@daniel_koss @amasad Yes and a lot of them just don’t understand AI.
English
Idiosync Capital
748 posts
Idiosync Capital
@Idiosync_Cap
Esoteric alternatives, some credit, some public equities. Flâneur.
Katılım Mart 2023
784 Takip Edilen355 Takipçiler
@amasad 100%, this shows the average AI investors is completely clueless.
English
$NVDA should be pumping on the K3 news. Instead it’s down.
English
If I have to rehash my original DD on $NBIS - this statement captures the biggest differentiator when comparing to all the other datacenter names back in 2025. That gap has increased considerably since then and the importance of software orchestration to margins is ever more obvious.
Dr. Tomislav Marinovic@DrTomsLens
Ophir Nave, COO of $NBIS: “Our disciplined, diversified approach — from owned data centers to asset-light partnerships — together with robust demand for our high-value software stack, will enable us to build a sustainable AI cloud business with strong and durable margins.” According to Ophir, there is strong demand for Nebius’s high-value software stack, while the asset-light model should significantly improve margins. Turns out building software for AI infrastructure wasn’t such a dumb idea after all. Many people thought it was.
English
Idiosync Capital retweetledi
The frenzy around Kimi K3 highlights how well timed $NBIS' acquisition of Eigen AI was.
As open-source models become more capable and are released at an accelerating pace, the ability to optimize them quickly and serve them efficiently becomes a major competitive advantage for any inference platform.
Eigen AI brings exactly that. In a recent Kimi K2.6 benchmark, Eigen led the field with 265 output tokens per second vs. 224 for the next-best provider, CoreWeave, despite using B200 GPUs rather than the newer GB300 platform.
This benchmark relates to an earlier version of the model, so we still need to see how it performs on K3, but the direction is telling.
Token Factory’s value will depend not only on offering the latest models, but also on running them faster, more efficiently, and with better economics than competing platforms.
Kimi K3 is exactly the kind of release that makes the strategic value of the Eigen acquisition even clearer.
$NBIS has been saying from the beginning that the market would move in this direction.
English
@daniel_koss Most people can’t even fathom the compute requirements that will come once the physical robots take over. 100-1000x.
English
Here’s what I believe AI bears get wrong at the most fundamental level:
“There will be an overbuild.”
Many AI bulls respond that demand far outpaces supply. I agree, but that answer is lazy because it does not explain why.
The deeper misunderstanding is that AI bears fundamentally do not understand what data centers and AI actually are.
They treat AI factories like clothing factories: there is only so much clothing people can buy. Produce too much, and prices collapse or inventory goes unsold.
Intelligence does not work that way.
First, demand for intelligence is effectively infinite. A unit of intelligence is never wasted. At worst, you allocate more thinking to existing tasks.
Second, the TAM for intelligence is the entire global services economy. Once AI massively outperforms humans across every domain, it becomes exponentially larger.
A more capable intelligence will not be constrained by the economic ceiling of lower-intelligence tasks.
Compared with future AI systems, humans are primitive monkeys.
Today, we live in the monkey economy, and many smart monkeys argue that we do not need more humans because all the bananas are already being picked and all the trees have already been climbed.
Saying there will be an overbuild does not merely misjudge supply and demand. It reveals a fundamental misunderstanding of intelligence’s potential and its enormous TAM, which I suspect not a single person currently estimates optimistically enough.
This may be one of the rare moments in history when almost everyone, including the dreamers, is underestimating reality.
Usually, it is the opposite.
The future is bright for the entire AI trade imho.
I'm very bullish all quality names like $NBIS $DRAM $MU $SKHY $BE etc.
English
@TheStockerMan You should add the value of their subsidiaries (ClickHouse, Avride, Toloka).
English
I have no idea when $NBIS will stop bleeding.
But here’s how I look at it:
$NBIS is guiding toward 7-9B of annualized revenue (ARR) by the end of 2026.
Let’s assume they land around the middle at 8B.
The market is currently valuing the company at roughly 43.61B.
