Vink柒
810 Posts
Vink柒
推特：@Vinkyu567
不定时分享交易心得
哈基米 Holder
Frequent Trader
1.7 Years
23 Following
1.3K+ Followers
1.4K+ Liked
Posts
All
Quotes
Videos
Live
PINNED
I recommend a charting app. It’s the best one I’ve used so far—it's a bit better than TV. OpenMarket—it's good for all kinds of styles.
For ETH, I still think we should watch the pullback. It hasn’t reached a point where the upward trend is starting yet. Keep waiting for a bottom-fishing opportunity.
$ETH
For ETH, I still think we should watch the pullback. It hasn’t reached a point where the upward trend is starting yet. Keep waiting for a bottom-fishing opportunity.
$ETH
Some Xin’s update on the prospectus is coming out tonight
Second-quarter net profit explodes
All the related sectors are surging
Let’s see if we can get an allotment
Second-quarter net profit explodes
All the related sectors are surging
Let’s see if we can get an allotment
Verified
MicroStrategy started selling coins again. Sure enough, once it happened the first time, it happened countless more times—though the CTO has already given a lot to other institutions.
Verified
I reviewed the market again on July 1:
Meta is up, while chip stocks are down.
At first glance it looks like two separate directions, but the market is actually pricing the same thing:
With so much money burned by this AI run, how will it pay back?
When many people see that Meta might sell AI compute, their first reaction is: Did they buy too much?
But I don’t think it’s that simple.
Meta is both increasing capital expenditures and pushing the commercialization of its compute. This feels more like “funding the new with the old”: turning some older compute into cash flow, then continuing to invest in the next-generation compute that is even more scarce.
So the issue isn’t that AI compute isn’t wanted—it’s that compute across different generations is starting to be repriced.
Meta is up because the market sees the possibility that AI infrastructure could shift from a cost center into a revenue-producing asset.
Chip stocks are down because the market is starting to worry about the supply-demand inflection point for the previous generation of chips, as well as the valuation that has been stretched after share prices ran up earlier.
A stock drop doesn’t mean demand has collapsed.
More accurately, July 1 wasn’t the end of the AI narrative—it was an upgrade of it.
In the past, the market was looking at:
Who bought the most GPUs.
Now the market is starting to look at:
Who can turn GPUs into cash flow.
AI infrastructure is moving from a “capital expenditure race” into a “compute asset operations race.”
This is the real part of the market change that deserves attention.
$META.US
Meta is up, while chip stocks are down.
At first glance it looks like two separate directions, but the market is actually pricing the same thing:
With so much money burned by this AI run, how will it pay back?
When many people see that Meta might sell AI compute, their first reaction is: Did they buy too much?
But I don’t think it’s that simple.
Meta is both increasing capital expenditures and pushing the commercialization of its compute. This feels more like “funding the new with the old”: turning some older compute into cash flow, then continuing to invest in the next-generation compute that is even more scarce.
So the issue isn’t that AI compute isn’t wanted—it’s that compute across different generations is starting to be repriced.
Meta is up because the market sees the possibility that AI infrastructure could shift from a cost center into a revenue-producing asset.
Chip stocks are down because the market is starting to worry about the supply-demand inflection point for the previous generation of chips, as well as the valuation that has been stretched after share prices ran up earlier.
A stock drop doesn’t mean demand has collapsed.
More accurately, July 1 wasn’t the end of the AI narrative—it was an upgrade of it.
In the past, the market was looking at:
Who bought the most GPUs.
Now the market is starting to look at:
Who can turn GPUs into cash flow.
AI infrastructure is moving from a “capital expenditure race” into a “compute asset operations race.”
This is the real part of the market change that deserves attention.
$META.US
After a big batch of liquidity absorption below, we still saw an uptrend. Now orders are relatively steady and we’ve also reached a resistance level. At the beginning of the third quarter, there should first be a rebound, and then it may drop again. The overall trend in the fourth quarter will move upward.
It all depends on how exactly Wosh reduces its balance sheet—how will it cut it?
And as I said before, no matter how much it falls, it won’t fall much. At worst, you’d lose at most 50%, and upside would be 200%—a 1:4 risk-reward ratio.
$BTC
It all depends on how exactly Wosh reduces its balance sheet—how will it cut it?
And as I said before, no matter how much it falls, it won’t fall much. At worst, you’d lose at most 50%, and upside would be 200%—a 1:4 risk-reward ratio.
$BTC
Vink柒
·
--
BullishStop, Azu
Below is all bargain hunting, there’s no more room to stack it$BTC
Below is all bargain hunting, there’s no more room to stack it$BTC
The current main storyline
Layer 1: The super cloud giants pay NVIDIA.
Microsoft, Meta, Amazon, and Google, in order to build AI data centers, have massively purchased GPUs, complete systems, networking, and AI infrastructure.
