Dwiki Pratama Putra retweeted
Elon Musk says there's no AI demand issue and explains why SpaceX rents its compute on short-term contracts: AI compute is worth more per watt every year, renewals are coming in at higher prices, and the market is a “cheese shop” with no chips left
“Our belief is that the value of compute per watt will increase over time, not decrease. And the reason for that, I think it should be obvious to people really, is that as you improve the AI software, as the models get better and better, they’re able to, you get more useful intelligence per watt. And the value of that compute is proportionate to the useful intelligence per watt. And that’s why you see things like even old computers like H100s renting for more than they rented when they were new.”
“And I think that’s likely to continue. I think we’re going to face a shortage of superintelligence compute. And that shortage is going to remain for a long time. So, I think the smart move is to do short-term contracts and to essentially place the bet on useful intelligence per watt increasing rapidly, which obviously is done, so that’s what’s happened so far.”
“And also if you look at, say, how are our short-term contracts renewing? Are they renewing at higher dollars per watt or lower? And we see them renewing at higher dollars per watt, not lower dollars per watt. We are also frankly hounded relentlessly by every hyperscaler and by dozens of smaller companies to rent more of our compute. So we could rent out our compute several times over if we said yes to everyone. That’s just the situation here.”
“That’s why I posted that sort of video meme of the Monty Python cheese shop sketch where the guy goes into a cheese shop to get some cheese and they actually have no cheese. And so he goes like, what about this kind of cheese? What about that kind of cheese? And they really just don’t have any cheese.”
“So we’ve actually tried to offload some of our inference compute to other cloud providers so that we can use more compute for training. And we get the cheese shop sketch. You know, they’ll say, like, well, we’ll ask them about what about some, you know, GB300s? Nope, don’t have any of those. What about, you know, some GB200s? Nope, don’t have any of those. What about some TPUs? Nope. What about some Trainiums? Nope. What about some, okay, how about some H100s? And like, well, we might have some overseas somewhere and then they’ll check and come back and say, nope, we don’t have that either. So it’s like ridiculous.”
“Yeah. So I don’t think we’re going to have a demand issue here. And like, it’s not impossible. I’m not saying this will happen. It’s not impossible that we will have a six-month payback period for compute next year.”
Elon Musk says Starlink connectivity revenue has “a clear path to several hundred billion dollars,” and promises a separate presentation on the plan
“Yeah, I think we probably should do a longer presentation on the game plan for Starlink connectivity and why we think it will ultimately be a business that generates several hundred billion dollars in revenue a year."
"Because you are missing some of the pieces that we know about internally. I can say we do see a clear path to several hundred billion dollars a year in connectivity revenue."
"But I think we should probably describe that in more detail in a forum that’s not limited to a few minutes at an earnings call.”
Dwiki Pratama Putra retweeted
Banger insights from the makers of this chart, who put IBM above Anthropic
“.. At the moment, we have more compute capacity than we have demand, and we have a lot of the tokens and processing power being given away in freemium solutions,” he told @CNBC. “Everyone in the AI race is running for a last-man standing position.”
cnbc.com/2026/10/19/gartner-…
Dwiki Pratama Putra retweeted
Fatimah Azzahra : "Kedepannya tolong jangan diklaim lagi MBG itu tujuan utamanya untuk stunting. Jujur saja sebut untuk ekonomi!" Kritik lugas dari Fatimah Azzahra ini menohok langsung ke jantung narasi pemerintah.
Dwiki Pratama Putra retweeted
Anthropic just made a MASSIVE announcement and I think people are overlooking how significant this could become, especially with the company reportedly looking to IPO at a $2 TRILLION valuation.
Anthropic announced that it plans to commit 5% of its shares to a new Public Benefit Fund, subject to shareholder approval, with the possibility of contributing an additional 2% for every 50% increase in its valuation. Think about how insane that is. At Anthropic’s most recent $965B private valuation, that initial 5% stake would already be worth approximately $48 billion. But at the $2T valuation being discussed for its IPO, we’re talking about $100B in equity potentially being allocated toward public benefit initiatives.
