R McGrath's Avatar

R McGrath

@themcgrath.bsky.social

He/Him. Uncle. Senior millennial. General Purpose Nerd. Once-owner of @modwiggler.com Now I just work there. He's just this guy, ya know?

36 Followers  |  45 Following  |  112 Posts  |  Joined: 24.11.2024  |  1.9458

Latest posts by themcgrath.bsky.social on Bluesky


feels cruel for Bluesky to have a Popular With Friends tab, if any of us were popular with friends we wouldn't be on Bluesky

16.02.2026 08:53 โ€” ๐Ÿ‘ 945    ๐Ÿ” 159    ๐Ÿ’ฌ 9    ๐Ÿ“Œ 14

It's so annoying when a bigot dies and we get to listen to people spouting a lot of pious crap about them. The damage they do will always outlive them.

16.02.2026 17:37 โ€” ๐Ÿ‘ 32    ๐Ÿ” 5    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

if someone is saying shit like "it's lowkey biphobic that you support AOC even though she won't legalize cannibalism" or "you realize that expecting your room to have windows a luxury belief right" what you're seeing are the charged particles being shot off from polycule fission

13.02.2026 05:15 โ€” ๐Ÿ‘ 1957    ๐Ÿ” 360    ๐Ÿ’ฌ 24    ๐Ÿ“Œ 26

I assure you; private sector information workers being forced back into the office are every bit as annoyed.

08.02.2026 04:26 โ€” ๐Ÿ‘ 5    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

If I were Mayor of Ottawa, I would drive the @octranspo.com management team to Westboro, dump them on a random side street and give them 1 hour to get back to City Hall using only a Presto card

Whoever makes it back in time gets to keep their job

04.02.2026 15:19 โ€” ๐Ÿ‘ 116    ๐Ÿ” 24    ๐Ÿ’ฌ 8    ๐Ÿ“Œ 13
Post image

For every retweet I will make Zuckerbergโ€™s creep glasses 5% thicker

01.02.2026 16:46 โ€” ๐Ÿ‘ 845    ๐Ÿ” 500    ๐Ÿ’ฌ 6    ๐Ÿ“Œ 25
Post image

FANTASY/DRACOLCH.GIF

23.01.2026 13:20 โ€” ๐Ÿ‘ 3726    ๐Ÿ” 440    ๐Ÿ’ฌ 45    ๐Ÿ“Œ 36

Both links in your original tweet cause that issue for me - but only on my desktop, not on mobile (on wifi behind the same firewall).
Incident ID: 1353000530404278711-482029458233560391 should let your correlate my error in your logs

05.01.2026 20:52 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
Post image

your links don't work

05.01.2026 19:32 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

๐Ÿ‘๐Ÿ‘๐Ÿ‘

03.01.2026 14:20 โ€” ๐Ÿ‘ 3556    ๐Ÿ” 571    ๐Ÿ’ฌ 46    ๐Ÿ“Œ 16

As a Canadian, I only feel this way about Toronto. Calgary and Quebec are standouts but I feel like they have defined Canada enough that they are now in the fold.

01.01.2026 21:12 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
Preview
MIT report: 95% of generative AI pilots at companies are failing Thereโ€™s a stark difference in success rates between companies that purchase AI tools from vendors and those that build them internally.

MITโ€™s NANDA initiative found that 95% of generative AI deployments fail after interviewing 150 execs, surveying 350 workers, and analyzing 300 projects. The real โ€œproductivity gainsโ€ seem to come from layoffs and squeezing more work from fewer people not AI.

20.08.2025 04:51 โ€” ๐Ÿ‘ 4625    ๐Ÿ” 1893    ๐Ÿ’ฌ 75    ๐Ÿ“Œ 439

It's only been 6 months, but dang it feels weird to suddenly have no beard

06.12.2025 01:30 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Because "go to college" = "get diploma" = "get higher paycheck"

Somehow it all comes back to devaluing the Trades.