That means investors today are effectively paying around 5.45x forward revenue if $NBIS reaches that target.
For one of the fastest-growing AI infrastructure companies in the world, I don’t think that’s expensive.
Let’s compare it.
Mature software companies growing 15-20% often trade around 5-8x sales.
High-quality cloud software growing 30-40% can trade 10-15x sales.
AI infrastructure businesses with years of hypergrowth ahead should arguably deserve a premium if they can prove margins and execution.
So let’s run a few scenarios.
6x Sales
$8B revenue = $48B valuation
That’s about 10% upside from today’s valuation.
In other words, if $NBIS simply hits management’s revenue target and the market only assigns an AVERAGE multiple, there isn’t much downside from valuation alone.
Now look at a more optimistic case.
8x Sales
$8B revenue = $64B valuation
That’s roughly 47% upside from today.
10x Sales
$8B revenue = $80B valuation
That’s roughly 83% upside.
Now imagine management exceeds expectations.
Suppose they reach $10B in annual revenue over the following years while expanding EBITDA margins as utilization improves.
Even an 8-10x sales multiple would imply an 80-100B company (+100% upside)
And remember…
This isn’t a company growing 20%.
The biggest reason the stock has sold off isn’t because demand disappeared.
It’s because investors are questioning whether $NBIS can finance and execute such an aggressive expansion.
I’m focused far more on revenue growth and execution than I am on the day-to-day price action.
Because if the numbers end up where management believes they can, today’s valuation could look very different a few years from now.
English
Idiosync Capital retweetledi
The mega bull case for AI infrastructure would be *if* market share shifted away from certain frontier labs with 90%+ inference margins toward cheaper models, whether open-source or closed.
It would increase the ROI on AI spend for end customers by increasing intelligence per dollar, which would drive incremental token demand. Margin dollars would effectively get redistributed from the frontier labs to AI infrastructure providers. The infra winners would be those with the lowest per token cost and the winners at the model layer would be those with the highest token efficiency.
There are many reasons Jensen is so focused on open source, but this is likely the most important one as I think he is probably less worried about a monopsony these days. Lower margin % at the model layer = more margin $ at the infra layer all else equal.
With SpaceX and Meta being vertically integrated and possessing the #3 and #4 models respectively it is more possible than ever. Note that Grok 4.5 is ahead of Fable for some useful tasks at a much lower cost, so ranking them #3 is conservative.
This is not happening yet. Cheap, mostly open source tokens are likely the majority of volume today but the majority of economic value is still accruing to the most intelligent models. Might change though.
We will see.
Cassandra Unchained@michaeljburry
This is true as I have heard this from contacts in the Valley. Goes with my pinned post. The AI race is shifting from bigger models to cheaper, smarter systems cnbc.com/2026/07/10/the…
English
I believe this is one of the reasons why IPO for OAI and Anthropic has become a moving target.
If your models are constantly in the winds of competitors which cost an order (or 2) magnitude less, your frontier model business will always be under threat.
There is an argument that there will always be token demand for the absolute frontier, and that open weight will command the most token demand, but this seems lazy to me.
If performance metrics are so close, the cost won’t justify or the economic output of that top end demand won’t be valued attractively because it will always be under some threat of displacement by a newer, more powerful OW model.
Zephyr@zephyr_z9
K3-type models are bad for frontier AI model companies and margins (K3 won't threaten OAI, cuz Sol is still very price competitive, but Opus will be evaporated) but they are great for cloud/TaaS (token-as-a-service) providers and infra Moonshot cannot keep up with the demand, and they will invest a lot on acquiring compute (good for A shares AI infra plays)
English
@realroseceline @InvestNorthwise It will only get even more difficult once you consider the newly introduced asset light model. Awesome work though.
English
This is probably the best $NBIS model I’ve ever seen (by far) — you did a great job @InvestNorthwise 👏. That said, it’s only as good as its assumptions. Small changes in revenue per mw, utilization, margins, depreciation, dilution, capex, or the exit multiple will produce dramatically different outcomes.
