$NBIS is tackling one of AI infrastructure’s biggest bottlenecks: power.
Nebius has activated its first Bloom Energy fuel cell power block at its 300 MW New Jersey data center site, marking a step toward dedicated, on-site electricity generation for its AI operations.
Building data centers and securing GPUs isn't enough. AI cloud providers also need reliable power to bring capacity online and deliver computing services to customers.
The multiyear Bloom Energy arrangement could help Nebius address that challenge as it expands its infrastructure footprint. If the power setup supports dependable operations and strong utilization, it could strengthen the company’s ability to turn capacity into recurring AI-cloud revenue.
Fuel cells and large-scale data center development come with substantial costs and long-term commitments. Nebius still needs to demonstrate that its expanding infrastructure can generate enough revenue and cash flow to justify the investment.
The activation is real operational progress, but it doesn't yet prove the economics will work at scale.
Elon Musk just delivered another blow to the neocloud trade: “Pointless.”
CRWV, NBIS and IREN extended their declines after Musk said SpaceX has no interest in acquiring a neocloud company, warning that more than 200 players could face consolidation and failures.
The overnight reaction was immediate: 🔻 $CRWV : -2.3% 🔻 $NBIS : -2.3% 🔻 $IREN : -2%
The bigger concern is funding. Musk argued that debt markets have become far less accessible for companies without substantial revenue streams independent of AI data centers. SpaceX, he said, is in a different position.
Venture capitalist Chamath Palihapitiya echoed the warning, highlighting weak capitalization, potential data-center delays and the risk that expensive computing capacity could sit underutilized in 2027.
That’s the real issue for neocloud investors. AI demand can be enormous, but demand alone doesn't guarantee that every infrastructure provider will earn an attractive return on billions of dollars in spending.
Still, this isn't proof that every neocloud will fail. Companies with strong customer commitments, reliable power, disciplined financing and high utilization could emerge stronger as the sector consolidates.
Morgan Stanley sees ~160 GW of U.S. data center power demand through 2029 against an insane ~77 GW shortfall leaving nearly half of that demand at risk of going unpowered.
That’s a massive opportunity for $GEV with its sold-out gas turbines and $BEB with onsite fuel cells that can generate electricity without relying on turbines.
And getting that power into AI data centers creates another opportunity for VRT in power delivery and cooling, FPS in transformers and prefabricated electrical systems and ON and NVTS in improving efficiency through SiC and GaN.
$SPCX CEO Elon Musk says we’re going to see “a lot of stranded compute in partially complete data centers” as GPUs arrive faster than power infrastructure needed to run them.
That sounds bearish but could actually keep GPU rental prices higher for longer as companies with power already secured benefit from tight supply while idle chips lose value with every new generation.
This is what I mean when I say $AAOI has a massive opportunity but an even bigger execution risk with Glassdoor showing just 1.7 stars and 1% CEO approval.
When management is trying to ramp monthly production from ~150K to over 650K transceivers keeping experienced workers becomes just as important as having strong customer demand.
$BTC Has meanwhile moved back to its 2 year moving average price after deviating below it for some time.
In my previous tweet I mentioned how Bitcoin tends to. historically trade at good high timeframe value when it is below that level. The big question is usually how long it remains below that area. In this case it didn't take a long time because since then, price moved up from $63K to $87K at its recent high in just 3 months.
Good reason why I always like to keep looking at these high timeframe "indicators" to determine when BTC is at historically good value levels.
Most of this gap opened up after April as token demand accelerated.
Think about what this means for $NBIS , $IREN and $CRWV .
AI compute demand is pushing rental prices beyond depreciation expectations, strengthening the case for sustained pricing power and stronger returns on GPU infrastructure.
The market is still underestimating what happens when demand keeps outpacing available capacity.