Compute retweeted
This is a really good question and I’ve done a lot of thinking on this myself. In theory, the cost of intelligence drops down to the cost of an electron as a function of time. So the question is, where is the value accrued at if it’s a race to the bottom?
First it’s important to look at the frontier labs as they are on everyone’s mind and it seems like a headline comes out every day talking about how they are going to fail. It’s important to see that they currently capture their value from the differential of intelligence over its competitors. As time goes on, assuming that these behind them are catching up that differential closes, especially with open source on their tail. Obviously the heads of Anthropic and OpenAI know this, and it’s the reason why they have been so heavily focused on preventing distillation from taking place. It was important that they cracked down on this, despite causing an interim flatlining in revenues, because their entire value proposition relies on it. The next step is to reach full RSI (Recursive Self Improvement) where the models train themselves. I included a graph from Leopold Aschenbrenner’s Situational Awareness paper here to demonstrate what happens when AI’s reach RSI. This will likely be the breaking point away from open source models, and it will give the frontier labs such a decisive lead in the intelligence race.
Now, the question is how much are OpenAI and Anthropic willing to spend to get to this point as soon as possible? The truth is, from a game theory perspective, their entire business model relies on it. On the other end of that curve is the cure to every single illness, physics problem, robotics, etc and that’s where the true value is accrued. How much do you believe they are willing to spend on compute here for the time being?
This brings me back to the original question you asked which was what does it mean for the price of compute/power. It depends on how much value you believe is on the other side of this and how fast they can achieve this. That’s why short term power is so valuable right now. The problem is on the capital side for the time being where many of these projects are not going to be completed due to power, driving up the $/MW and financing rates. But this then goes back to more capital for Anthropic and OpenAI to pay to fund their future commitments. For these reasons don’t see how the price can peak unless lenders stop lending, which the Gov can’t let happen in the current economic state.
Compute retweeted
Beyond our project in Hawesville ( Justified) - 2027 is simply not available. Scarcity! That’s why we were able to lock in a great customer on great terms so quickly for Justified. And I agree - the opportunity for customers is huge. They are not nickel and diming the collocation partner because they need compute yesterday. If anything, they are prepared to spend what it takes to accelerate already contracted projects.
Compute retweeted
Blackwell Ultra (B300) GPU rental prices up 17 consecutive days to a new all-time high, according to Compute Desk's index of verified transactions from U.S. neoclouds.
Compute retweeted
Doomberg AI of course mangled the message. It is this kind of reporting that discredits their brand. Well filtered X (control your inputs!) is far superior.
Hock said compute demand exceeds supply into 2030
The casino line came during a discussion on OAI capital raise, and Hock rejected the idea that it's overvalued. He blamed investors who set extreme expectations w/o understanding the tech. Down in the actual text of the article it reports him saying OAI has a viable biz model at 50bn ARR and that the losses are training spend.
Compute retweeted
“Small public companies are first discovered by smart retail, then by small institutions, then by large institutions, and then finally by dumb retail.”
— @iancassel
Compute retweeted
Last week I shared 10 takeaways from our recent investor meetings. The response was tremendous, and so were the follow-up questions.
Many were with respect to the market; price action, volatility and sentiment. Interest rates. Geopolitics. Financing concerns. Rising costs. Community opposition. There's a lot weighing on the sector.
Listen - I can't control the market. But I can drive how we build this business, allocate capital and create shareholder value.
So let's address a few things head-on.
Compute retweeted
$IREN returns after being on the longest red streaks in the past. Now 7 red days in a row going into Friday.
Average returns:
1 day out: +12.0%
1 week out: +15.5%
1 month out: +15.3%
Compute retweeted
$nuai @StreamDataCentr JV probability decreases daily
Evidence continues to pile up suggesting that $nuai will build its own team in house to complete construction rather than partner with Stream.
Jose Rodriguez and Evan Pierce specifically played a more significant role in the prior earnings call as management wanted to signal to the market that $nuai possesses in-house talent with hyperscaler backgrounds.
The hiring of Doug Hofmockel, who most likely worked with Evan and Jose at ByteDance, as well Doug’s recent LinkedIn post seeking electrical engineering expertise to design hyperscaler data centers paints a picture of a firm evolving in real time and looking to adopt the $iren $cifr $wulf approach of in-house DC construction rather than rely upon third party contractors as heavily.