As a result, NVIDIA first captures the largest incremental revenue and profit leverage—this is also the core pricing logic behind the AI bull market over the past two years.
Layer 2: NVIDIA’s growth turns storage into the new bottleneck.
The more GPUs are sold and the more AI servers are built, the greater the demand for HBM, high-end DRAM, NAND, and advanced packaging.
At this point, the profit pool begins to spread from “compute chips” to “storage supply.”
Micron, SK hynix, and Samsung aren’t rising just because the cycle is simply recovering—they’re rising because they have become indispensable links in the AI factories.
Layer 3: After storage companies start making money, they must keep expanding capacity.
To meet AI customers’ long-term orders, Micron, SK hynix, and Samsung have to increase capital expenditures—build plants, buy equipment, do advanced packaging, and improve yields. So new orders start flowing to
$AMAT.US
$LRCX.US
$KLAC.US
ASML and other equipment companies, as well as upstream materials, components, and facilities engineering firms.
The migration of profits doesn’t happen overnight—it passes along the industry chain, layer by layer.
In the first stage, the super cloud providers’ capital expenditures flow to NVIDIA, and GPUs become the first valuation anchor of the AI bull market.
In the second stage, as NVIDIA delivers more AI systems, HBM and high-end storage are pushed into the new bottleneck position, and Micron, SK hynix, and Samsung begin to capture the profit leverage.
In the third stage, storage companies, to lock in future orders, are forced to increase capital expenditures, and equipment and materials companies then stand at the entry point of the next round of order flow.
In other words,
first, capital buys the most visible leading players;
then, it buys the bottlenecks tightened by the leaders;
finally, it buys the shovels behind the bottleneck’s capacity expansion.
Layer 1: The super cloud giants pay NVIDIA.
Microsoft, Meta, Amazon, and Google, in order to build AI data centers, have massively purchased GPUs, complete systems, networking, and AI infrastructure.
As a result, NVIDIA first captures the largest incremental revenue and profit leverage—this is also the core pricing logic behind the AI bull market over the past two years.
Layer 2: NVIDIA’s growth turns storage into the new bottleneck.
The more GPUs are sold and the more AI servers are built, the greater the demand for HBM, high-end DRAM, NAND, and advanced packaging.
At this point, the profit pool begins to spread from “compute chips” to “storage supply.”
Micron, SK hynix, and Samsung aren’t rising just because the cycle is simply recovering—they’re rising because they have become indispensable links in the AI factories.
Layer 3: After storage companies start making money, they must keep expanding capacity.
To meet AI customers’ long-term orders, Micron, SK hynix, and Samsung have to increase capital expenditures—build plants, buy equipment, do advanced packaging, and improve yields. So new orders start flowing to
$AMAT.US
$LRCX.US
$KLAC.US
ASML and other equipment companies, as well as upstream materials, components, and facilities engineering firms.
The migration of profits doesn’t happen overnight—it passes along the industry chain, layer by layer.
In the first stage, the super cloud providers’ capital expenditures flow to NVIDIA, and GPUs become the first valuation anchor of the AI bull market.
In the second stage, as NVIDIA delivers more AI systems, HBM and high-end storage are pushed into the new bottleneck position, and Micron, SK hynix, and Samsung begin to capture the profit leverage.
In the third stage, storage companies, to lock in future orders, are forced to increase capital expenditures, and equipment and materials companies then stand at the entry point of the next round of order flow.
In other words,
first, capital buys the most visible leading players;
then, it buys the bottlenecks tightened by the leaders;
finally, it buys the shovels behind the bottleneck’s capacity expansion.
Codex Orange Skin Book also got a 2k star badge
Truly teaches you codex from scratch—so you can directly build your own stock-monitoring software. Find the data you need. If you want what you need, you can check my Twitter.
Truly teaches you codex from scratch—so you can directly build your own stock-monitoring software. Find the data you need. If you want what you need, you can check my Twitter.
Look at the liquidation map and see at what price levels large liquidations occur. Then determine which price levels are most likely to become a short-term bottom or top. After the liquidations, price will consolidate for a while.
Below are the 40,600 and 17,000 levels where massive liquidations occur—today just triggered one large liquidation.
Above, the nearest is around 64,000 (from the nearby level), where there are 1,800 BTC to be liquidated.
Below are the 40,600 and 17,000 levels where massive liquidations occur—today just triggered one large liquidation.
Above, the nearest is around 64,000 (from the nearby level), where there are 1,800 BTC to be liquidated.
BeanBag Launches Pro Version
Standard Package Monthly 68 RMB, Annual Subscription 688 RMB
Enhanced Package Monthly 200 RMB, Annual Subscription 2048 RMB
Premium Package Monthly 500 RMB, Annual Subscription 5088 RMB
Can you HODL?
