Investments & market insights 📈 Daily updates + my portfolio growth 💸 YouTube 🎥 https://xitter.cf/t.co/dUuKweywva
Joined March 2024
- Tweets20.7K
- Following345
- Followers25.8K
- Likes11K
Daniel retweeted
My 2030 price target for $NBIS is $3,000.
Here’s why.
Nebius expects to have at least 5 GW of active compute capacity deployed by 2030.
And the economics are getting stronger:
• Annual revenue used to be around $10M per MW.
• Long-term contracts are now closer to $20–25M per MW.
• Short-term capacity can reach $40–50M per MW.
And this is far from over.
The imbalance between AI compute supply and demand is expected to keep widening for at least another 18–24 months.
For 2030, I’m assuming $25M (conservatively) in annual revenue per MW.
5,000 MW × $25M = $125B in annual revenue.
Now comes profitability.
AWS currently operates at ~40% operating margins, implying an estimated net margin of around 30%.
At scale, I’m modeling a 25% net margin for Nebius.
$125B revenue × 25% = $31.25B in net income.
And the growth won’t stop in 2030.
Management expects to add at least 1.5–2 GW of incremental compute capacity annually.
From a 5 GW base, that’s another 30–40% in annual capacity growth.
A company generating over $31B in annual net income while expanding capacity at that pace deserves a premium valuation.
I’m assigning it a 30x P/E.
$31.25B × 30 = ~$940B market cap.
I’m also factoring in approximately 25% additional share dilution by 2030 to account for the massive capital required to fund this expansion.
That brings my price target to approximately $3,000 per share.
A nearly $1 trillion company by 2030.
And potentially still growing capacity by 30–40% annually.
The market is massively underestimating what Nebius is building.
$META is still the best buy among the MAG 7.
Excluding $NVDA, no other MAG 7 company is growing revenue faster.
Q3 revenue: +26% YoY.
$META’s biggest competitive advantage is distribution.
3.6 billion people use Meta’s platforms every single day.
Nearly half the world’s population.
Facebook, Instagram, and WhatsApp are the three most used social platforms in the world.
While other AI companies spend billions acquiring users, Meta already has direct access to billions of them.
And with Muse gaining traction, this advantage is becoming even more powerful.
Other AI companies are fighting for users.
$META already has them.
Some predictions:
• $NBIS becomes a hyperscaler, the AWS of AI.
• $ORCL becomes a $1T+ company within the next 3 years.
• $META smart glasses replace the iPhone.
• $GRAB becomes the operating system of Southeast Asia.
• $AMZN becomes a $10T company by 2030.
• $LMND becomes the largest insurance company in the world.
$SPCX earnings just OBLITERATED the bear case for neoclouds.
$CRWV, $NBIS, and $IREN lost roughly 25% of their value last week.
The selloff was driven by expectations of higher interest rates, which would increase financing costs for an industry that needs to raise hundreds of billions of dollars to fund its expansion.
HIGHER COMPUTE PRICES WILL MORE THAN OFFSET HIGHER INTEREST RATES.
And the numbers discussed during $SPCX earnings illustrate exactly why.
The CFO revealed that new compute deployments are generating payback periods of LESS THAN 12 MONTHS.
Elon Musk went even further, suggesting that payback periods could fall to just 6 MONTHS in 2027.
AI compute demand continues to grow faster than supply, driving prices higher.
Look at the economics:
• Earlier this year, AI compute contracts were generating roughly $10–15M in annual revenue per MW.
• Today, pricing ranges from the high $30Ms to $50M per MW.
The economics of AI infrastructure are improving dramatically, even as financing costs rise.
THE NEOCLOUD BULL CASE HAS NEVER BEEN STRONGER.
If $IREN delivers Horizon 2–4 to $MSFT in December, I expect a significant re-rating.
We already know $IREN has the power, with a pipeline of at least 5.8 GW.
And the company is proving that it no longer needs to rely on massive shareholder dilution to fund its expansion.
Its financing strategy is becoming increasingly clear:
• Customer prepayments.
• GPU-backed financing.
• Future operating cash flow.
• Data center-backed financing.
