Hashrate / ASICs: A collection of recent writing in case you want to be informed on hashrate, ASICs, and ASIC production (aka the arms dealers of BTC Miners) The Hashrate Almanac: Why hashrate always goes up, the hashcurve, forecasts, etc x.com/penny_ether/status/197… A study of ASIC CapEx / TAM / margins / economics / etc: Basically, selling ASICs is awesome. x.com/penny_ether/status/196… About $BTDR's ASIC manufacturing prospects (selling and self-mining): x.com/penny_ether/status/196… Some estimates on ASIC TAM for the coming year: x.com/penny_ether/status/197… Bonus: Historic hashprice-adjusted network hashrate CAGR: x.com/penny_ether/status/197… -------------- Also, we just adjusted up 6% to ~1,080 EH/s, in a period where hashprice was lower. Up and to the left is not good for any miner. In case you're curious how to read this chart: - Up: miners making the same amount of money, but now require BTC to stay higher-for-longer - Up and to the right: expected. hashprice up, marginal hashrate comes online. Too much up is bad... that hashrate can sting later on. - To the right: bull case for miners. however, improved economics means increased chances miners buy more ASICs (eg: up will happen later on, though at some maximum rate limited by infra and ASIC manufacturing) - Down and to the right: super bull case, but won't last for long: the hashrate that came off will be back soon. - Down: good, but the hashrate that came off will be back soon - Down and to the left: expected.. hashprice down means marginal hashrate comes offline - Left: BTC down, miners still making money though - Left and up: Misery. Why did all you greedy MFers buy hashrate and ruin it for the rest of us?

$MARA's O&O hashcost is barely any cheaper than what they pay for hosting.. even if you attribute 50% of SG&A towards "hosting". (And this doesn't count the costs of initially acquiring those MWs.) Additionally, YTD, the amount of O&O hashrate has barely budged... despite SG&A increasing significantly. Hosted ticked up, though. But go ahead and believe that SG&A is so much because they are building "the next iteration of $MARA" or whatever else. Maybe it's for a push for O&O AI/HPC, despite them throwing dollars at undisclosed French AI/HPC cashflows, talking down AI/HPC for over a year (and CSPs in the latest earnings call), and their only engagement in AI/HPC is "10 racks of AI in Granbury". I'd have no qualms with them shelling out millions more in SG&A and the MWs if they had a clear plan for monetizing it... but so far all I hear is "inference at the edge" and "electrons are oil"... and "10 racks of AI" in Granbury. If any $MARA bull can clearly explain how they will get cashflow from their MWs, I'm all ears. Every other miner has said they will pursue colocation or do inhouse AI/HPC... $MARA instead says it's a secret because they don't want people to copy them yet again. LOL

Slight correction to the first chart.. I forgot to update $/BTC (present) to account for current network hashrate.

As for why I "pick on" $MARA... they're currently the most bullshitty miner out there. The amount of wool they try to pull is ridiculous. I do not abide. I've given the same treatment to other miners.