Models don’t predict the future, they translate assumptions into numbers. The further out you forecast, the more humble you should become. I still stand by what I’ve said: to me, based on the risks, margin of safety, etc $NBIS becomes attractive in the low $100s/share (roughly a $25–30B market cap). 🌹
Northwise Project@InvestNorthwise
northwiseproject.com/nbis-stock-for… needs to be updated for the new $NBIS business line announced this week, but otherwise pretty fresh. We track every individual site and build everything on top the biggest variables up for debate being ARR per MW and enterprise customer mix (we raised these considerably based on interviews from the executive team over the last few months)
English
News of Kimi K3 dropped today. Better than Opus 4.8, not as good as Fable 5. Model competition: relentless. Feels like top open-weight Chinese models are closing the 3 month gap. Gap now feels like 1.5 months.
English
Idiosync Capital retweetledi
AI will ultimately lead to the collapse of wealth scarcity.
Fable 5 and GPT 5.6 Sol charge premium margins today because frontier capability is scarce. Extrapolating out, what costs dollars per million tokens now will run locally for pennies within a few years time. And eventually, cognitive labour approaches a cost of $0 - paralegals, analysts, entry level developers. Add humanoids, and physical labour follows the same trend.
At that point, "wealth distribution" stops meaning what economists think it does because goods and services whose cost is mostly embedded labour (which is most of them) deflate toward the cost of energy and raw materials. Then, the average person's purchasing power explodes even if their income never moves.
I agree with Elon's UHI (universal high income) theory here, where rather than UBI, you'll have UHI delivered through lower prices across the board. All driven by AI and robotics which will flood the economy with goods and services at a rate faster than the money supply can catch up to it.
Therefore, the real distribution question isn't labour vs. capital. Rather, who controls the two inputs that stay scarce throughout this paradigm AI shift: energy and compute.
Eventually, the nations that build terawatts and fabs will win, while the nations that write AI regulations without building either will lose. This is why I'm extremely bearish on Europe (despite being European myself), which is wrapped up and insulated by risk aversion and red tape.
Conversely, this is the reason why I am extremely bullish for the US, Korea, Taiwan and Japan - and is why those equity markets are on an upward trajectory long-term compared to European markets which will suffer a slow, painful demise over the next few years.
English
Great thread. I remember 3 things from this book: 1) pay top end, 2) how hard would you fight to keep me on the team, 3) trust people to do right thing (on expenses etc) and audit to sniff abusers. Good book, but specific to creative talent.
Gappy (Giuseppe Paleologo)@__paleologo
"No Rules Rules" is not a book about Netflix or reinvention. To me, the main lesson from Netflix's success is that talent is underrated. Sure, there's a lot of fluff and stuff I missed. But the talent discussion is interesting and has implications for hedge funds. A long 🧵 1/
English
@daniel_koss @babyfolio @itsalasdairmann Congrats gents, this is great and look forward to following and exchanging more
English
Big personal announcement!
@babyfolio, @itsalasdairmann and I are joining forces to build the best investor community on the internet: Edelbridge Alpha.
We believe in one thing above all: performance.
Ali: +320% YTD
Baby: +270% YTD
Me: +360% YTD and now +3,365% since Jan 2024
There is endless investing content online. Much of it is generic, consensus-driven, hard to act on, or AI-repackaged by people who never show their true results.
Edelbridge Alpha is built to be the opposite: original alpha, actionable, short and specific posts, transparent numbers and accountability.
What subscribers get:
- High-conviction investment ideas
- Actionable trade ideas
- Full portfolio updates
- Peer-reviewed deep dives
- Financial models and price targets
- Industry insights
- Macro thoughts and analysis
- Exclusive insights from executives at companies we follow
Each of us writes with our own style, opinions, and approach. No committee takes! No hedged consensus "should, could, would, might". We share real price targets and our actual numeric assumptions! We'll be right or wrong and you can tell! :)
And the three of us are only the founding contributors. We will selectively invite more high-performing investors and traders on X, many of whom you likely already follow ;)
The bar: consistent performance, independent thinking, and original analysis that actually generates alpha.