Now there is only ONE thing left to prove: execution.
$IREN needs to demonstrate that it can actually deliver the compute capacity it has promised.
Successfully delivering Horizon 2–4 to $MSFT in December would prove that $IREN can execute on its commitments.
And once that happens, 99% of the bearish arguments against $IREN will lose their relevance.
The power is secured. The financing is falling into place.
Execution is the final piece of the puzzle.
And December could change everything.
$CBRS won’t stop at $AMD and $AMZN.
More partnerships are coming, and I expect $GOOGL to be next.
AI inference has two main stages:
• Processing the prompt.
• Generating the answer.
GPUs and $AMZN’s Trainium are extremely good at the first.
But when it comes to generating tokens at extreme speed, Cerebras is built for exactly that job.
And this is the key to the entire $CBRS thesis:
Cerebras doesn’t need to replace GPUs.
It can COMPLEMENT them.
Because Cerebras supports open standards, its chips can work alongside other AI accelerators.
We’re already seeing this approach take shape.
$AMD Helios processes the prompt.
$CBRS generates the answer.
One inference flow. Extreme speed. Massive throughput.
Cerebras is pursuing the same approach with $AMZN Trainium on AWS.
The pattern is becoming obvious:
• $AMD + $CBRS
• Trainium + $CBRS
And eventually:
• TPU + $CBRS
$GOOGL is the obvious next partner.
Cerebras is positioning itself to become the token-generation layer of the entire AI cloud ecosystem.
$28B market cap. Over $25B in backlog. And the potential to become the token-generation layer for every major cloud provider.
$CBRS is now my favorite semiconductor stock.
My 2030 price target for $NBIS is $3,000.
Here’s why.
Nebius expects to have at least 5 GW of active compute capacity deployed by 2030.
And the economics are getting stronger:
• Annual revenue used to be around $10M per MW.
• Long-term contracts are now closer to $20–25M per MW.
• Short-term capacity can reach $40–50M per MW.
And this is far from over.
The imbalance between AI compute supply and demand is expected to keep widening for at least another 18–24 months.
For 2030, I’m assuming $25M (conservatively) in annual revenue per MW.
5,000 MW × $25M = $125B in annual revenue.
Now comes profitability.
AWS currently operates at ~40% operating margins, implying an estimated net margin of around 30%.
At scale, I’m modeling a 25% net margin for Nebius.
$125B revenue × 25% = $31.25B in net income.
And the growth won’t stop in 2030.
Management expects to add at least 1.5–2 GW of incremental compute capacity annually.
From a 5 GW base, that’s another 30–40% in annual capacity growth.
A company generating over $31B in annual net income while expanding capacity at that pace deserves a premium valuation.
I’m assigning it a 30x P/E.
$31.25B × 30 = ~$940B market cap.
I’m also factoring in approximately 25% additional share dilution by 2030 to account for the massive capital required to fund this expansion.
That brings my price target to approximately $3,000 per share.
A nearly $1 trillion company by 2030.
And potentially still growing capacity by 30–40% annually.
The market is massively underestimating what Nebius is building.
$IREN CEO couldn’t care less about higher interest rates.
The economics are simply too strong.
With GPU payback periods of around 2 years on 3-year contracts, even higher borrowing costs barely move the economics.
And the real bottleneck in AI infrastructure isn’t financing. It’s POWER.
Look at what happened with $MSFT and Nscale.
Microsoft reportedly walked away from a UK data center project because the grid connection wouldn’t be available until 2030.
This is exactly why $IREN’s access to power is such a massive competitive advantage.
And its financing model is just as impressive:
• Customer prepayments + GPU-backed financing can cover more than 100% of GPU costs.
• Once its data centers become operational, $IREN expects to unlock additional financing backed by those same assets.
$IREN is signing contracts at more than DOUBLE last year’s prices, with demand for 2027 capacity exceeding supply several times over.
The market is obsessing over financing risks while $IREN’s biggest challenge is bringing capacity online fast enough to meet demand.
The bottleneck is power. And $IREN has it.
Daniel retweeted
$IREN at $26 is one of the biggest opportunities in the entire stock market.
