Andy retweeted
$NBIS
Nebius CEO Arkady Volozh at GTC Berlin:
“Yes, the prices are still there, demand is still there. Actually, it’s the main thing which is happening on the market, demand is out there. There is a lot of problems we see. How to finance, how to build, how to finance the thing, where to get the chips."
Is there a possibility in 1-2 years computers start shipping with a very rudimentary OS ran by an LLM that builds interfaces / tools on the go with no windows / macOS?
What is the first principles reason that we need an OS as developed and rigid as windows or macOS are? What would it mean for $MSFT and $AAPL ?
@lxe rebuilt Linux for ~$5k worth of tokens. In 6 months, this will be $500 - why do consumers need an operating system in a years time if their desired outcomes can be achieved with low-level computer code and agents working in the background, communicating through a constantly evolving interface
@dhh created @omarchy in ~6 months, and yes it was done with a team of top engineers, but what if a bunch of agents can build a ‘you-specific’ OS bit by bit that’s tailored to your taste, tasks, etc (assuming compute is cheap/accesssible) with an interface that evolves to you and the specific task at hand?
Last night, a swarm of Opus 5.5 agents working for 9 hours and burning 254.2 million output tokens were able to fully reverse engineer and decompile the Linux source code
Full source archive available here kernel.org
Grab it before they take it down
Interesting take, but what’s implication for $ that neoclouds $NBIS & $CRWV charge/MW? They don’t do cost+ pricing like consulting firms, the price is set by demand. NBIS even auctions contracts. Agree on supply/cost, but demand/price independent of credit costs?
I've been critical in the past of AI compute indexes and the willingness to price compute based on the spot rental price, but now that the AI reset is sinking into the repricing of AI buildout debt its time to realize that the cost of capital is what determines the price of compute.
The most important number in GPU pricing right now isn't $/GPU-hour. It's the gap between what a hyperscaler pays to borrow (~90bp over Treasuries) and what a neocloud pays (~1,100bp).
The cost of compute is now set by the cost of capital, and the spot rental rate no longer tells you much.
The price of an H100 isn't one number. The same chip in the same rack costs a different amount depending on whose balance sheet it sits on. A ~1,000bp funding gap is a structural cost advantage that no amount of operating skill closes.
For example, take one GPU-hour apart. Most of it isn't power or space. It's paying back the chip, the interest on the loan that bought it, and the return the equity expects. A $40–55k all-in GPU, 75% debt-funded, at ~7,000 utilized hours a year is a financing instrument that happens to do matrix multiplication.
As the +500bp on neocloud debt (10%→15%) adds $0.20–0.30 to the break-even cost of every GPU-hour, which is then added on to the rental price.
That's 8–10% of the H100 index, from a credit spread alone, driven almost entirely by the cost of capital.
That means that the compute spot index is a derivative of the AI infra debt credit curve and the squeeze on neocloud margins.
Super crisp and impressive take
I've been critical in the past of AI compute indexes and the willingness to price compute based on the spot rental price, but now that the AI reset is sinking into the repricing of AI buildout debt its time to realize that the cost of capital is what determines the price of compute.
The most important number in GPU pricing right now isn't $/GPU-hour. It's the gap between what a hyperscaler pays to borrow (~90bp over Treasuries) and what a neocloud pays (~1,100bp).
The cost of compute is now set by the cost of capital, and the spot rental rate no longer tells you much.
The price of an H100 isn't one number. The same chip in the same rack costs a different amount depending on whose balance sheet it sits on. A ~1,000bp funding gap is a structural cost advantage that no amount of operating skill closes.
For example, take one GPU-hour apart. Most of it isn't power or space. It's paying back the chip, the interest on the loan that bought it, and the return the equity expects. A $40–55k all-in GPU, 75% debt-funded, at ~7,000 utilized hours a year is a financing instrument that happens to do matrix multiplication.
As the +500bp on neocloud debt (10%→15%) adds $0.20–0.30 to the break-even cost of every GPU-hour, which is then added on to the rental price.
That's 8–10% of the H100 index, from a credit spread alone, driven almost entirely by the cost of capital.
That means that the compute spot index is a derivative of the AI infra debt credit curve and the squeeze on neocloud margins.
A great take on memory stocks from @ren_stocks in my perspective - the R/R at these levels is astonishing and the quality of these businesses is only becoming stronger every day. The stock moves seem to be pure sentiment to me. $MU $SNDK $SKHY
$MU $SNDK $SKHY
Samsung and SK Hynix are both down more than 50% from their June highs, and they now trade at 2–3x their 2027–28 earnings.
In every previous memory upcycle (2001, 2008, 2018, 2022), the low end for these two was 4–7x. Their price-to-book is now near 1x, a level that usually shows up only at the bottom of a real bust, with warehouses full of unsold chips, DRAM prices cut in half and quarterly losses.
None of that is happening this time.
Memory inventories at both companies fell below 10 days in September, and KB Securities has their earnings staying near record highs through 2028, with only a single-digit dip that year.
Every new generation of GPUs and custom chips also needs new memory: HBM, SOCAMM, LPDDR5 and enterprise SSDs. As the installed base gets refreshed, each upgrade cycle turns into another round of orders.
The market is pricing in a bust but the demand is ever increasing.
Long $MU and $SNDK, and not selling.
This is a good point - I wonder if bitcoin:native is a short now. Imagine tomorrow a model can hack it - is it going to 0 then?