Unlike a traditional charitable donation, Anthropic is committing ownership in the company itself, meaning the value of those shares could continue growing as Anthropic becomes more valuable. The idea is that a portion of the enormous wealth being created by AI could eventually be used to benefit society, including potentially supporting people and communities affected by AI driven economic disruption. We’re entering a period where AI agents are becoming increasingly capable of performing tasks that previously required human labor. Everything from customer service and software development to research, financial analysis, and administrative work is beginning to change. Now, I don’t believe this means everyone is suddenly going to lose their jobs. In fact, I think AI is going to create entirely new industries and opportunities that we haven’t even imagined yet. But we also have to recognize that this transition could be extremely disruptive, particularly for workers whose responsibilities can increasingly be automated.
And this is WHY I think Anthropic is making this move. The company recognizes that AI has the potential to generate trillions of dollars in economic value, but that doesn’t necessarily mean everyone is going to benefit equally. By setting aside equity today, Anthropic is creating a mechanism that could potentially help address those challenges as AI adoption accelerates.
My biggest takeaway from all of this is that we’re reaching a point where AI companies are not only competing to build the most powerful models in the world but are also beginning to think about how the enormous wealth created by those models should be distributed. And I think we’re only scratching the surface of how much AI is going to change our economy over the next decade.
Dwiki Pratama Putra retweeted
Here’s why agent swarms are going to be a very big deal. We can finally make compute a critical variable in a large number of important tasks in the economy which wasn’t possible before.
There’s a large amount of work in the world that is essentially only constrained by the amount of time that we put into research, data processing, discovery work, running tests, and more. All of this work is bottlenecked by our ability to plan, execute, review, and coordinate everything happening in these workflows. Which means we basically do very little of them today relative to their economic value.
Agent swarms are the first real ability to parallelize this work in a way where - in theory - the only parameter that matters is how much compute you throw at the problem. Cyber security, life sciences or material science research, trading, QA testing, data processing, and many other fields are essentially compute constrained problems in the limit once models are good enough at these tasks.
The breakthrough is that in a world where agents can use tools, write code, process data, have strong domain depth, and can coordinate with each other, you can begin to solve arbitrarily complex problems by setting the right goals and applying enough tokens to the problem. This changes a lot of fields as well as even the competitive dynamics in these fields.
Dwiki Pratama Putra retweeted
Honestly, I can’t help but laugh whenever I see this kind of modeling. How many people can actually predict what will happen beyond 2027 with any accuracy? Is there even any point in building these models in the first place? Especially the idea of “forecasting” demand… I think it’s utterly absurd.
This quoted post is unavailable.
Dwiki Pratama Putra retweeted
Now that I have successfully rage baited what feels like every AI bear on X, let's systematically break down the most common counterarguments because they are very off the mark.
#1. The AI trade is the same as the dotcom bubble or the 2008 crisis
This is the most frequent rebuttal, and it is a lazy comparison. When you look under the hood, the numbers do not support it.
Looking at the S&P 500 top 10 at each peak:
March 2000: 53.6x trailing earnings, 9.1% net margin
October 2007: 16.4x trailing earnings, 13.3% net margin
Today: 26.5x trailing, 20.9x forward, 33% net margin
Unlike the dotcom bubble, today's leaders have the earnings to support their build out. In the three years before the 2000 peak, the tech leaders' stocks rose 7.4x on 2.1x earnings growth, and the rest of the market was priced even higher: the Nasdaq traded at about 175x earnings. Today's AI leaders rose 2.9x on 3.1x earnings growth, and on average they pay about a fifth of the 2000 price per unit of growth.
PEG on trailing three year growth:
2000: 1.6 to 5.7
Now: 0.2 to 1.6
2007 was never a valuation problem. It was a leverage problem, which I cover in #4.
Yes, today's concentration at the top of the index is real, and it is higher than in 2000, but this time it is backed by profits. Today the top 10 are 40.6% of the S&P 500's value and 36.1% of its earnings. In March 2000 they were 27% of the value and only about 15% of the earnings.