I’m buying.
It’s no longer just about having 5.8 GW of power capacity. It’s about turning that power into revenue-generating AI infrastructure.
And $IREN is delivering.
• Horizon 1, the first tranche of its $MSFT contract, is already operational.
• Horizon 2, 3, and 4 are scheduled for delivery by December.
The era of massive shareholder dilution is coming to an end.
$IREN is building a financing flywheel designed to fund its explosive growth without relying on massive dilution.
Customer prepayments. GPU-backed debt. Operating cash flow. And potentially data center-backed financing.
But here’s where things get even more interesting.
Uncontracted capacity is an ASSET, not a liability.
AI compute demand continues to outpace supply, pushing annualized revenue per MW from roughly $10M earlier this year to $25M today, with short-term contracts reaching as high as $50M.
IREN hasn’t locked in yesterday’s prices. It gets to capture tomorrow’s.
$10 BILLION market cap for a company with 5.8 GW of power capacity and these economics?
Absolutely ridiculous.
$META is the best buy among the Magnificent 7.
Just like $GOOGL was in 2025.
Since its April 2025 lows, $GOOGL has rallied 190%.
Meanwhile, $META is up just 30% from its July lows.
I expect $META to follow a similar trajectory.
And the fundamentals speak for themselves:
• Excluding $NVDA, $META is the fastest-growing Magnificent 7 company by revenue.
• Revenue grew 28% YoY in Q2 2026.
• $META is on track to surpass $GOOGL in net ad revenue this year: $243B vs. $239B.
• With Muse and Meta AI, $META has established itself as a frontrunner in AI.
• Meta glasses are on track to sell over 20M units this year.
And you’re paying just 22x forward earnings.
Most of the time, making money in the stock market is this simple.
$CRWV, $NBIS, and $IREN crashed 25% last week.
Meanwhile, Nebius CEO Arkady Volozh sees a completely different picture:
“Demand significantly exceeds supply. You can sell everything you build. And the gap is widening.”
And that growing imbalance is driving prices higher.
Look at the economics:
• AI compute contracts that previously generated $10–15M in annual revenue per MW are now averaging $20–25M.
• Some short-term contracts have reached $40–50M per MW annually.
And prices are expected to climb even further.
According to Arkady, this supply-demand imbalance will continue to widen for at least another 18–24 months.
If AI continues to deliver increasing economic value across industries, this imbalance could persist for years beyond that.
Neocloud stocks have collapsed, but their investment thesis hasn’t deteriorated.
It’s gotten STRONGER.
Demand is accelerating. Supply can’t keep up. Pricing power is increasing.
The stocks are collapsing. The fundamentals are getting stronger.
This selloff is driven by sentiment, not deteriorating fundamentals.
You can now buy $GRAB at the same price as its CEO.
In September, Anthony Tan bought $30M worth of shares at an average price of $2.89.
Meanwhile, $GRAB is trading at just 16.9x NTM earnings, despite what management is projecting:
• Revenue CAGR of over 30% through 2028.
• EPS CAGR of approximately 40% through 2028.
• Financial Services reaching Adjusted EBITDA breakeven in H2 2026.
• Financial Services generating $500M in Adjusted EBITDA by 2028.
On top of that, the company announced a $900M share buyback, equivalent to 7.5% of its market cap.
40% annual earnings growth at just 16.9x forward earnings.
Very few opportunities in this market make me as bullish as $GRAB.
$CBRS is becoming an extremely attractive opportunity.
Down 67% from its IPO-day peak, with a market cap of just $30B.
Look at the valuations:
Based on projected FY2028 revenues:
• $CBRS: 4.01x P/S
• $NVDA: 5.73x P/S
• $AMD: 6.97x P/S
Cerebras is trading at a significant discount to both Nvidia and AMD.
And the margin argument doesn’t tell the whole story.
$CBRS reported a 41% gross margin in Q2 2026, but management has repeatedly outlined a medium-to-long-term target of 60%.
That’s below $NVDA’s ~75%, but ABOVE $AMD’s ~54%.
