Melvin

AI Analyst @MilkRoadAI | Co-Founder @_Investinq + @optionality_ | Finding opportunities across AI, photonics, defense, space, and tech.

[Melvin @MelvinInvests]

Nebius will be the first $1 trillion neocloud company (Save this).

Most people have never heard of Nebius and they should have because it is the fastest growing infrastructure company on the planet right now

AI cloud revenue grew 841% year over year in Q1 2026 to $389.7 million, and the company exited Q1 with an annualized run rate of $1.9 billion that management expects to reach $7 to $9 billion by year end.

For context, t took AWS over five years to grow from $1 billion to $7 billion in annual revenue while Nebius is doing it in a single year.

The neocloud category itself is the key to understanding why Nebius's ceiling is so high.

Neoclouds are clean slate AI cloud providers, built from the ground up for training and inference instead of retrofitting legacy infrastructure like AWS, Google Cloud, or Azure.

This matters because AI workloads are fundamentally different from traditional cloud workloads because they require extremely high bandwidth GPU interconnects, low latency storage, optimized networking fabrics, and custom cooling, none of which legacy hyperscalers were built for.

The neocloud market generated roughly $25 billion in revenue in 2025, up 223% year over year and Synergy Research projects it will approach $400 billion by 2031, compounding at 58% annually, one of the fastest sustained growth rates ever recorded for an infrastructure category of this scale.

And Nebius is the best positioned pure play in that market.

Now look at the Wells Fargo capacity chart.

Nebius was running 172 MW of active capacity at the end of 2025 and by 2030, Wells Fargo projects 7.5 GW, a 43x expansion in five years.

The two flagship buildouts driving that number are enormous in scale and ambition.

The Independence, Missouri campus is a 400-acre AI factory campus approved under Chapter 100 incentives, the largest of its kind in the Midwest scaling to 1.2 GW when fully built.

The Pennsylvania campus, announced in Q1 2026, is an identically sized 1.2 GW site that will begin contributing revenue in the first half of 2027 with 250-300 MW online by year-end.

Both campuses have a combined planned capacity of 2.4 GW and those are just the flagship self-build sites.

The Wells Fargo model has Nebius adding another 3.4 GW of other/unallocated capacity through 2030, which represents future site announcements not yet disclosed.

The financial mechanics of why $1 trillion is achievable are not complicated.

Wells Fargo's model implies that at 7.5 GW of active capacity and using the industry standard of approximately $10 million per MW per year in revenue, Nebius would be generating roughly $75 billion in annualized revenue by 2030.

At a conservative 5x revenue multiple ,well below what infrastructure scale AI businesses command today that is a $375 billion market cap.

At the revenue multiples Nebius currently trades at, which reflect the scarcity of pure play AI infrastructure assets, you reach $1 trillion.

I’m extremely bullish on Nebius.

Follow me @MelvinInvests for more on $NBIS and other underrated AI infrastructure gems.

[Melvin @MelvinInvests]

@femu I also have a position in CoreWeave too :)

[Melvin @MelvinInvests]

Nebius will be the first $1 trillion neocloud company (Save this).

Most people have never heard of Nebius and they should have because it is the fastest growing infrastructure company on the planet right now

AI cloud revenue grew 841% year over year in Q1 2026 to $389.7 million, and the company exited Q1 with an annualized run rate of $1.9 billion that management expects to reach $7 to $9 billion by year end.

For context, t took AWS over five years to grow from $1 billion to $7 billion in annual revenue while Nebius is doing it in a single year.

The neocloud category itself is the key to understanding why Nebius's ceiling is so high.

Neoclouds are clean slate AI cloud providers, built from the ground up for training and inference instead of retrofitting legacy infrastructure like AWS, Google Cloud, or Azure.

This matters because AI workloads are fundamentally different from traditional cloud workloads because they require extremely high bandwidth GPU interconnects, low latency storage, optimized networking fabrics, and custom cooling, none of which legacy hyperscalers were built for.

The neocloud market generated roughly $25 billion in revenue in 2025, up 223% year over year and Synergy Research projects it will approach $400 billion by 2031, compounding at 58% annually, one of the fastest sustained growth rates ever recorded for an infrastructure category of this scale.

And Nebius is the best positioned pure play in that market.

Now look at the Wells Fargo capacity chart.

Nebius was running 172 MW of active capacity at the end of 2025 and by 2030, Wells Fargo projects 7.5 GW, a 43x expansion in five years.

The two flagship buildouts driving that number are enormous in scale and ambition.

The Independence, Missouri campus is a 400-acre AI factory campus approved under Chapter 100 incentives, the largest of its kind in the Midwest scaling to 1.2 GW when fully built.

The Pennsylvania campus, announced in Q1 2026, is an identically sized 1.2 GW site that will begin contributing revenue in the first half of 2027 with 250-300 MW online by year-end.

Both campuses have a combined planned capacity of 2.4 GW and those are just the flagship self-build sites.

The Wells Fargo model has Nebius adding another 3.4 GW of other/unallocated capacity through 2030, which represents future site announcements not yet disclosed.

The financial mechanics of why $1 trillion is achievable are not complicated.