Put today's earnings at 2000's prices and the gap is massive:
Nvidia at Cisco's peak multiple (178x) would be about $1,210 a share, a $29T company, 5.7 times today's price.
Today's top 10 at the 2000 top 10's 53.6x would be worth $43T even on earnings excluding gains, against $28T today.
The S&P 500 at 2000's forward P/E of 25.2x would be about 10,160, against 7,659.
(See Chart 1)
#2. Nvidia is the next Cisco, or the next Enron.
From January 1998 to the March 2000 peak, Cisco's stock rose 7.6x while its trailing earnings rose 1.7x, ending at 178x pro forma earnings. Over the last 26 months, the same length of time, Nvidia's stock rose 1.8x and its trailing earnings 3.3x. It trades at 31x.
Enron is an even worse comparison. In 2000 it reported $100.8B of revenue, much of it gross trading volume, and kept $979M of it, a 1% margin, while its stock traded at about 74x earnings. Its profits leaned on mark to market accounting and debt kept off its balance sheet, and it filed for bankruptcy in December 2001, then the largest in US history. Over the last 12 months Nvidia turned $303B of revenue into $193B of net income and $134B of operating cash flow, paid in cash by the largest companies in the world.
#3. Today's companies aren't profitable
These are some of the most profitable companies in history. Over the last 12 months Nvidia earned $166B excluding investment gains, more than Saudi Aramco's $161B in 2022, the largest annual profit any company had ever reported. Until 2024 only Aramco had earned more than $100B in a year. Nvidia, Apple, Microsoft and Alphabet now each do, even excluding gains.
The suppliers are the same. Micron earned $85B at a 64% margin and TSMC $78B at 50%. SK hynix made $85B of operating profit, and Samsung made $80B of operating profit in the third quarter alone, a quarter only Aramco has ever topped.
The ten largest tech companies of March 2000 earned $37B combined, at a 15% margin. Cisco kept 17 cents of each revenue dollar and Lucent 9. Nvidia keeps 55 cents and Micron 64.
To be fair, memory is cyclical. These are shortage profits, and Micron's previous record year was $14B in fiscal 2018.
(See Chart 2)
#4. The AI trade is built on debt and is not sustainable
The dotcom buildout ran on borrowed money. Debt owed by US telecom carriers rose from $158B in 1996 to $380B in 2000, and the junk rated carriers had negative operating cash flow as a group. The seven pure network builders (Global Crossing, Level 3, Williams, XO, 360networks, McLeodUSA, Winstar) owed $31.8B plus $6.5B of preferred stock, and six of the seven did not cover their interest bill even on their own EBITDA numbers. Telecom made up 55% of US junk bond defaults in 2001, and 23 telecoms went bankrupt by mid 2002.
In 2000 WorldCom needed 3.2 years of cash flow to repay its debt, Global Crossing 8, and XO's cash flow was negative. Today Alphabet, Microsoft, Amazon and Meta owe $351B, hold $533B of cash, and could repay it all with about six months of cash flow.
In the last year the 10 largest US tech companies generated $1.10T of operating cash flow and had $523B left after $574B of capex. At the 2000 peak, the top 10 tech names had about $34B of free cash flow, $66B in today's dollars, and traded at about 57x cash flow. Today it is 24x.
2008 was leverage inside the financial system. In 2007 financial sector debt was 144% of GDP, broker dealers ran 46x assets to equity, and Citigroup ran 18.6x and Bank of America 11.4x, both top 10 companies at the time. Today financial sector debt is 85% of GDP, broker dealers run 17x, and the S&P 500 top 10 run 1.8x.
The debt that does exist sits at the edges, with names like Oracle. The four largest spenders are paying for this out of cash flow.
(See Chart 3)
#5. Demand is inflated, and they are overbuilding capacity like the dotcom bubble. AI provides no long term value and adoption is overstated.
In 2000, WorldCom said internet traffic doubled every 100 days. In reality it doubled about once a year, and US backbone networks ran at about 10 to 15% of capacity. The carriers built for demand that did not exist yet, in the hope that it would show up.
