token pricing/consumption.

Water Tower Research

Nvidia Earnings Read-Across; Now for Something (Very) Different!

Nvidia’s 3QFY26 results materially exceeded consensus expectations, with revenue of $57.0 billion (+22% Q/Q, +62% Y/Y) versus Street estimates of...

Water Tower Research

NBIS: "Neo" Of the Cloud?

. We also expect the shift from training to inference business to be positive for the margins of infrastructure providers like Nebius.

Water Tower Research

Price Gauge

52-week Range

Market Cap (USDm)

P/E

Dividend (Yield)