01.12.2025 06:38 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

School admin won't let teachers flunk students at that rate. If universities started refusing to let these students pass/graduate, it would be reflected on the school & teacher's reputation - undeservingly so, but that is the world & metrics we have built for ourselves.

01.12.2025 06:37 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

All chess is 4D chess that's why they have that time clock thingy.

28.11.2025 18:56 โ€” ๐Ÿ‘ 895    ๐Ÿ” 152    ๐Ÿ’ฌ 15    ๐Ÿ“Œ 3
Post image Post image Post image Post image

Writing code is the act of weaving a series CPU instructions. Every row of pixels is a weft of a byte groups retrieved from computer memory.

20.11.2025 18:29 โ€” ๐Ÿ‘ 55    ๐Ÿ” 10    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Aww yiss! Name the time and the place!
And the setting, and starting level, and race/class restrictions

20.11.2025 01:31 โ€” ๐Ÿ‘ 1    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Today is the day to repost this meme about International Men's Day, one of the most unironically wholesome memes that has ever been made.

Shoutout to all my fellow champs, chiefs, and kings. โœŠ

19.11.2025 18:38 โ€” ๐Ÿ‘ 8018    ๐Ÿ” 3169    ๐Ÿ’ฌ 57    ๐Ÿ“Œ 95

That sounds like an easy thing to fix โœ‹

20.11.2025 00:02 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0
A LinkedIn ad showing Jensen Huang shaking hands with some vague fashion executive:

L'OREAL
6,204,444 followers
Promoted
We're thrilled to announce a collaboration with NVIDIA to supercharge beauty with next-gen Al.

WELCOME TO
NVIDIA

L'ORร‰AL
GROUPE

A LinkedIn ad showing Jensen Huang shaking hands with some vague fashion executive: L'OREAL 6,204,444 followers Promoted We're thrilled to announce a collaboration with NVIDIA to supercharge beauty with next-gen Al. WELCOME TO NVIDIA L'ORร‰AL GROUPE

I donโ€™t want to seem out of touch but I donโ€™t actually understand the economy anymore.

18.11.2025 03:23 โ€” ๐Ÿ‘ 12751    ๐Ÿ” 2245    ๐Ÿ’ฌ 484    ๐Ÿ“Œ 555

In lieu of my usual rewatch of You're The Worst, I have the perfect film response to a recent lack of closure;
The Banshees of Inisherin.

It has everything; Adorable & loyal animals. Homesickness for somewhere you've never been and for people you may never see again. Wailing against the inevitable.

13.11.2025 08:56 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Weird that we have a planet-wide magnetic defense system that self activates when the sun decides to attack but it is very pretty

12.11.2025 02:21 โ€” ๐Ÿ‘ 2604    ๐Ÿ” 407    ๐Ÿ’ฌ 15    ๐Ÿ“Œ 10

"I'm so, so tired of ads for Ozempic, Viagra, and AI garbage, please give me something else"

A finger on the monkey's paw curls

"Draft Kings is your number one destination for every sport!"

10.11.2025 06:49 โ€” ๐Ÿ‘ 1891    ๐Ÿ” 170    ๐Ÿ’ฌ 23    ๐Ÿ“Œ 4

Sometimes when people ask me why Iโ€™m wearing a mask I say Iโ€™m traveling or have some important thing soon and canโ€™t afford to get sick and miss it and thatโ€™s pretty much always true but I think it would be nice if it were more normalized to just say โ€œI donโ€™t want to get sickโ€ and leave it at that

08.11.2025 16:49 โ€” ๐Ÿ‘ 3633    ๐Ÿ” 484    ๐Ÿ’ฌ 6    ๐Ÿ“Œ 109

Gosh, explaining complex ideas by way of Dialogues? How insufferable and unprecedented.

Huh? Plato? Socrates? Never heard of them.

02.11.2025 19:04 โ€” ๐Ÿ‘ 2    ๐Ÿ” 0    ๐Ÿ’ฌ 1    ๐Ÿ“Œ 0

EV owner lesson learned;
I thought not using my brakes would make the pads & disks last longer, but they decided to just corrode instead. Turns out they need regular exercise to keep clean and healthy.