Yes, AI spend is larger now, but it sits against $2.3T of signed backlog at AWS, Google Cloud, Microsoft and Oracle. Demand is running ahead of supply: TSMC says demand is still far above its ability to supply, Micron says it has no line of sight to when supply and demand balance, and Nvidia's CFO expects supply to remain a bottleneck at least through fiscal 2028.
Rental prices for the H100, a chip that shipped in 2022, are up 17% over the past year after halving from their 2024 peak. Hardware normally gets cheaper as it ages; over the past year we have seen the opposite, because demand is outrunning supply.
The suppliers see this demand first. They would be the canary in the coal mine here, and they just reported record quarters: TSMC, ASML and Micron each posted the highest quarterly revenue in their history. TSMC raised its 2026 growth guidance at every update, to about 44%. ASML lifted its 2026 sales guide from €34 to 39B to €44 to 45B, says 2027 is close to fully covered with orders for its standard EUV machines, and has customers asking to order for 2029 and 2030. By press estimates a High NA machine costs about $380M and a standard EUV machine about $180M. Nobody books those years ahead on a whim.
Memory makers are seeing the same and are signing take or pay agreements years out. Micron has 26 of them, some running to 2031, covering over 35% of its revenue through 2030, and over 75% of its 2027 output is already committed. Samsung's agreements run at least five years, and SanDisk has $93.9B of minimum contracted revenue.
For comparison, in 2000 Lucent had committed $8.1B of financing to its own customers. Now the money runs the other way: Micron holds $12.7B of customer cash deposits, and ASML's long term customer down payments rose from €3.6B to €9.3B last quarter. These commitments are funded by the cash flow of the largest companies in the world.
Generative AI reached 55% of US adults aged 18 to 64 within three years of launch. The internet was at 30% at the same point, and the PC at 20%. About 2B people use AI worldwide, 55.9% of US businesses on Ramp pay for AI tools, and Census data shows 23.8% of US businesses used AI in the past two weeks. Consumers spend about $40B a year on it, roughly $20 per user, so monetization has barely started.
Physical AI is earlier still. Waymo runs about 500,000 paid rides a week, up from 50,000 in May 2024, and Nvidia's physical AI revenue passed $9B over the last 12 months. Each robot or autonomous vehicle needs far more compute and memory than a phone: Micron says an L4 vehicle typically carries over 200 GB of DRAM, about 24 times a phone, and expects humanoids to need as much. Counterpoint expects humanoid memory demand to grow 20x from 2026 to 2030. Volumes are still small; this layer of demand has barely started.
(See Chart 4)
#6. Today's tech leaders aren't smart
This is the most insane one for me. The idea that random people on X can question the intelligence of the people running these companies is nuts. ASML's EUV machines fire a laser at a drop of molten tin twice, 50,000 times a second, creating a plasma hotter than the surface of the sun to print chips. TSMC is mass producing chips at 2nm. These companies turn science fiction into reality on a regular basis.
The leadership is also nothing like 2000. Executives at Enron and WorldCom were convicted of fraud, and Congress passed Sarbanes Oxley in 2002 because of them. Since then every public company CEO and CFO has had to personally certify their company's financial statements. For people to throw stones at the intelligence of these leaders is just loony.
In summary, I am very optimistic about the transformative nature of AI for our lives and our economy. So as I originally stated, I will continue to be long my picks and shovels favorites like TSMC, ASML, GEV and others. We are still in the early innings; don't let lazy fear mongering tell you otherwise. I am not saying "it is not the same" out of hope or cope. The math and data support me.
Dwiki Pratama Putra retweeted
jack’s pretty spot on here, you don’t need to be an ai researcher to see it either:
- 3 years ago ai models couldn’t do simple math. couldn’t count the letters in “strawberry”
- today the most powerful models have solved coding and are solving the most challenging problems in maths, physics and genetics. soon (and i mean very soon) it’ll expand for medicine, cancer, knowledge work etc
- model’s are improving at a blistering pace, new releases every 13 days. this eventually snowballs into rsi.