$MARA - Likely mining at a loss right now. They're spending ~$43/PH/s per day, but hashprice is ~$41.50. At present, likely paying more cash to mine BTC than the BTC is worth, despite having spent ~$1b on ASICs last year. These figures omit other cash expenses... R&D and taxes other than on income (used to be an SG&A expense, but they now break it out) and other write-offs. -------- Curiously, they are either over-reporting their hashrate (resulting in a lower self-reported hashcost), or their hashrate is not producing as much BTC as it "should" (bad luck, poor block construction, etc). They provide this table that shows their costs (excluding SG&A), from which you can infer the average daily hashrate: The math: $31.30 per PH/s per day, times 92 days, is $2,880/PH/s for the quarter. With $145,054 in costs, that comes to 50.37 EH/s. However, you can also infer the average daily hashrate from the amount of BTC mined. Eg, if you ran 1 EH/s on an FPPS, you'd have earned 45.04 BTC for the quarter. $MARA mined 2,115 BTC, so the "effective" hashrate they had was ~46.96 EH/s. Lower than 50.37 EH/s they claim. They mined less BTC than 50.37 EH/s would have. So.. bad luck, or their numbers or off, or they primarily mined during below-quarter-average hashprice conditions. Part of a consistent trend: It's possible I have some numbers wrong here -- but I doubled checked and the numbers and math seem right. Doesn't help that one of the last remaining pureplay miners stopped reporting average operational hashrate over a year ago. But, even when they did (2024 - Q1, Q2) the numbers didn't tie out. ------ As it pertains to "energized hashrate" vs the above "effective hashrate" -- here's how it's panning out for them: Leaving tons of revenue on the table. ------ What ultimately matters is the amount of BTC they mine and what it costs to do such. In that regard, refer to the first table provided. They truly suck at mining. ------ Lastly, it's been mind blowing listening to them recently. They seem to actually believe they are doing a good job, and they seem to think people will lap up anything they say... which is a lot. Anyone remember 2PIC? Dead. How about "Twin Turbo"? Not a treasury anymore: "In prior periods, we presented certain bitcoin yield metrics to illustrate trends in the growth of our bitcoin holdings. Beginning this quarter, these metrics are no longer presented, as management determined they are no longer meaningful given our decision to sell bitcoin from production to fund operations and our focus on the active management of our digital assets rather than passive bitcoin ownership." Also no comment on CTO being gone or not. (Isn't it material information if he is leaving or has left?) I get the sense that being the market cap leader for so long has deluded them into thinking "easy mode" will have lasted forever. Almost every sentence uttered is bullshit, inconsistent, contrary to their actual business strategy, or not remotely related to their bottom line. Of course, you might not know this because nobody has the balls (or opportunity) to grill them on this. What information is provided is sparse. Didn't even mention the debt situation... let's just say the early 2030's are not looking pretty for them if BTC doesn't cooperate. I expect that current management will have sailed off into the sunset by then. Fred and Salman have sold a combined $12.6m in shares since June 2024 -- voluntarily (eg, not for tax reasons). -------------- Ran out of images, but their EV is now ~$2.5b (if you assume all notes convert) or ~$3.0b if you assume the ones ITM or NTM convert, or ~$3.5b if you assume none of them convert. Seems quite rich for ~50EH/s of high-cost hashrate (with abysmal operating history) that pulled in $51m this quarter, and is pulling in $0 today. Then again, perhaps they find a way to make HPC and/or power generation work. I wouldn't trust them with a dollar of my capital.

$MARA - Delusional

An example of ongoing semi fab tech, fueled by the investment (and production) of chips for AI: piped.video/watch?v=vkx2zIan… Along with other advancements, this will inevitably result in lower J/TH ASICs (as has always been the case). Long network hashrate.

Wasn't too far off

I feel like you guys would find this proof interesting: @dampedspring @LynAldenContact

A simple proof that ATM-issuance treasuries have a rational mNAV of 1.0x: Assume a treasury has 10m BTC, and will procure 11m more BTC. At what mNAV would the BTC Yield equal (or exceed) the mNAV? There is only one answer: 1.0x, and 0% BTC Yield. If you use any higher of an mNAV, the BTC Yield will not be enough. Eg, if you assume 1.10x, it is impossible for there to be 10% BTC Yield. Not enough BTC. Therefore, it would not be rational for anyone to pay 1.10x. So, at 10m BTC, the rational mNAV is 1.0x. Now assume the treasury has 5m BTC, and will procure 5m more to get to 10m. (We know that at 10m the mNAV should be 1.0x.) So: At what mNAV would the BTC Yield equal (or exceed) the mNAV, assuming this mNAV holds the entire time while they procure 5m BTC? Once again, there is only one answer: 1.0x. Now assume it has 2.5m BTC and will get to 5m. Etc, etc. All the way down to 0 BTC. In other words, if a treasury just relies on a positive mNAV and equity issuance for BTC Yield, there is no rational justification for it to have a positive mNAV. This applies to any strategy that solely relies on the mNAV premium to generate BTC Yield: ATM, convertible notes, and warrants. Not surprising, given there is no cashflow and no actual value generated... but IMHO this is an interesting proof nonetheless. ----------------------- The caveat here was in bold: any strategy that solely relies on mNAV premium. If there is cashflow (which is then used to by BTC, or not) the assumption fails. As I see it, fixed income products provide cashflow -- eg, the company issues them at a higher credit spread than what actually is realized -- so long as BTC number go up. If you loan me $1m at 10% and I invest that $1m and get 25%, that's actual cashflow. Has nothing to do with mNAV. Best of luck to all BTC Treasuries.

ChatGPT's product retention curves is a product manager's wet dream. Their 1 month retention has skyrocketed from <60% 2yrs ago to an unprecedented ~90%! Youtube was best-in-class with ~85%. 6mo retention is trending to ~80%. Rapidly rising smile curve. Generational product.

