Olivia
Online
White Collar Exit
@WhiteCollarExit
Analyst @MilkRoadAI | C-suite advisor | Studying AI’s full value chain
München, Deutschland
Joined February 2022
119 Following
601 Followers
1.4K Posts
Pinned Tweet
WHY BLOOM ENERGY $BE COULD BE THE AI PLAY FOR 2026
——
The AI boom has a boring problem.
Electricity.
You can’t scale power like software.
Wires, transformers, permits, years.
But hyperscalers are competing on months.
That timing gap is creating a new category: „on-site power that shows up fast“
That’s where Bloom Energy enters.
——
A) What Bloom actually does
Bloom sells on-site power for data centers.
Think “power plant in a box” next to the facility.
Fuel cells produce electricity via an electrochemical process (not the
burn-and-spin turbine setup).
For operators, the use case is simple.
If the grid can only give you 50 MW,
but your campus needs 300 MW,
you can supplement the missing power on-site.
——
B) Why the market suddenly cares
Two forces collided.
Grid expansion is slow and political.
AI demand is compounding fast.
Global data center electricity demand is projected to more than double by 2030 to ~945 TWh.
That scale forces new solutions.
Also because traditional solutions like gas turbines are sold out and nuclear takes too long!
In one recent discussion, Bloom cited a survey showing data centers planning on-site generation jumped from 13% to 38% in about six months.
That’s a real “tipping point” signal.
⸻
C) The 3 moats that matter
1) Speed and uptime-style reliability
Speed is the first moat.
Bloom can deploy its systems at scale within ~90 days.
Also they build these systems as modular ~65 kW building blocks.
So maintenance events are small, not one big turbine down, ensuring 99% uptime.
2) Low emissions and easier community acceptance
Second moat: local footprint.
They emphasize “no combustion” at the point of generation, and they highlight the system being quiet (around ~65 dB).
That matters when communities push back on noisy, dirty backup-style setups.
3) Lower total cost of ownership
Third moat: economics.
If you include TCO (not just fuel price), Bloom is often framed as ~20% lower TCO vs. gas turbines and others.
Especially once you factor in lead times, maintenance windows, and the cost of delay.
In an AI race, time-to-power is part of the bill.
——
Key takeaway
Bloom Energy is perfectly positioned to benefit in an environment where customers need
- 24/7 reliable energy
- fast energy deployment
- low emission energy
- grid not able to supply enough
= AI datacenter boom
$Q sits behind one of the biggest AI shifts: squeezing more tokens out of every powered rack.
We cannot build AI data centers fast enough.
Demand is still huge, but the bottleneck is shifting to the physical layer and societal issues with AI.
That is why efficiency matters.
Every datacenter needs to max out output from the racks that actually gets powered.
Two shifts matter most for Qnity: 800VDC racks and 3D chip packages.
800VDC helps move more power through AI racks more efficiently.
As racks become more power-dense, they need better insulation, thermal management, EMI shielding, and reliability materials.
That's Qnity’s Interconnect Solutions business.
3D chip packages stack and connect chips closer together, so data moves faster and less energy gets wasted.
That's Qnity’s Semiconductor Technologies business, which sells process materials used in advanced chip manufacturing.
So, the thesis is: AI efficiency is becoming a materials-intensity story!
We at Milk Road called Nebius, Credo, Bloom Energy, AAOI, AMD and others before their big run ups. You can join us for just $1 and leave anytime you want!
https://t.co/WqBpzQXqch
$Q sits behind one of the biggest AI shifts: squeezing more tokens out of every powered rack.
We cannot build AI data centers fast enough.
Demand is still huge, but the bottleneck is shifting to the physical layer and societal issues with AI.
That is why efficiency matters.
Every datacenter needs to max out output from the racks that actually gets powered.
Two shifts matter most for Qnity: 800VDC racks and 3D chip packages.
800VDC helps move more power through AI racks more efficiently.