- all while the cost of intelligence is plummeting. the frontier models of last quarter is now being served to you at 1/8th the cost.
and the craziest part is barely anyone is even using AI. there’s a whole wide world of jobs, tasks, problems that can (and will) be eaten up by the singularity
we just don’t notice it because it hasn’t affected “us individually” but doesn’t mean it’s not coming.
anyways, think people will be very surprised next year when what ai did to software engineering comes barrelling into modern day humanity.
I’m speaking at the IMF in an hour about the singularity and why I believe it has begun and the choices we'll need to make. Watch live here: youtube.com/live/Rqt9kWoTJbc…
Dwiki Pratama Putra retweeted
There’s a fundamental reason why TSMC has always been notoriously conservative in its forecasts and investment planning
It’s not because they are the only adults in the room. It’s because they are the ones writing the checks to build those incredibly expensive fabs, packed with ASML lithography equipment
If AI demand turns out to be overestimated, Nvidia would certainly take a significant valuation hit. And with $279bn in disclosed supply and capacity commitments, its potential financial exposure is hardly trivial either
But TSMC faces a different problem entirely: tens of billions sunk into fabs that could sit underutilized for years, with depreciation and fixed costs crushing returns on capital
You can adjust a product roadmap, renegotiate some supply commitments or write down inventory. You can’t simply return a $20bn semiconductor fab because demand didn’t materialize
That’s why TSMC’s forecasts deserve particular attention
These guys have enormous visibility across the semiconductor supply chain, strong incentives to avoid overbuilding and decades of experience navigating the industry’s brutal cycles
And yet, they are planning to spend $60–64bn this year alone, with even heavier aggregate investment over the next three years, because AI demand keeps exceeding their expectations
For all these reasons, as far as AI demand visibility goes, $TSMC inputs carry much more informational content than W St vibes…
AI Infra: Are we confusing a U.S. power shortage with a global supply glut?
One of the biggest debates around the AI buildout is whether we have enough power and data center capacity to absorb all the compute coming online in 2027.
I think we're missing an important distinction: U.S. supply constraints ≠ global supply constraints.
Global demand: ~50-55GW in 2027 based on most estimates.
[chart below]
Global supply: ~50GW. Dylan Patel / @SemiAnalysis_ forecasts roughly 50GW of new capacity coming online, including 10-20GW attributable to @OpenAI and @Anthropic.
[chart below]
U.S. supply: ~20-25GW. @FundaAI estimates ~17-23GW. If the U.S. accounts for ~70% of global demand (~35GW), we're looking at a potential 10-15GW power shortfall. That gap is real, and I don't dismiss it.
[chart below]
But here's what I think the market is overlooking: the rest of the world.
Based on channel checks, Southeast Asia could potentially bring online ~15GW, Japan ~5GW, Europe another ~5GW, with additional capacity across China, Canada and increasingly Latin America.
These are indicative estimates, not guaranteed deliveries. But the point is that global capacity to absorb AI compute may be considerably larger than a U.S.-centric analysis suggests. Of course, power availability doesn't automatically translate into operational data centers. Permitting, grid connections, equipment deliveries and construction timelines all matter. And compute demand isn't perfectly fungible across geographies.
Still, a U.S. power shortage doesn't necessarily mean the world cannot absorb 50GW of compute.
Another important data point: $TSMC.
Of all the companies in the semiconductor supply chain, $TSMC is probably the one whose judgment I trust most. It is famously conservative and has exceptional visibility into customer demand.
On its latest earnings call, CEO C.C. Wei said TSMC has gone as far as sending its own teams to verify data center construction progress and power availability in the U.S.
"The feedback is so far so good, and the CSPs have been well prepared."
My take: I'm cautiously optimistic that sufficient capacity exists globally to absorb the incoming compute supply. The RoW opportunity is larger than many appreciate.
My biggest concern remains execution: whether enough land, power and data center shells actually get delivered on time. Delays could leave chips temporarily idle and cause near-term disappointments for certain suppliers.
But that's a very different bearish thesis from structural global oversupply.