$MARA - Without HPC, they seem quite overvalued. Some quick stats. Enterprise (ex-BTC) valued at ~$4.4b. What do you get when you buy this? - Produced $51m in Q2. In Q2, hashprice was $51.15, hashcost was $38.19 ($41.20 if you include R&D and other). Hashprice is now $46 and on with diff adj may drop to $44. - At $51m EBITDA, EV/EBITDA stands at ~20x. This means even if BTC doubles every 4 years, and there is ZERO upwards adjustment to difficulty (which averages around 40% CAGR), and they cease diluting right now... they'd produce as much cash as you pay for the enterprise, in 20 years time. - 60 EH/s of ASICs. ~50 EH/s of them running on average. At 18.5 J/TH. 28 EH/s of them were bought and installed last year, costing ~$843m. (ROI will almost assuredly be negative. They paid $30/TH/s and are making ~$0.004/TH/s per day) - Let's assume 60 EH/s. That is currently ~750 BTC per month. That's currently a BTC Yield of about 1.4% per month. A year from now, that'd be 1.2% per month. Factoring in 40% CAGR on difficulty going up, it'd be more like 0.7% per month. - The above is going off of 370m fully diluted shares. It's assuredly higher than that, given they have $1.6b ATM left, and must be fully aware this currently valuation is a gift. - Stock based comp is fucking insane. The performance alone is embarrassing enough (sitting on an ivory tower blowing smoke up everyone's ass while ops/costs suck) ... but to get paid more in a day than the average American makes in a year, for mainly distributing shareholder's cash to bankers and ASIC manufacturers, would personally cause me existential dread. - $2.875b in debt that will not convert under $20/sh. (BTC might save them, here... but you'd better off owning BTC). Even if we pretended like the next halving would never occur, why pay $4.4b for this? If you apply a reasonable, but still stupidly high, EV/EBIDTA of 5x, that shreds ~$3.5b of value. Puts them at around $15/sh. Which I'd still not pay for, but you do you. --------- They've rallied ~30% since Sep 1, during which hashprice (their only revenue) dropped 15.56%. Ignoring the volatility of BTC, difficulty adjusted up by ~16%... meaning they will permanently mine 16% less BTC per day than they did on Sep 1, from now until forever. And that decay seems to show no signs of stopping. The only reasonable explanation for their rally is the market valuing their HPC/AI prospects. Which include: - Not having a CTO, as far as I can tell. - 2PIC is dead, as far as I can tell. So much for this (March '24: "We look forward to rolling out MARA 2PIC700 and setting a new standard for data centers with our next generation technologies.") At any rate, could happen. I'm sure they will at least announce some BS related to HPC/AI pretty soon. Will the market will forget that as recently as August (KeyBanc Forum) Fred has been throwing shade at HPC pivoters? "And they think, oh, I can convert something that was trading at, let’s just say, dollars 1,000,000 a megawatt into something that trades at 10,000,000 a megawatt overnight by just saying it could be an HPC spot. The truth is, of all the people that have pivoted, only really two have been able to announce deals." I recall some other talk that was, paraphrasing, "heh, $MARA is so strong that we can rely on our mining. Huge vertical stack. The other companies are probably jealous of us -- they'd love for us to share our technology!" Fred likes to talk about "inference at the edge"... but none of the curated pre-screened analysts (whose firms make money dumping $MARA shares) seem to have the balls to ask: WTF does this have to do with $MARA? They seem content with "wow, you sound smart." CAN YOU JUST FOR ONCE SAY WHAT THE FUCK $MARA IS GOING TO DO TO MAKE MONEY. (Note the distinction: $MARA, not management) We get it... you know how to SPEND money. --------- Lastly, a friendly reminder of MARA is not a treasury company. They made a PR post on linked in to make that clear. Banger music in the background. (linkedin.com/posts/maraholdi…) If you like BTC yourself go buy some, not the stock. Otherwise you are paying ~$4.2b for their actual operations that are optimistically worth like $1.2b or something. Yes, you're also getting their "datacenters" and contracted MWs. What do you think they are worth? Without Fred making a compelling case for anything (rather, he's insisted on throwing shade at HPC), I'd say they're worth the underlying operations of mining BTC. Oh, and a windfarm. You can mine BTC on some days! Anyway... $MARA will probably announce some HPC initiative soon, which will be bullshit just like every other thing the talk about. Good luck with that.