Standard Package Monthly 68 RMB, Annual Subscription 688 RMB
Enhanced Package Monthly 200 RMB, Annual Subscription 2048 RMB
Premium Package Monthly 500 RMB, Annual Subscription 5088 RMB
Can you HODL?
What does it mean to chase a buy order? Are we that eager?
But I don't want it to pump so quickly; a dip would be healthier. I haven't accumulated enough in my DCA yet.
$BTC
But I don't want it to pump so quickly; a dip would be healthier. I haven't accumulated enough in my DCA yet.
$BTC
There are some big orders propping things up, but we've hit a holiday period, so let's see if they pull those orders next week.
If they don't pull out, we could take a small long position for a rebound at these levels. If the price tanks, no worries, just treat it as dollar-cost averaging. Right now, the max downside is about 20%, but the upside potential is 100%, though it might take a while to play out.
As I mentioned before, once we fill those liquidity gaps on the way up, a drop is definitely coming. Bears take their time; they don’t just shoot up out of nowhere.
The short position I opened on ETH has also been resolved; I'm currently in profit but still planning to hold on a bit longer.
$BTC
If they don't pull out, we could take a small long position for a rebound at these levels. If the price tanks, no worries, just treat it as dollar-cost averaging. Right now, the max downside is about 20%, but the upside potential is 100%, though it might take a while to play out.
As I mentioned before, once we fill those liquidity gaps on the way up, a drop is definitely coming. Bears take their time; they don’t just shoot up out of nowhere.
The short position I opened on ETH has also been resolved; I'm currently in profit but still planning to hold on a bit longer.
$BTC
mrvl to the moon $MRVL
Vink柒
·
--
Just added more to my MRVL stack, I gotta shout this out again - I'm bullish on it.
Forget all those complex business models, let’s break it down with the simplest example.
Think of AI as a never-ending Double 11 sale. Everyone's eyes are glued to those factories (GPUs, computing power) hustling to crank out products, watching who can ramp up production and whose products shine the brightest. The glory of production is visible to all.
But if you think about it, what really chokes the Double 11 every year? It's not that the factories can’t produce, it’s that they can’t ship it out.
Warehouses overflowing, shipping services overwhelmed, your package stuck at some sorting center for three days - when that happens, no matter how much product is made, it just sits in the warehouse and goes stale.
Marvell is like the “logistics master” behind Double 11, fixing sorting centers, laying down major routes, and boosting transport capacity.
What they do, in a nutshell, is get each piece of “cargo” (AI inference data) flying swiftly, stably, and efficiently between factories (chips).
And this AI Double 11 is hitting that critical point - the bottleneck is shifting from “producing goods” (computing + storage) to “delivering goods” (connections).
As models get larger and more complex, a single Q&A session needs to shuttle massive data back and forth among thousands of chips; at this point, what determines the speed of the whole system is no longer just the power of a single chip, but whether this “sorting network” can handle the load.
What’s great about Marvell is: they don’t just fix the network. They’ve taken on the entire logistics chain - major city transport (optical interconnects), central sorting hubs (switching chips), and various endpoint routing, plus they customize fulfillment centers for some of the biggest e-commerce clients (custom chips) - they cover it all in one go.
Buying below $MRVL 300 is a no-brainer, the lower it drops, the more I buy. A few months down the line, you’ll see how cheap 300 is, just like Bitcoin dropping to 20k.
Forget all those complex business models, let’s break it down with the simplest example.
Think of AI as a never-ending Double 11 sale. Everyone's eyes are glued to those factories (GPUs, computing power) hustling to crank out products, watching who can ramp up production and whose products shine the brightest. The glory of production is visible to all.
But if you think about it, what really chokes the Double 11 every year? It's not that the factories can’t produce, it’s that they can’t ship it out.
Warehouses overflowing, shipping services overwhelmed, your package stuck at some sorting center for three days - when that happens, no matter how much product is made, it just sits in the warehouse and goes stale.
Marvell is like the “logistics master” behind Double 11, fixing sorting centers, laying down major routes, and boosting transport capacity.
What they do, in a nutshell, is get each piece of “cargo” (AI inference data) flying swiftly, stably, and efficiently between factories (chips).
And this AI Double 11 is hitting that critical point - the bottleneck is shifting from “producing goods” (computing + storage) to “delivering goods” (connections).
As models get larger and more complex, a single Q&A session needs to shuttle massive data back and forth among thousands of chips; at this point, what determines the speed of the whole system is no longer just the power of a single chip, but whether this “sorting network” can handle the load.
What’s great about Marvell is: they don’t just fix the network. They’ve taken on the entire logistics chain - major city transport (optical interconnects), central sorting hubs (switching chips), and various endpoint routing, plus they customize fulfillment centers for some of the biggest e-commerce clients (custom chips) - they cover it all in one go.