Some people on X realised that OpenAI's math drop didn't contain much cryptography. @dpaleka notes that a few more fields are under-represented: largely the ones that are useful for AI research
Replying to @finkd @johnternus
Bird’s eye view, if this pace continues, we won’t need data centers in 10 years. #datacenter
The GPUs becoming obsolete argument doesn’t hold up to logic or what’s truly happening in business’s today. There are plenty of tasks with varying compute and latency requirements so workload can be spread efficiently
Replying to @RealMattMoney
“Asset backed”….GPU’s rapidly depreciating asset. We went from Blackwell to Vera in under 2 years…
Good factual on the characteristics of different debt tranches but 15% cost of unsecured debt impacts cost of equity (and thus multiples) thru risk of default, not scare of assets’ quality or underlying business strength - equity investors are buying stocks not rev/profit sharing
CoreWeave's debt, explained simply, because I like to speak plainly:
Secured debt = the mortgage. Backed by the house. The GPUs and the contracts they serve are the collateral.
Unsecured debt = the credit card.
Backed by a promise.
Everyone is panicking that $CRWV UNSECURED yield to MATURITY hit 15%
But a credit card rate spiking doesn't mean the house (or in this case the GPUs/data centers) lost value.
The secured lenders or the ones who underwrote the actual GPUs and contracts aren't selling here.
Now zoom out, because this is analogous to the entire AI market in wrapped up in a single company.
Trillions in secured, asset-backed debt are loaned against GPUs, data centers, power contracts, and signed capacity deals with hyperscalers. This money doesn't care about stock prices or investor sentiment.
It gets paid if the contracts pay. (Sounds pretty similar to $next if you ask me…)
Below that in line is unsecured debt, a bet on the companies themselves.
On top of everything is the equity, a bet on the story.
So when you see scary headlines about "AI debt," ask one question: which layer?
Yields spiking on unsecured debt and equities dropping while secured debt sits calm simply means the market is repricing the story.
The assets are fine.
The buildout keeps getting funded because the secured debt holders still believes in the collateral.
The day that changes… well the day secured lenders start questioning GPU values and contract quality, is the day the AI trade actually breaks and we should run for the hills.
Until then, volatility at the top of the stack (unsecured, equities) is just the shock absorbers doing their job.
Agree - stock has history of massive volatility on sentiment and jumping back up to more logical valuation levels. Still plenty of financing runway and no leverage concerns, at least for now
Using a VERY conservative estimate for the value of the non-core business units, $NBIS' core business is effectively trading at ~1-2x YE2026 ARR at today’s market cap, with half of that ARR already pre-booked.
It’s been a long time since the stock was this cheap.
Also, this marks the third time this year $NBIS has experienced a 40%+ drawdown. For stories like this, periodic sentiment resets are normal, and often where the best opportunities emerge.
Good read on $NBIS - top mgmt team paired with best-in-class software product (much more value than just renting GPUs), customer loyalty, and long remaining financing runway.
Despite recent market volatility and the AI selloff, $NBIS remains my highest-conviction investment and is once again the largest holding in our portfolios.
It’s rare to find a generational opportunity at the intersection of seemingly insatiable demand, low-cost financing, and management execution that continues accelerating despite the company’s growing scale.
The Street expects roughly $16B in exit ARR for 2027. Following Q2, our full-company model projected approximately $31B. I increasingly believe even that may prove too low.
Management has discussed short-term capacity deals approaching $50M in annualized revenue per MW. If Nebius can capture more of this demand through short-term auctions rather than locking capacity into lower-priced multiyear agreements, revenue per MW could meaningfully exceed our assumptions.
Combined with accelerating capacity deployment, $NBIS could potentially exit 2027 at more than 2x current Street expectations.
I’ve been deploying significant cash from our flagship and active Growth portfolios during this selloff.
Looking beyond 2027, our models suggest $NBIS could surpass $100B in annual revenue by 2030. I increasingly believe even that extraordinary forecast could be exceeded.
Two developments are changing my expectations.
First, Vera Rubin could materially increase the revenue potential of newly deployed capacity, particularly in short- and medium-term contracts. Alongside Nebius’s expanding software stack, including Token Factory and zero-day model support, the company is moving further beyond commodity GPU rentals.
Second, capacity acquisition is accelerating. Much of our original forecast depended on greenfield developments across the US and Europe coming online in 2028 and 2029. Management’s hybrid approach, incorporating brownfield acquisitions, colocation, and faster near-term deployments, could pull substantial revenue forward.
Higher revenue per MW combined with faster capacity deployment creates enormous upside to our existing scenarios. New management announcements continue strengthening the probability of our bull case.
Of course, none of this comes cheaply. Our CapEx forecasts are also substantially above Street expectations. But unlike the financing concerns recently surrounding $CRWV , Nebius benefits from a strong cash position, attractive debt terms, and significant flexibility across debt, convertibles, equity, and potential subsidiary monetizations.
Management has demonstrated an impressive ability to balance these financing sources without compromising growth.
I suspect the upcoming Q3 earnings could be a major catalyst if management provides formal 2027 ARR and CapEx guidance. Both figures could catch Wall Street badly off guard.
The market continues to underestimate not only the scale of infrastructure Nebius is building, but also the increasingly valuable software platform sitting on top of it.
$NBIS remains my highest-conviction long-term holding, and recent volatility has only strengthened that conviction.