Lastly, consider how the company commented on negotiations in the past few weeks:
NUAI is actively discussing lease arrangements directly with multiple potential anchor tenants, including one of the world's largest consumers of AI infrastructure. (Most recent 8-k)
That is why I find the most recent analysis from sell side shops that claim 45-50% JV interest to be incorrect as it $nuai appears to be looking for more economic upside post PPA.
Compute retweeted
We heard big news from the Ector County Commissioners Court today about $NUAI and its TCDC project! The judge said that now that the 200 MW power deal is signed, "this is starting to become a reality." New Era asked for no tax breaks and will pay the full rate, which could mean HUNDREDS OF MILLIONS in property taxes, back to the community. The county is getting ready for this to be built, and it's a huge win for the local community. @LovedropV @fnitefiend
Compute retweeted
Built a $NUAI TCDC valuation model with @optionology & @lovedropv.
We modeled 3 lease rates across P1+P2, using 12x after signing & 15x stabilized.
Base case: $22.99 → $33.69/share.
Directional model, not a PT. Assumptions below 👇
Compute retweeted
$NUAI at $7. The asymmetry here is fucking enormous and I don't think the market has figured out what this could become.
I've followed it since Jan. It was quite literally a different co then.
The thesis is simple. AI needs an obscene amount of power and power is the bottleneck.
493 acres in Texas, permits in hand, and a path to 757MW of BTM power. The land and the power to build a massive AI data center campus w/out sitting around waiting for the grid.
Phase 1 has 207MW contracted under a 20 yr PPA w @VistraCorp. Signed. Not classified, not conditional, not pending a study, actually signed.
That's the distinction people are missing. $CIFR just got conditional Batch Zero classification on 3.2GW and ripped 9%. Conditional means projects can still be removed if they fail verification, and ERCOT's final transmission plan for that group isn't due until fall 2027.
Contracted power and conditional power are not the same asset.
Now they need one thing. A tenant.
Mgmt. has guided to an investment grade hyperscaler. That's a small universe, all of whom have said time to power is their number one constraint.
Sign one for a meaningful chunk of the site and this stops being a speculative land + power story and becomes contracted AI infra. Completely different co.
I own plenty of $IREN, $NBIS and the other names everyone already knows. Nothing wrong w/ crowded trades. But this is one of the few I've found recently where the gap between what it is and what it's priced as is this wide.
I own 103,138 shares and the Jan27 $7.50s and wonder if it's enough.
Compute retweeted
I think one of the best things you can do to become a better investor is to learn about everything.
The more you understand history, economics, technology, science and even engineering, the better you understand how the world works and how it is changing. History gives you context, economics helps you understand how money moves through the economy, while science, technology and engineering help you understand what is being built and where the world could be going.
You don't need to become an expert in any of them. You just need to keep learning, because the more you understand about the world, the easier it becomes to connect the dots and recognise opportunities before they become obvious to everyone else.
Compute retweeted
The presumption that the Fed raising short-term rates reduces inflation is predicated on the belief that higher rates reduce demand and investment.
But what if higher rates don’t reduce demand and investment because the demand for intelligence and energy is unaffected by higher rates because winning the race for super intelligence has a near infinite ROI and the demand for compute will remain incalculable.
Why won’t higher rates at this unique moment in history therefore lead to more inflation as interest costs are embedded in everything?
And the problem is compounded as the more the Fed raises rates, the more inflation we will have and the more the Fed will need to raise rates further and so on.
But what if the old models don’t apply to the current paradigm and the Fed is wrong?
I think the Fed might have just made a mistake. Am I right or am I wrong?
Compute retweeted
The ATM solves for the Macquarie covenant. It doesn’t automatically mean they need to sell $100mm in stock. No dilution has happened yet.
Macquarie has to make a credit decision to handle the LOC and wrap it in the deal. That’s the next announcement before Oct 9th.
The real focus should be what they said in the filing today:
“
The Company has prioritized negotiating towards one or more transactions directly with a tenant partner and is focused on discussions with multiple potential tenants, including one of the “world’s largest consumers of AI”infrastructure to serve as its anchor tenant for its flagship project pursuant to a direct lease.
If we enter into a letter of intent with a tenant, it will be non-binding…
“
—Additionally, the Company may continue to explore joint ventures with development partners.
…particularly in the event of a direct lease without a development partner.
Don’t be surprised if NUAi attempts to do this without the JV, taking 100% of the deal and recapitalizing the company with help from Macquarie or the HPC directly.
The JV is no longer the required wrapper for $NUAI.
“Think Bigger”
That’s exactly what the mgt team is doing.
$NUAI announced a new $100M ATM program today.
My view: this is neither purely bullish nor purely bearish.