31.10.2025 21:46 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0
So, big tech has spent hundreds of billions of dollars of capital expenditures, and much of those expenditures are on GPUs. These GPUs, once installed, immediately begin decaying in value, and the more you use them, the more likely they are to crap out, and even when sitting idle, these GPUs have some power consumption โ€” somewhere between 7% and 14% of their power in the case of the H100 โ€” making them a consistent burden on the infrastructure of any purchaser whether they have customers or not. This becomes an even bigger problem for more expensive GPUs like the B200, which has an idle power draw of 140W per GPU. 

Extrapolate that out to thousands, tens of thousands or hundreds of thousands of them, or to the 50,000-per-building of them inside Stargate Abileneโ€™s NVIDIAโ€™s , and youโ€™ve got yourself a problem. 

Even without customers, even without IT load, these servers sit burning power, and while a company could turn them off and boot them on for a period of time, these racks are often run in clusters โ€” servers of GPUs that are synced together using high-speed networking โ€” and while I canโ€™t find anything firm as far as how long it takes to set up a cluster, this presentation on cluster design from the CEO of Lambda details a clusterโ€™s sheer complexity, involving discrete networking, software layer and storage methods just to get the bloody thing working. 

In simple terms, it may be expensive to leave them on, but itโ€™s going to take a bunch of time and energy to restart these things, meaning that they are more than likely left on, slowly burning energy.

Imagine how bad that is at a scale of hundreds of thousands of GPUs? 

Yet things might be a little wonkier, as the reality may be that while hyperscalers have customers, those customers might not be using all their GPUs at all times. Although there may be some customers that agree to a minimum spend or throughput, itโ€™s hard to imagine that these contracts are enough to surmount the destructive margins. 

And when โ€ฆ

So, big tech has spent hundreds of billions of dollars of capital expenditures, and much of those expenditures are on GPUs. These GPUs, once installed, immediately begin decaying in value, and the more you use them, the more likely they are to crap out, and even when sitting idle, these GPUs have some power consumption โ€” somewhere between 7% and 14% of their power in the case of the H100 โ€” making them a consistent burden on the infrastructure of any purchaser whether they have customers or not. This becomes an even bigger problem for more expensive GPUs like the B200, which has an idle power draw of 140W per GPU. Extrapolate that out to thousands, tens of thousands or hundreds of thousands of them, or to the 50,000-per-building of them inside Stargate Abileneโ€™s NVIDIAโ€™s , and youโ€™ve got yourself a problem. Even without customers, even without IT load, these servers sit burning power, and while a company could turn them off and boot them on for a period of time, these racks are often run in clusters โ€” servers of GPUs that are synced together using high-speed networking โ€” and while I canโ€™t find anything firm as far as how long it takes to set up a cluster, this presentation on cluster design from the CEO of Lambda details a clusterโ€™s sheer complexity, involving discrete networking, software layer and storage methods just to get the bloody thing working. In simple terms, it may be expensive to leave them on, but itโ€™s going to take a bunch of time and energy to restart these things, meaning that they are more than likely left on, slowly burning energy. Imagine how bad that is at a scale of hundreds of thousands of GPUs? Yet things might be a little wonkier, as the reality may be that while hyperscalers have customers, those customers might not be using all their GPUs at all times. Although there may be some customers that agree to a minimum spend or throughput, itโ€™s hard to imagine that these contracts are enough to surmount the destructive margins. And when โ€ฆ

And when I say โ€œdestructive margins,โ€ I mean so much more than just the power to run and cool them. Thereโ€™s the construction of the data centers to house them, the necessary maintenance, and, of course, any and all debt used to buy them. 

While I can (and will!) say that for every $50,000 GPU bought, big tech needs to make another $100,000, I believe that the hole created by the GPUโ€™s existence is likely larger than just $100,000, with every construction delay, power outage, or moment of idle compute adding dollars to the hole for each and every GPU. 