$MSTR - Update. Currently at 1.434x mNAV, via ATM, they need to sell 21,405 BTC ($2.3b*) of stock to generate 1% BTC Yield. They need to sell 521,283 BTC ($57.0b*) of stock to generate 19.75% BTC Yield (mid-way point), or 1.5m BTC ($160.4b*) of stock to generate 43.40% BTC Yield. Between Aug 18 and Oct 19 they sold $1159.7m of stock, implying a 43.40% BTC Yield would take ~23.4 years.** This assumes all ATM proceeds are used to buy BTC, and not to pay back convertible notes, or dividends (>$550m/yr). It also assumes mNAV would stay at 1.434x the entire time, despite the fact subsequent buyers would see the above numbers increase exponentially. Eg: BTC needed for 1% BTC Yield would double every 23.34% BTC Yield accrued. If MSTR did buy another 1.5m BTC through ATM issuance, and mNAV stayed at 1.434x, they would need to sell 70,434 BTC of stock to generate 1% BTC Yield, and 4.8m BTC of stock to generate another 43.40% BTC Yield. -------------- On the preferred front, $MSTR would have to sell 277,941 BTC ($30.4b*) of prefs to 1.434x BTC/sh. At $100/sh, that's 304.0m of them. Between Aug 18 and Oct 19 they sold around $255.9m of them, implying such a task would take ~20.1 years.** At $9/yr dividend, that would incur $2.7b/yr in additional dividend liabilities. -------------- * Based on current price of BTC. If BTC goes up, so does this figure (proportionally). ** Assuming sales (in $'s) remain flat: Longer time if BTC price goes up. Shorter time if BTC price goes down.

"We are 2.5 weeks into Q4 and SWC has already delivered +0.5% BTC Yield without including all warrants, and +1.5% if you include all warrants." Can you explain how your BTC per *share* ends up higher when you include all warrants as *shares*? Seems like the denominator would get larger.

"Even if you take a somewhat pessimistic view and assume that SWC will generate only +10% BTC Yield each quarter into the future, how do you value that? In 4 years, you would have multiplied your BTC position by 4.6x." Jesse, how many BTC-worth of shares would people have to buy in order for your BTC/sh to go up 4.6x?

$CRWV / $CORZ - A fun little 10m study to guess $CORZ price if vote is no. Take $CORZ beta to $WGMI prior to June 26th, and apply that to $CORZ price at the time, and change in $WGMI between then and now. That comes to around $21.02/sh. An interest "gut check" here is as follows: - we know the price if deal is "yes" ($CRWV x 0.1235) - we estimated the price if deal is "no" (above) So, given current price, what odds might the market be pricing to yes / no? Comes to around 38.4% yes... which matches very closely to the below poll, which of the voters that didn't abstain, was 35.3% I don't really place much confidence in this $21/sh price. I just thought it was interesting that the implied odds of yes/no (based on simple beta analysis) matched what the poll said. In all likelihood, $CORZ actual "beta" isn't applicable to the basket of $WGMI.. but more like some basket of HPC-focused miners, but adjusted for the fact that $CORZ already had a large amount of MWs contracted, whereas those others were getting rerated from 0 MWs HPC. In any event... 🍿

As much as I love seeing everybody win, there is something extra satisfying about making money on a red day. Biggest position is ZQ Q3 2026, and gold. If AI works, weaker jobs. If AI infra build slows, weaker jobs. Scared of inflation? Gold. Weaker dollar? Gold. (Copper is interesting here as well... as it captures upside to AI as well as inflation/weak dollar). To be clear what ZQ contracts are: market is pricing in about 3.00% FFR in Oct next year. A full 1.00% less than now. These contracts go up in value as the expectation for FFR goes down (eg, expectation of more cuts by then). What effects that expectation? Listen to what the Fed says and how they are changing their messaging. Fed is increasingly signalling that cuts would not be "mission accomplished" cuts.. but "bad cuts". Eg, we're back to the "fed puts" era. (But they are worried about inflation.) The way it's felt for the past three months: Market wants to rip, and will rip. AI exposure good. Only time it doesn't rip is when it panics over the economy. And when that happens, "fed puts" pay off. Icing on the cake for "fed puts" is "trump puts". Trump wants lower rates. Trump will bully Fed dissenters. JPow out in May. Mid-terms next year (gotta pump those numbers). Even if more cuts is a bad choice, I don't mind having exposure to this kind of irrationality. I just cannot see a scenario where we get less cuts unless the economy is absolutely ripping (equities up). So, these ZQ contracts have provided the perfect barbell. It'll work until it doesn't, then you can all throw pie in my face.

If $CRWV wants the market to believe they will not renegotiate with $CORZ, they should agree to pay a $270m "renegotiation fee" if the vote is "no" and then a new deal is cut. Payment has to go to a third party -- me, obviously.

$CORZ - Situational Awareness loading up sec.gov/Archives/edgar/data/…

The rare 🚨