Paid subscribers will (next week) also get optional access to a private, actively moderated Discord to discuss the ideas together, red team them, openly discuss risks and build conviction or spot scary risks or when a thesis breaks.
Spam, insults, and low-quality noise get removed instantly, even from paying members. The room stays high-signal.
I started investing because I refused to watch inflation eat my savings, and it became an obsession because, like competitive gaming, there is an objective scorecard. But the messages from readers who paid off debt or started funding their kids' education made me realize this could be so much bigger. That's what I want to work on now!
Link to the Edelbridge Alpha Substack is in my bio.
Join us from day one!
English
@EndicottInvests @InvestNorthwise FWIW it seems like so far we are holding 220. This will clear a large call resistance and for the next 3 weeks there is little resistance above (some pockets but nothing compared to what we have cleared today). If we hold by EOD you should be adding. I did that today.
English
Definitely an exciting statement. But it’s very difficult to extrapolate. The best every shareholder can do is make an assessment on connected capacity until then. Refer to @InvestNorthwise projections and DYDD. Biggest contributors to capacity until then should be Vineland and Finland. Use a revenue multiplier / MW and get to an estimate.
English
Okay lets do some fun speculation on $NBIS
Marc Boroditsky, the CRO, said the conversation in next 13 months will be Nebius generating 10s of Billions in revenue.
I would say 10s of Billions is a minimum of $20B
13 months from now is August 2027.
Now i think there are three ways you can frame this:
1. All he is simply saying is a conversation about them generating 10s of billions of revenue, which could mean more towards a guide for 2028, not that they WILL be. (Most Likely)
2. "Revenue" could be ARR, which means $20B of ARR by August 2027, if you extrapolate that out, that's $30B of exit ARR at the end of 2027. (Could happen but definitely a bullish scenario)
3. Literal revenue. 10s of Billions of revenue, meaning $20B revenue MINIMUM by August 2027. (highly, highly unlikely)
What do we think?
Molly O’Shea@MollySOShea
Marc Boroditsky (@marcboroditsky), Chief Revenue Officer of @nebiusai ( $NBIS ), says the next phase of AI is shifting from.. 'What are AI natives building?' to 'When are enterprises adopting?' "Lots of discussion about the CapEx, power & stock prices, which I think, at the end of the day, is a bit of a distraction because this is a marathon that we're running." "Actually being able to drive that enterprise adoption, that's the critical next stage for the entire industry." On the data center build-out: "The biggest challenge that many of our competitors are facing is being single-threaded, which is strange when you're in a multi-threaded parallel processing industry." Looking ahead.. "The next 12 months in AI is an eternity."
English
@Funmentalist The last time I saw this number for institutional holding it was 52%.
English
$NBIS
Institutions +23% in one month
Guess why we went down to $181
They are not buying at ATH like retail
Sorry guys, you got played again if you sold
Top Tech AI Stocks 🤖 🔥 📈@TopTechStocksUS
$NBIS The Institutions meteoric rise last month is equal to them stealing retail shares with all the shakeout. Hold on to your shares $NBIS is going to release major news in 2-3 weeks max!
English
English
@TheLongInvest I have a request from @reelyphishy to unblock him. He says he’s sorry and he’s now a believer. 😬
English
Idiosync Capital retweetledi
Mark Zuckerberg just confirmed it:
$META “needs all the computing power it can get.”
But in an AI market starved for compute, external offers are “so high” that Meta may sometimes rent out capacity instead of using it internally.
This was never bearish.
M. V. Cunha@mvcinvesting
What triggered this selloff in AI infrastructure stocks was the "excess capacity." Algos saw the headline and immediately assumed the AI buildout was slowing down. But is it really excess capacity if you can resell it at higher prices than you paid for it? I don’t think so.
English