Need a very real example? The Information reported recently that Oracle lost $100 million between June and August 2025 on rentals of NVIDIAโ€™s Blackwell chips:

In the three months that ended in August, Oracle lost nearly $100 million from rentals of Nvidiaโ€™s Blackwell chips, which arrived this year. Thatโ€™s partly because there is a period between when Oracle gets its data centers ready for customers and when customers start using and paying for them, the documents show. Itโ€™s not clear what causes the gap or how Oracle plans to shorten it.

So, while we canโ€™t say for sure, thereโ€™s a chance a large chunk of these losses were caused by the spread-out deprecation of those Blackwell GPUs โ€” after all, its most recent quarterly earnings included $1.35 billion of depreciation, right?

Sidenote: For context, Oracle reported depreciation of $804m in the same period last year, and depreciation of $3.867bn for the whole of FY2025. This should give you a sense of how rapidly these GPUs are becoming a millstone on its earnings. 

โ€ฆis there any way to actually recover that $100 million through the margins on these chips? Oracle only has 4.5 years of depreciation on these Blackwell Chips, and The Information reports that those GB200 server racks have a negative 100% gross margin according to internal Oracle documents. If thatโ€™s the case, we have a much, much bigger problem, but even if the margins arenโ€™t literally negativeโ€ฆwill therโ€ฆ

And when I say โ€œdestructive margins,โ€ I mean so much more than just the power to run and cool them. Thereโ€™s the construction of the data centers to house them, the necessary maintenance, and, of course, any and all debt used to buy them. While I can (and will!) say that for every $50,000 GPU bought, big tech needs to make another $100,000, I believe that the hole created by the GPUโ€™s existence is likely larger than just $100,000, with every construction delay, power outage, or moment of idle compute adding dollars to the hole for each and every GPU. Need a very real example? The Information reported recently that Oracle lost $100 million between June and August 2025 on rentals of NVIDIAโ€™s Blackwell chips: In the three months that ended in August, Oracle lost nearly $100 million from rentals of Nvidiaโ€™s Blackwell chips, which arrived this year. Thatโ€™s partly because there is a period between when Oracle gets its data centers ready for customers and when customers start using and paying for them, the documents show. Itโ€™s not clear what causes the gap or how Oracle plans to shorten it. So, while we canโ€™t say for sure, thereโ€™s a chance a large chunk of these losses were caused by the spread-out deprecation of those Blackwell GPUs โ€” after all, its most recent quarterly earnings included $1.35 billion of depreciation, right? Sidenote: For context, Oracle reported depreciation of $804m in the same period last year, and depreciation of $3.867bn for the whole of FY2025. This should give you a sense of how rapidly these GPUs are becoming a millstone on its earnings. โ€ฆis there any way to actually recover that $100 million through the margins on these chips? Oracle only has 4.5 years of depreciation on these Blackwell Chips, and The Information reports that those GB200 server racks have a negative 100% gross margin according to internal Oracle documents. If thatโ€™s the case, we have a much, much bigger problem, but even if the margins arenโ€™t literally negativeโ€ฆwill therโ€ฆ

Premium Newsletter tomorrow: Meta, Google, Amazon and Microsoft must add $2 Trillion in AI revenue by 2030 or they wasted their capex.

The cost of data centers GPUs may make it impossible to make a profit from AI. Here's $10 off annual.

edzitronswheresyouredatghostio.outpost.pub/public/promo...

30.10.2025 22:01 โ€” ๐Ÿ‘ 991    ๐Ÿ” 158    ๐Ÿ’ฌ 20    ๐Ÿ“Œ 15

I'm starting to think The Great Beard experiment has run its course. Might keep it until my birthday, but a more youthful face would support the energy I'm trying to bring into the new year.

29.10.2025 14:30 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

Why does the ejecta cloud appear to be polar? I would imagine that, if anything, the poles are the last bastion of the collapsing magnetosphere,so the ejected gas cloud would be closer to a torus shape.

29.10.2025 12:15 โ€” ๐Ÿ‘ 0    ๐Ÿ” 0    ๐Ÿ’ฌ 0    ๐Ÿ“Œ 0

@themcgrath is following 20 prominent accounts