++
Better formatted: robonomics.substack.com/p/ai…
Dwiki Pratama Putra retweeted
deddy sitorus anggota komisi ii dpr fraksi pdip bilang kopdes merah putih berpotensi gagal lebih dari 90 persen
ia bilang program ini bisa jadi beban negara bertahun-tahun
ia khawatir nanti banyak orang berurusan dengan hukum karena pelaksanaannya “sembrono”
alasannya
konsepnya salah kaprah dari awal
koperasi tidak lahir dari keinginan warga desa, tapi dipaksa dari atas (top-down)
dibangun pakai template yang sama di seluruh indonesia, tanpa model bisnis yang cocok dengan kebutuhan tiap desa
modalnya, kata dia, diambil dengan cara “merampok” dana desa yang seharusnya untuk infrastruktur desa
proyek juga menyedot apbn dari kementerian dan bank milik negara
lahannya, kata dia, “merampas” tanah milik desa, dibangun tanpa lihat kelayakan lokasi dan desain yang cocok dengan calon pembeli
Dwiki Pratama Putra retweeted
David Vélez on why banks can't compete with Nubank:
> 140M customers, 10,000 employees
> Incumbents: 100M customers, 100,000 employees, and 6,000 branches
> ~3% of the cost to serve
Next step: AI-run banks with 500M customers & 5,000 employees.
The rest is history.
Dwiki Pratama Putra retweeted
The HBM market is setting up for one of the BIGGEST supply demand imbalances in the entire AI buildout.
Just look at what Bloomberg Intelligence is projecting. HBM shipments could grow from roughly 2.5 billion GB in 2025 to nearly 18 billion GB by 2030. That's approximately a 7X increase in volume in just FIVE years. Normally, when an industry dramatically increases production, you would expect prices to decline as supply catches up with demand. Yet Bloomberg's projections are showing the complete opposite. Average HBM prices are expected to climb from roughly $13 per GB in 2025 to $35 per GB by 2030. Think about how insane that is. We're potentially looking at a market where shipments increase roughly 7X while average selling prices nearly triple.
And this is exactly why I'm so bullish on memory. Every new generation of AI accelerators requires increasingly sophisticated memory to deliver the bandwidth needed to keep those GPUs running efficiently. And as models become larger, inference workloads explode, and autonomous agents and physical AI begin scaling, the demand for memory is only going to accelerate. What makes HBM particularly interesting is that supply cannot simply expand overnight. Manufacturing HBM requires advanced packaging, complex stacking processes, enormous capital investment, and significant production capacity. Not to mention, expanding HBM production can consume manufacturing capacity that would otherwise go toward conventional DRAM, potentially tightening supply across the broader memory market as well. This is why I continue to believe memory is one of the most attractive opportunities in the AI infrastructure trade.
South Korea's stock market is in FREE FALL and the panic surrounding the memory trade is getting completely out of control.
The KOSPI just plunged another 7%, bringing its decline to roughly 40% from its June highs. Samsung and SK Hynix are getting absolutely destroyed, and the selling pressure is spreading across the entire AI infrastructure trade. I went on Making Money with @cvpayne today to talk about the memory selloff and why I believe investors are losing sight of what's actually happening in this industry.
And now we're getting ANOTHER warning out of South Korea. A new column published by Maeil Business Newspaper is raising concerns that the global memory industry could face oversupply beginning in 2029 as new manufacturing capacity starts coming online. The argument is that South Korea has become increasingly dependent on Samsung and SK Hynix, and that the country has roughly two to three years to develop other growth industries before the memory cycle potentially reverses. Now, I think this is something investors should absolutely pay attention to but here's where I disagree with the panic.
We're talking about a potential oversupply problem in 2029 while the industry is still struggling to meet demand TODAY. Memory manufacturers are already dealing with capacity constraints, AI customers are securing long term supply agreements, and demand for HBM continues to grow as Nvidia $NVDA and other AI chipmakers push toward increasingly memory intensive systems. And people need to understand that building additional manufacturing capacity doesn't automatically mean we're going to have an oversupply problem. These fabs take years to construct, equipment needs to be installed, production yields need to improve, and manufacturers still have to qualify their products with customers. Not to mention, HBM manufacturing is substantially more complicated than traditional DRAM, and not all additional memory capacity can immediately be used to produce the most advanced AI memory products.