As racks become more power-dense, they need better insulation, thermal management, EMI shielding, and reliability materials.
That's Qnity’s Interconnect Solutions business.
3D chip packages stack and connect chips closer together, so data moves faster and less energy gets wasted.
That's Qnity’s Semiconductor Technologies business, which sells process materials used in advanced chip manufacturing.
So, the thesis is: AI efficiency is becoming a materials-intensity story!
@milkroaddaily Fully agree on the chemicals angle!
Did a full deep dive on the Milk Road AI YT-Channel on $Q.
Massively benefiting from 800VDC and 3D chip design!
@KyleReidhead @MilkRoadMacro token usage will go up, for sure!
The key question: Do you need a Ferrari for every ride or is a Honda enough sometimes?
Think about it!
The AI buildout trade is moving toward deployability of chips & efficiency!
2 other interesting arenas to look at:
1st, downstream adoption of AI, e,g,. in healthcare or finance! Eli Lilly a name to watch 👀
2nd, as the agentic economy scales, the economic rails for them get interesting!
The biggest IPO wave since the internet is here: SpaceX, Anthropic, and OpenAI.
But our lead AI analyst @WhiteCollarExit is buying none of them.
Here's his thinking on why (save this):
Only 5% of the AI chips already shipped are actually running.
The other 95% sit idle, waiting on power and data centers that don't exist yet.
But the demand is real...
Bot and agent traffic just passed human traffic on the internet for the first time, 57.4% machine against 42.6% human.
The companies selling that intelligence are scaling faster than any software category in history.
The problem is everything underneath them.
The power and data centers to run all those chips can't be built fast enough, and neither can the gear inside the rack. That physical ceiling caps how fast every layer above it can grow.
To find where the money actually lands, @WhiteCollarExit uses one map:
The 5-Layer AI Cake that NVIDIA's Jensen Huang built to track value through the AI economy.
1. Energy: the electricity, cooling, and power that run everything above it.
2. Chips: the GPUs and hardware that energy feeds.
3. Infrastructure: the data centers, storage, and networks that house the chips.
4. Models: the foundation models that run on that infrastructure.
5. Applications: the AI products and use cases where the money finally shows up.
Each layer pulls its demand from the one before it.
And at the top, agents are taking over the enterprise.
AI began with individuals, consumers on ChatGPT and developers on Claude - but that phase is ending.
At its Build conference in late May, Microsoft started rebuilding Windows as a runtime for AI agents, with an "Agent Computer" that lets agents write code and operate cloud PCs inside set rules.
NVIDIA showed off the RTX Spark, a one-petaflop chip built to run agents on your desktop instead of in a data center.
A whole new business is forming just to babysit these agents: Coralogix raised $200M in early June at a $1.6B valuation to watch them for hallucinations, runaway costs, and security holes.
More agents means more inference, which means more chips, which means more power.
Drop down to the models, and that's where the IPO money is rushing in.
The numbers are hard to believe. Anthropic filed its S-1 on June 1 at a $965B valuation, with revenue running near $47B a year, up from about $10B a year ago.
Dario Amodei says the growth has beaten Anthropic's own forecasts by 8x. Eight of the Fortune 10 are now Claude customers.
Stack Anthropic and OpenAI together and they already out-earn Google Cloud and Azure on a run-rate basis, with some estimates passing AWS by year-end.
Then there's SpaceX. The old record IPO raise was Saudi Aramco - pulling in $29.4B. SpaceX went for more than double that, powered partly by xAI and a compute business pulling in over $2B a month from its Google and Anthropic deals.
@WhiteCollarExit is still passing on all of it.
Run-rate isn't the same as durable, profitable revenue, and the model layer is priced entirely on growth.
Anthropic committed up to $100B to Amazon for compute, with a matching setup at Google, so part of every lab's revenue is the cloud provider paying the lab while the lab pays it right back.
From the outside, a $47B run-rate with weak retention looks identical to one with strong retention, and the two are worth wildly different amounts.