An ATM is a financing tool, not an immediate $100M dilution.
The company is not obligated to sell any shares, and management can decide when, how much, and at what price to issue.
That said, the market is right to price in some near-term pressure:
• potential future dilution
• additional share supply on rallies
• uncertainty around how aggressively the ATM will be used
At ~$7/share, fully using the program would require roughly 14M new shares, so the dilution is not trivial.
But the other side of the equation matters too.
$NUAI is moving from development-stage planning into actual execution:
• Phase 1 power is now contracted
• permitting continues to advance
• construction readiness is improving
• future project capital requirements are becoming more real
Having access to equity capital reduces financing risk, even if it introduces dilution risk.
For me, the interesting part is how this interacts with my options strategy.
Most of my earlier short puts are now close to expiration or have already captured nearly all of their premium.
So if the ATM announcement creates short-term pressure on the stock — while the underlying thesis remains unchanged — it may create another attractive put-selling window.
That is exactly the type of setup I look for:
temporary price pressure + elevated IV + improving fundamentals
I am not trying to predict whether $NUAI trades at $6.50, $7, or $7.50 next week.
I am asking a different question:
At what effective price would I be comfortable owning more shares if assigned?
If the market gives me attractive premiums at strikes where I am willing to own the stock, I am happy to let volatility work for me again.
The ATM may create a short-term ceiling.
But it may also create the next volatility opportunity.
The key questions now are not simply “Is there dilution?”
They are:
How much is actually issued?
At what price?
And what value does that capital create?
Personal investment view, not financial advice. $NUAI
Compute retweeted
🧵 Northland’s new $NUAI note has a few very important takeaways:
They expect management to spend the next couple of weeks commercializing the initial ~757 MW of TCDC capacity, not just the 207 MW Phase 1.
207 MW Vistra PPA + 550 MW Thunderhead BTM.
That is a much bigger story than a standalone Phase 1 lease.
Compute retweeted
AI infrastructure is the unsexiest part of the sexiest AI industry, as we like to say :)
But this part has its own (unsexy:) innovations too.
One of them is pricing. A more programmatic approach through auctions is the way forward.
We were the first AI cloud to introduce a successful auction.
(I can talk all day about why the marketplace approach doesn’t work that well, btw. Hint: compute is not a commodity!)
And now we have the auction built directly into the platform.
More to come.
Compute retweeted
$NUAI was built for exactly this! Abbott’s letter today paused Texas data center permits until ERCOT finishes its audit and laid out the rules: pay your own electrical infrastructure, don't touch community water, don't lean on the grid.
Ok, now look at what NUAI did this morning: 20 year PPA with Vistra, behind the meter, no grid hookup. NUAI pays for the substation. Closed loop cooling on its own well. Phase 1 needs no air permit!
A lot of Texas projects just went into a queue. But TCDC and $NUAI just got stronger, and should make more big announcements soon. NFA.
Compute retweeted
For those wondering what this means, this is one of the few sites in the US with 207 MW of live energy that is fully approved by the large load interconnection process.
As a result, $NUAI is poised to sign a long term off-take agreement with a hyperscaler or Neo Cloud in very short order.
This 207MW was originally committed to $CIFR and then to a hyperscaler. Both failed and NUAI prevailed. That is a testament to the quality of this management team. In doing so, NUAI now has the upper hand to negotiate a hell of a deal in what is the hottest datacenter market in history.
And it doesn’t stop there. With another 550 MW of on-site generation capacity already contracted, NUAI’s TCDC site will quickly become a mega project that values the company at ~$5-7bn (or greater) vs yesterday’s ~$0.6bn market cap.
Good day to be an $NUAI shareholder. Congrats to all.
Compute retweeted
Don't miss my brand new show with Jordi Visser.
@jvisserlabs says he sees Bitcoin differently than anyone he knows, and that includes Saylor. He agrees it will serve as the collateral of the system, but he disagrees about what gets it there.
We get into the latest AI warnings, whether the data center spending ends the way 2008 did, what actually starts Bitcoin's next bull run, and why he thinks worrying is waste of energy.
Let us know what you think in the comments.
Timestamps:
00:00 Trailer 00:56 Was the AI safety scare coordinated? 6:24 Who really benefits from regulation 9:49 Why the government can't let prices fall 15:18 Is the data center boom another housing bubble? 21:05 Why so few people own so much 26:40 What comes next for Bitcoin and digital money 32:36 The $184 trillion nobody talks about 43:40 The one thing he says survives everything 54:52 Why worrying is the worst investing habit