Now, could we eventually see oversupply in 2029? ABSOLUTELY. Memory is a cyclical industry and I've never argued that these companies will enjoy record pricing and margins forever. But there's a massive difference between acknowledging that possibility and acting like the memory cycle is already over. I still believe the fundamentals of the memory trade remain incredibly strong, especially for companies positioned to benefit from HBM and the broader AI infrastructure buildout. Bullish on memory and if you want to see how I’m investing through this selloff and check out my full portfolio below. link.milkroad.com/tx5qip
Dwiki Pratama Putra retweeted
$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.
Dwiki Pratama Putra retweeted
Anthropic co-founder Jack Clark tells the IMF the world is about to go from 10 million scientists to 10 billion (almost all virtual):
“According to UNESCO estimates, there are approximately 10 million scientists in the world. And even with only 10 million scientists, we have too few humans to oversee the work of today, too few laboratories to do material science analysis, too few biological labs to develop and test new drugs, and too few large pieces of scientific equipment in space or on the Earth for the experiments we want to run.
"Tell me what happens when instead of 10 million scientists, we have a billion virtual scientists directed by our own 10 million. Or 10 billion [virtual scientists]. That's the world predictably coming right for us. And we have to get ready. If we do get ready, we are going to be in a better position to translate the intelligence of AI systems into gains for us all. You know, longer, healthier lives, more energy, more food, more wealth.”
Dwiki Pratama Putra retweeted
The biggest misread of anyone predicting the end of software engineering was imagining that the world only needed a certain amount of software, and bringing automation to the field would reduce the need for labor.
Exactly the opposite is happening where essentially most things are actually turning into software. Because agents can use tools, work with computers, write software on the fly, and process data like humans (but far faster), we can now throw “software” at problems that never could reasonably be digitized before.
What that leads to is the role of engineering taking on greater importance in the economy, because we can now engineer far more systems and processes than were possible before.
If you’re a bank trying to do new forms of trading, engineers are often your first bet. If you are in life sciences, you expect more and more research and lab work to be engineering centric. If you’re in manufacturing, you’re trying to automate more of the supply chain, all that is engineering. More and more work is shaped like software problems.
Yes we’re not going to be coding manually in the future. But almost everything will be code, which means those that understand it are in a fantastic position.
Some observations about the "total" automation of software engineering:
1. There are more software engineers now than in 2025, and demand is growing.
2. Software engineers have never worked harder than they do now.
3. They largely don't write code anymore, agents write ~100% of the code.
4. The bottom line of the companies that employs these engineers and agents (i.e. the terminal purpose of all this code) is typically not budging as a result of AI adoption.
This is why you have to think in terms of systems. Automating one aspect of a system creates its own friction and moves the bottlenecks elsewhere.
Dwiki Pratama Putra retweeted
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.
Dwiki Pratama Putra retweeted
It is only a slight exaggeration and twist to say that my primary AI bull thesis is: Most* AI bears whose comments I see - among smart professional investors - have garbage, theses-destroying levels of factual knowledge and make garbage logical inferences based on the true facts they do know. They are smart enough to make up a stupid theory of how each new fact is bearish.
Whereas some of the AI bulls put out content like this. There are too many important and highly-backed factual tidbits and insights in here to count.
It's a good general discussion of AI and AI infrastructure, wrapped around - bonus! - a deep discussion of $CRWV, which is ground zero of the key bear focus.
Thank you, @goodalexander!
*P.S. My mirror-image meta-worry is that some genuinely knowledgable and growth-tech-fluent AI bears have a great thesis but are keeping quiet. I keep looking and not finding them. The best I can find is a few "non-garbage" ones - credible thesis, makes logical sense, moderately well supported factually, but probably wrong given even my level of knowledge. (People like goodalexander are far deeper. I'm just a wizard with a PA spread across 6 or 7 other investing topics).