Wells Fargo's model implies that at 7.5 GW of active capacity and using the industry standard of approximately $10 million per MW per year in revenue, Nebius would be generating roughly $75 billion in annualized revenue by 2030.

At a conservative 5x revenue multiple ,well below what infrastructure scale AI businesses command today that is a $375 billion market cap.

At the revenue multiples Nebius currently trades at, which reflect the scarcity of pure play AI infrastructure assets, you reach $1 trillion.

I’m extremely bullish on Nebius.

Follow me @MelvinInvests for more on $NBIS and other underrated AI infrastructure gems.

[Melvin @MelvinInvests]

Nebius will be the first $1 trillion neocloud company (Save this).

Most people have never heard of Nebius and they should have because it is the fastest growing infrastructure company on the planet right now

AI cloud revenue grew 841% year over year in Q1 2026 to $389.7 million, and the company exited Q1 with an annualized run rate of $1.9 billion that management expects to reach $7 to $9 billion by year end.

For context, t took AWS over five years to grow from $1 billion to $7 billion in annual revenue while Nebius is doing it in a single year.

The neocloud category itself is the key to understanding why Nebius's ceiling is so high.

Neoclouds are clean slate AI cloud providers, built from the ground up for training and inference instead of retrofitting legacy infrastructure like AWS, Google Cloud, or Azure.

This matters because AI workloads are fundamentally different from traditional cloud workloads because they require extremely high bandwidth GPU interconnects, low latency storage, optimized networking fabrics, and custom cooling, none of which legacy hyperscalers were built for.

The neocloud market generated roughly $25 billion in revenue in 2025, up 223% year over year and Synergy Research projects it will approach $400 billion by 2031, compounding at 58% annually, one of the fastest sustained growth rates ever recorded for an infrastructure category of this scale.

And Nebius is the best positioned pure play in that market.

Now look at the Wells Fargo capacity chart.

Nebius was running 172 MW of active capacity at the end of 2025 and by 2030, Wells Fargo projects 7.5 GW, a 43x expansion in five years.

The two flagship buildouts driving that number are enormous in scale and ambition.

The Independence, Missouri campus is a 400-acre AI factory campus approved under Chapter 100 incentives, the largest of its kind in the Midwest scaling to 1.2 GW when fully built.

The Pennsylvania campus, announced in Q1 2026, is an identically sized 1.2 GW site that will begin contributing revenue in the first half of 2027 with 250-300 MW online by year-end.

Both campuses have a combined planned capacity of 2.4 GW and those are just the flagship self-build sites.

The Wells Fargo model has Nebius adding another 3.4 GW of other/unallocated capacity through 2030, which represents future site announcements not yet disclosed.

The financial mechanics of why $1 trillion is achievable are not complicated.

Wells Fargo's model implies that at 7.5 GW of active capacity and using the industry standard of approximately $10 million per MW per year in revenue, Nebius would be generating roughly $75 billion in annualized revenue by 2030.

At a conservative 5x revenue multiple ,well below what infrastructure scale AI businesses command today that is a $375 billion market cap.

At the revenue multiples Nebius currently trades at, which reflect the scarcity of pure play AI infrastructure assets, you reach $1 trillion.

I’m extremely bullish on Nebius.

Follow me @MelvinInvests for more on $NBIS and other underrated AI infrastructure gems.

[Melvin @MelvinInvests]

@Sandeman52 Hell yea man! Goal is to retire very soon like you haha

[Melvin @MelvinInvests]

@MindfulDataPath very true! They are much more!

[Melvin @MelvinInvests]

@Sandeman52 Been a fan of you for a while, man! So that means a lot!

[Melvin @MelvinInvests]

@__thebenjamins @MilkRoadAI Exactly! People are still sleeping on this!

[Melvin @MelvinInvests]

@homer_b_boomer @apmoww Lol I am on podcasts too

[Melvin @MelvinInvests]

Micron will be a $4,000 stock (Save this).

Three memory companies, Micron, SK Hynix and Samsung are on a path to generate a combined $945 billion in operating income by 2029.

A decade ago that number would have been laughed out of any analyst meeting, but today it's already baked into contracted supply agreements.

For most of the last 20 years, memory was a terrible business, cyclical, commoditized, boom then bust then crash.

AI broke that cycle permanently and here's the mechanism.

A single AI server requires 8 to 10 times the DRAM of a traditional server, and HBM, the premium memory stacked directly on top of AI chips is now required for every major GPU architecture on earth.

The problem is that HBM takes roughly 3x the wafer area of standard DRAM to manufacture, so every time a fab shifts capacity toward HBM to meet AI demand, it simultaneously creates a shortage of conventional DRAM for everything else.

The result is that prices are rising everywhere at once.

DRAM contract prices surged over 170% year over year by end of 2025, and Q2 2026 prices are projected to rise another 58 to 63% quarter over quarter, the steepest single quarter jump in a decade.

And HBM is not a commodity but rather a qualification gated infrastructure.

To sell HBM to a hyperscaler, your product has to pass a rigorous multi-month qualification process tied to their specific chip architecture.