Nobody can tell which is which until the S-1 shows the real margins.
Chinese open-source is squeezing the labs from the other side.
Alibaba's Qwen passed Meta's Llama in total downloads, and a16z estimates that 80% of startups building on open models now run Chinese ones like DeepSeek, at a fraction of the cost.
That eats straight into model-layer margins.
The SpaceX deal has its own catch. The IPO sells only new shares, so no insider sales on day one. Founders Fund turned a $20M bet from 2008 into roughly $182B.
The public is buying in after the bet has already won, right before the December lock-up expires and early backers can finally offload billions.
Underneath all of that sits the infrastructure, and it can't keep up.
JPMorgan pulled satellite images of planned data centers and found that more than 60% of the capacity meant to be finished in 2027 hasn't even broken ground.
For 2027, only 6.3 GW are under construction against 21.5 GW announced.
The reasons are physical.
Grid hookups in the US now take three to five years.
The large transformers these sites need carry 18 to 24 month wait times.
Jensen Huang says a 1 GW AI factory now costs $50B to $60B and is heading toward $100B, and Goldman Sachs puts total AI spending at $8T over six years.
Software ships in weeks → concrete and grid connections move on construction timelines.
If you can't build faster and you can't get more power, one lever is left: pull more compute out of every watt you already have.
That's why the chips layer is shifting. The old way of moving power through a data center wastes about 17% of it as heat before it ever reaches the chip.
A new 800V DC design erases most of those losses, lifting efficiency from around 83% to over 92% while using 45% less copper.
NVIDIA's new Vera Rubin platform just entered full production on that design, shipping this summer to Amazon, Google, Microsoft, and Oracle, with 2.5 to 5x the inference speed of today's chips.
A Hopper rack drew about 40kW. Blackwell pushed it to 120kW. Vera Rubin racks will hit 600kW to 1MW, a 25x jump in three years.
At that density, three parts suddenly matter as much as the GPU itself:
The power semiconductors that handle the voltage, the optical fiber that moves the data, and the materials that insulate against the heat.
You don't have to guess which companies make them, because Jensen Huang has been buying them himself.
He's confirmed stakes in Corning and Lumentum for fiber, a power-chip deal with Infineon, an R&D tie-up with Qnity on materials, and a public nod to Marvell that moved its stock 34% in a single day.
His own money is tracing the bottleneck in real time.
At the bottom of the stack sits the hardest limit of all: power.
Only 5% of deployed GPUs are switched on. The other 95% wait for electricity.
The industry is maybe 15 to 20% into an $8T buildout, and the grid is already straining at 15%. Adding power to the grid is slow, full of multi-year permits and queue backlogs.
Going around the grid is fast. Solid oxide fuel cells can be deployed in 55 to 90 days and skip the grid entirely, and they fit the new chips perfectly, since Bloom Energy's cells already produce 800V DC, the exact voltage Vera Rubin racks run on.
Regulators are piling on too.
Alberta added a 2% levy on hardware for big data centers, New Jersey now makes large loads fund their own new power, and at least 18 US states have bills creating special rates for heavy users.
Off-grid power is becoming a requirement just to get these sites online.
Adding it all up...
The IPOs everyone is chasing sit at the very top, capped by a physical layer that can't grow fast enough to feed them.
The real winners are sitting underneath, in the companies squeezing more compute out of every watt and solving the power problem, while the headlines focus on SpaceX.
@WhiteCollarExit has built his whole portfolio around that idea, with every position mapped to a layer of the cake.
He made real moves this month, trimming one big winner and buying others into weakness, and he's hunting in 3 new areas for his next buys.
If you want the full portfolio, the exact positions, and the 3 areas he's buying into next, the link is in the first comment.
We at Milk Road called Nebius, Credo, Bloom Energy, AAOI, AMD and others before their big run ups.
You can join us for just $1 and leave anytime you want!
https://t.co/WqBpzQXqch
The dangerous part of the AI CapEx trade is that the thesis can be right while the stocks need to digest!