Much of $CBRS’s recent selloff has been driven by the fact that OpenAI’s latest ultra-fast GPT models are powered by Nvidia GPUs rather than Cerebras chips.
But Cerebras’ CEO explained why: the company is SOLD OUT.
Cerebras simply didn’t have the available capacity to allocate to OpenAI.
Sam Altman himself reiterated that Cerebras remains a close partner of OpenAI.
And this is where the opportunity gets particularly interesting.
The demand for ultra-fast inference is going to be enormous.
With Cerebras chips delivering inference speeds up to 13x faster than Nvidia GPUs in reported comparisons, I have little doubt that $CBRS will capture a massive share of this market.
The stock is down 67%.
$CBRS at a $30B valuation is a massive opportunity.
The market is underestimating just how big the fast inference market will become, and how much of it Cerebras will capture.
$IREN at $26 is one of the biggest opportunities in the entire stock market.
I’m buying.
It’s no longer just about having 5.8 GW of power capacity. It’s about turning that power into revenue-generating AI infrastructure.
And $IREN is delivering.
• Horizon 1, the first tranche of its $MSFT contract, is already operational.
• Horizon 2, 3, and 4 are scheduled for delivery by December.
The era of massive shareholder dilution is coming to an end.
$IREN is building a financing flywheel designed to fund its explosive growth without relying on massive dilution.
Customer prepayments. GPU-backed debt. Operating cash flow. And potentially data center-backed financing.
But here’s where things get even more interesting.
Uncontracted capacity is an ASSET, not a liability.
AI compute demand continues to outpace supply, pushing annualized revenue per MW from roughly $10M earlier this year to $25M today, with short-term contracts reaching as high as $50M.
IREN hasn’t locked in yesterday’s prices. It gets to capture tomorrow’s.
$10 BILLION market cap for a company with 5.8 GW of power capacity and these economics?
Absolutely ridiculous.
$AVGO is becoming impossible to ignore.
Down 35% from its highs, Broadcom is now my favorite semiconductor stock.
ASICs are expected to capture 25% of the AI accelerator market by 2030, with $AVGO designing the vast majority of them.
CEO Hock Tan expects $AVGO to exceed $30 in EPS by FY2028.
Put a 30x multiple on $30+ in EPS, and you’re looking at a $900+ stock.
Today, $AVGO trades at just $328.
That’s nearly 200% upside in just over two years.
And the growth projections are staggering:
• FY2027 AI semiconductor revenue: $115B
• FY2028 AI semiconductor revenue: $230B
After this week’s brutal selloff, the AI trade is presenting extraordinary opportunities.
$AVGO stands out as one of the biggest.
Tom Lee believes the market has found its support and that we are NOT at the beginning of a larger downturn.
He expects the market to have bottomed yesterday or to bottom on Monday.
He sees a buying opportunity across the entire AI trade:
• Neoclouds: $CRWV, $NBIS, $IREN
• Semiconductors: $NVDA, $AMD, $AVGO
• Memory: $MU, Samsung, SK Hynix
All of these have pulled back to key support levels.
While sentiment around the AI trade has deteriorated, the fundamentals and long-term drivers of AI remain intact.
As Lee puts it:
“When fundamentals hold up and markets become fearful, the risk/reward for investors becomes very attractive.”
Tom Lee is back to doing what he does best: BEING BULLISH.
$IREN is building a financing flywheel that validates its entire investment thesis.
GPUs account for roughly two-thirds of total AI infrastructure CapEx. The remaining third goes toward building the physical data centers.
Now look at the economics:
• Customer prepayments cover 45–55% of GPU CapEx, roughly equivalent to the entire cost of building the physical data centers.
• GPU-backed debt finances up to 90% of GPU costs.
• Operating cash flow and future data center financing fund the remaining requirements and fuel further expansion.
More contracts → More prepayments → More GPUs → More operating cash flow → More capacity → More contracts.
As this flywheel gains momentum, $IREN accelerates compute deployment, reinvests cash flows into further expansion, and reduces its dependence on shareholder dilution.
Once this flywheel reaches scale, $IREN stops being just another capital-intensive AI infrastructure company and becomes a self-funding growth machine.