The moment you're qualified, you're essentially the only approved supplier for that platform until the next generation ships, 18 to 24 months later.

The competitive moat isn't brand loyalty or price. It's a technical certification that physically locks out everyone else.

Samsung's semiconductor division posted a 47-fold increase in operating profit in Q1 2026 alone, and SK Hynix's operating margins are approaching 80%, the highest of any major company on the planet.

Now here's why Micron is going to be the biggest winner of all of them.

SK Hynix holds 62% HBM market share and got there first but Micron is the most compelling investment, and the gap is closing fast.

Micron is the only US domiciled memory company, which means every CHIPS Act grant, every Pentagon preferred supplier contract and every buy American infrastructure mandate flows through it by default.

Its Idaho and New York fabs are receiving $6.1 billion in direct federal grants plus billions more in subsidized loans, government funded capacity expansion that competitors in Korea simply don't get.

It also has the most operating leverage of anyone in this cycle.

Micron started at 21% HBM market share after years of underinvestment, which means every point it gains from here drops directly to the bottom line because the fixed cost base is already set.

Its HBM4 is now shipping with speeds exceeding 11 Gb/s, bandwidth above 2.8 TB/s, and 20% better power efficiency than HBM3E which matters enormously to hyperscalers paying $10 million per month per data center in electricity costs.

The profit projection in that chart puts Micron at $160 billion in operating income by 2029 alone, which would make it one of the most profitable companies on the planet.

The current stock price doesn't reflect that reality yet.

[Melvin @MelvinInvests]

@cronosbro I did that

[Melvin @MelvinInvests]

@colon4893 Slow down in AI capex and power

[Melvin @MelvinInvests]

@colon4893 CoreWeave and lambda

[Melvin @MelvinInvests]

@paulguidry @_Investinq Yea about 14!

[Melvin @MelvinInvests]

@philipp317 Consider dropping a follow!

[Melvin @MelvinInvests]

@colon4893 About 14x but yes!

[Melvin @MelvinInvests]

@MilkRoadAI good, you def should!

[Melvin @MelvinInvests]

@homer_b_boomer Lmaooo I already made a shit on money with MU so I am fine. This is an educational post.

[Melvin @MelvinInvests]

@gibby_joe52901 I am very familiar but the US won’t.

[Melvin @MelvinInvests]

@terrified2024 lol okay

[Melvin @MelvinInvests]

@shanhaisiwa Supply and demand!

[Melvin @MelvinInvests]

Micron will be a $4,000 stock (Save this).

Three memory companies, Micron, SK Hynix and Samsung are on a path to generate a combined $945 billion in operating income by 2029.

A decade ago that number would have been laughed out of any analyst meeting, but today it's already baked into contracted supply agreements.

For most of the last 20 years, memory was a terrible business, cyclical, commoditized, boom then bust then crash.

AI broke that cycle permanently and here's the mechanism.

A single AI server requires 8 to 10 times the DRAM of a traditional server, and HBM, the premium memory stacked directly on top of AI chips is now required for every major GPU architecture on earth.

The problem is that HBM takes roughly 3x the wafer area of standard DRAM to manufacture, so every time a fab shifts capacity toward HBM to meet AI demand, it simultaneously creates a shortage of conventional DRAM for everything else.

The result is that prices are rising everywhere at once.

DRAM contract prices surged over 170% year over year by end of 2025, and Q2 2026 prices are projected to rise another 58 to 63% quarter over quarter, the steepest single quarter jump in a decade.

And HBM is not a commodity but rather a qualification gated infrastructure.

To sell HBM to a hyperscaler, your product has to pass a rigorous multi-month qualification process tied to their specific chip architecture.

The moment you're qualified, you're essentially the only approved supplier for that platform until the next generation ships, 18 to 24 months later.

The competitive moat isn't brand loyalty or price. It's a technical certification that physically locks out everyone else.

Samsung's semiconductor division posted a 47-fold increase in operating profit in Q1 2026 alone, and SK Hynix's operating margins are approaching 80%, the highest of any major company on the planet.

Now here's why Micron is going to be the biggest winner of all of them.

SK Hynix holds 62% HBM market share and got there first but Micron is the most compelling investment, and the gap is closing fast.

Micron is the only US domiciled memory company, which means every CHIPS Act grant, every Pentagon preferred supplier contract and every buy American infrastructure mandate flows through it by default.

Its Idaho and New York fabs are receiving $6.1 billion in direct federal grants plus billions more in subsidized loans, government funded capacity expansion that competitors in Korea simply don't get.

It also has the most operating leverage of anyone in this cycle.

Micron started at 21% HBM market share after years of underinvestment, which means every point it gains from here drops directly to the bottom line because the fixed cost base is already set.

Its HBM4 is now shipping with speeds exceeding 11 Gb/s, bandwidth above 2.8 TB/s, and 20% better power efficiency than HBM3E which matters enormously to hyperscalers paying $10 million per month per data center in electricity costs.

The profit projection in that chart puts Micron at $160 billion in operating income by 2029 alone, which would make it one of the most profitable companies on the planet.

The current stock price doesn't reflect that reality yet.