AI demand and the build-out are real, but the easy phase is probably behind us.
Think about hardware, the fact that we cannot power datacenters as we would like to and the grid not scaling fast enough.
Also, datacenter operators get increasingly more pushback from society, as people do not like, actually fear AI!
>60% of datacenters to be operational in 2027 have not yet broken construction ground (according to JPM satellite data).
= we cannot deploy the chips we are producing!
Finally, efficiency is a point. The longer a bottleneck persists (e.g., memory) the more likely companies will put attention on it to find workarounds.
= we may need less spend on given layers in the stack!
So, I don't wanna sound smart and be bearish as I think AI remains a structural trade, but the obvious CapEx winners may need time to digest the run.
I'm spending more time looking at the next layer of the trade which the market hasn't priced.
That's the efficiency and reliability trade in an 800VDC world and increasingly downstream AI adoption and economic rails for agents.
I need you to stop scrolling and try to understand something.
@jvisserlabs hasn't been perfect on everything, but he has called markets much more accurately than most. For decades.
The whole market has been two steps behind him on AI investing since this all started.
What is Jordi buying right now?
Ethereum.
Jordi Visser is buying $ETH.
Not loudly. Not even aggressively. He's just calmly building a position while everyone else top blasts AI.
Read this over and over until you understand what's happening here.
Two of the most prominent macro analysts alive, who built their careers at Goldman and Morgan Stanley, agree on this.
Raoul said Ethereum was, "fking obvious."
You don't have to agree with them on everything. I'm not telling you what to do.
What I am saying is simply, try to understand this.
Over the past few weeks, the data behind Bloom Energy has started to shift...
Our PRO analyst, @whitecollarexit, flagged five factors driving the change:
1. GPU utilization is reportedly running near 5%.
2. More than 60% of planned 2027 data center capacity hasn't started construction.
3. The Silicon Data index shows the frontier model premium stalling.
4. DeepSeek cut token costs by 75%.
5. Ramp's spending data shows buyers moving to cheaper open-source models.
In his view, none of that kills the power thesis.
What it does is raise the odds that the market starts questioning how much data center capex actually shows up in the near term. 👇
Our PRO analyst @whitecollarexit called Bloom Energy $BE on Feb 17, at ~$131 a share.
The stock ran as high as $322.83, a 146% gain from his entry - but on Tuesday, he started selling.
Here's why (save this)...
His original thesis was pretty straight forward:
AI data centers need huge amounts of electricity, and the public grid takes 3 to 5 years to deliver it.
Bloom's fuel cells put power on site in about 90 days.
Hyperscalers were never going to wait - they were going to pay for speed.
And the market caught on to this fast.
Bloom posted 130% revenue growth in Q1, signed a supply deal with Oracle, and saw its stock jump 20% on the report.
By late April it was @whitecollarexit's top-performing holding, up more than 100% since he posted his entry to PRO members.
He held anyway, because he believed the trade had a second leg...
Bloom's platform runs on 800V DC, the same power architecture NVIDIA's Rubin Ultra rack systems are expected to need in 2027.
As AI racks grow from 0.1 MW toward 1 MW each, older power setups won't be able to keep up.
@whitecollarexit argued Bloom was built for where the hardware roadmap was headed, and that the market hadn't priced that part yet. So he sat through the double.
But over the past few weeks, the data started to shift...
@whitecollarexit flagged five factors driving the change:
1. GPU utilization is reportedly running near 5%.
2. More than 60% of planned 2027 data center capacity hasn't started construction.
3. The Silicon Data index shows the frontier model premium stalling.
4. DeepSeek cut token costs by 75%.
5. Ramp's spending data shows buyers moving to cheaper open-source models.
In his view, none of that kills the power thesis.
What it does is raise the odds that the market starts questioning how much data center capex actually shows up in the near term.
Bloom is the most capex-sensitive name in his book.
It's a direct bet on the AI buildout running at full speed.
If spending plans slip, it gets hit first and hardest.
So on June 9, he trimmed. He took profits on part of the position, held the rest, and kept the 800V DC thesis intact.
The next day, Bloom dropped 8.6%.
The timing came from a framework @whitecollarexit has been writing about for months:
→ The AI buildout is a sequence of bottlenecks.
→ The market already repriced chips and memory.
→ Power is getting repriced now.
Once one layer gets fully priced in, the money moves to the next one.
The cash from this trim is already earmarked for that next layer.
PRO members saw every step of this trade in real time:
The February entry, the exact position size, the reasoning for holding through +100%, and the trim before the recent drop.
They'll also be first to see where the freed-up capital goes.
If you want that alert the moment it hits, the link is in the first comment below 👇
@AndreasSteno The export data shows chips are being shipped, not necessarily that they are all being deployed.
@Aaron_552 @MilkRoadAI My argument is not about demand!
The export data shows chips are being shipped, not necessarily that they are all being deployed!
We at Milk Road called Nebius, Credo, Bloom Energy, AAOI, AMD and others before their big run ups. You can join us for just $1 and leave anytime you want!
https://t.co/WqBpzQXqch
The SpaceX IPO is not really a space trade.
It is an AI infrastructure trade.
SpaceX becomes the independent (sovereign) stack AI can run on:
Terafab chips.
xAI compute.
Grok models.
X data.
Starlink connectivity.
Starship launch.
Orbital data centers eventually.
The problem is timing.
This reminds me of Tesla in 2019. Robotaxis, autonomy, robots... The vision held up, but here in mid-2026 none of it shows up in the P&L yet.
Similar setup with SpaceX.
Phenomenal company, expensive IPO, and the big revenue unlock sits somewhere between 2030 and 2050.
You're making a decade-long bet that Elon's AI + space stack actually comes together.
My angle: I hold $TSLA and I'm betting on a merger for exposure. SpaceX needs Tesla's future FCF to fund the build-out, which makes it likely in my view.
Market potentially realising some headwinds for AI buildout coming from:
- Silicon Data Index shows first dip (token cost)
- >60% of 2027 datacenters not yet broken ground
- Ramp’s June data shows shifting toward open-source
Aka physical bottlenecks slowing the buildout and demand shifting to lower open source models?
Do you see any issues in the (broader) trade from:
- Silicon Data Index shows first dip (token cost)
- >60% of 2027 datacenters not yet broken ground
- Ramp’s June data shows shifting toward open-source
Aka demand physical bottlenecks slowing the buildout and demand shifting to lower open source models?
Do you think that trades around efficiency, aka everything that increase compute output per watt (e.g., power semis, chemicals (materials), etc.) will actually benefit from this?
Because, if we are in a power constrained world, squeezing everything out of the given volume we have becomes more important?
Trends for you
Most Popular Users
Elon Musk
@elonmusk
240.3M followers
Barack Obama
@barackobama
119.2M followers
Donald J. Trump
@realdonaldtrump
111.6M followers
Cristiano Ronaldo
@cristiano
109.7M followers
Narendra Modi
@narendramodi
106.9M followers
Rihanna
@rihanna
97.4M followers
NASA
@nasa
92.1M followers
Justin Bieber
@justinbieber
90.7M followers
KATY PERRY
@katyperry
87.2M followers
Taylor Swift
@taylorswift13
81M followers
Lady Gaga
@ladygaga
72.6M followers
Kim Kardashian
@kimkardashian
69.6M followers
Virat Kohli
@imvkohli
69.2M followers
YouTube
@youtube
68.6M followers
Bill Gates
@billgates
63.6M followers
The Ellen Show
@theellenshow
62.5M followers
CNN
@cnn
61.9M followers
Neymar Jr
@neymarjr
61.8M followers
X
@x
60.9M followers
Selena Gomez
@selenagomez
60.3M followers