Buying below $MRVL 300 is a no-brainer, the lower it drops, the more I buy. A few months down the line, you’ll see how cheap 300 is, just like Bitcoin dropping to 20k.
Verified
Just added more to my MRVL stack, I gotta shout this out again - I'm bullish on it.
Forget all those complex business models, let’s break it down with the simplest example.
Think of AI as a never-ending Double 11 sale. Everyone's eyes are glued to those factories (GPUs, computing power) hustling to crank out products, watching who can ramp up production and whose products shine the brightest. The glory of production is visible to all.
But if you think about it, what really chokes the Double 11 every year? It's not that the factories can’t produce, it’s that they can’t ship it out.
Warehouses overflowing, shipping services overwhelmed, your package stuck at some sorting center for three days - when that happens, no matter how much product is made, it just sits in the warehouse and goes stale.
Marvell is like the “logistics master” behind Double 11, fixing sorting centers, laying down major routes, and boosting transport capacity.
What they do, in a nutshell, is get each piece of “cargo” (AI inference data) flying swiftly, stably, and efficiently between factories (chips).
And this AI Double 11 is hitting that critical point - the bottleneck is shifting from “producing goods” (computing + storage) to “delivering goods” (connections).
As models get larger and more complex, a single Q&A session needs to shuttle massive data back and forth among thousands of chips; at this point, what determines the speed of the whole system is no longer just the power of a single chip, but whether this “sorting network” can handle the load.
What’s great about Marvell is: they don’t just fix the network. They’ve taken on the entire logistics chain - major city transport (optical interconnects), central sorting hubs (switching chips), and various endpoint routing, plus they customize fulfillment centers for some of the biggest e-commerce clients (custom chips) - they cover it all in one go.
Buying below $MRVL 300 is a no-brainer, the lower it drops, the more I buy. A few months down the line, you’ll see how cheap 300 is, just like Bitcoin dropping to 20k.
Forget all those complex business models, let’s break it down with the simplest example.
Think of AI as a never-ending Double 11 sale. Everyone's eyes are glued to those factories (GPUs, computing power) hustling to crank out products, watching who can ramp up production and whose products shine the brightest. The glory of production is visible to all.
But if you think about it, what really chokes the Double 11 every year? It's not that the factories can’t produce, it’s that they can’t ship it out.
Warehouses overflowing, shipping services overwhelmed, your package stuck at some sorting center for three days - when that happens, no matter how much product is made, it just sits in the warehouse and goes stale.
Marvell is like the “logistics master” behind Double 11, fixing sorting centers, laying down major routes, and boosting transport capacity.
What they do, in a nutshell, is get each piece of “cargo” (AI inference data) flying swiftly, stably, and efficiently between factories (chips).
And this AI Double 11 is hitting that critical point - the bottleneck is shifting from “producing goods” (computing + storage) to “delivering goods” (connections).
As models get larger and more complex, a single Q&A session needs to shuttle massive data back and forth among thousands of chips; at this point, what determines the speed of the whole system is no longer just the power of a single chip, but whether this “sorting network” can handle the load.
What’s great about Marvell is: they don’t just fix the network. They’ve taken on the entire logistics chain - major city transport (optical interconnects), central sorting hubs (switching chips), and various endpoint routing, plus they customize fulfillment centers for some of the biggest e-commerce clients (custom chips) - they cover it all in one go.
Buying below $MRVL 300 is a no-brainer, the lower it drops, the more I buy. A few months down the line, you’ll see how cheap 300 is, just like Bitcoin dropping to 20k.
Vink柒
·
--
BullishMicron hit its last peak around June '24 at about 160. Due to the tariff war, this key event caused it to drop to 60, and now it's over 1000;
MRVL's last peak was in January '25 at around 125. The tariff war also impacted it, causing a dip to 50, and now it's peaked at 300.
Micron has surged about 17x from its lowest point, and has risen 6-7x from its last peak;
Marvell has increased about 6-7x from its lowest point, and has less than 3x growth from its last peak.
So is Marvell overpriced? Whether to grab this price mainly depends on whether you believe it has the potential to be the next Micron.
The top market cap player, Huang, personally recommends the next trillion-dollar company.
You can only go in with low leverage or spot trading; high-leverage contracts definitely won't hold for you.
$MRVL
MRVL's last peak was in January '25 at around 125. The tariff war also impacted it, causing a dip to 50, and now it's peaked at 300.
Micron has surged about 17x from its lowest point, and has risen 6-7x from its last peak;
Marvell has increased about 6-7x from its lowest point, and has less than 3x growth from its last peak.
So is Marvell overpriced? Whether to grab this price mainly depends on whether you believe it has the potential to be the next Micron.
The top market cap player, Huang, personally recommends the next trillion-dollar company.
You can only go in with low leverage or spot trading; high-leverage contracts definitely won't hold for you.
$MRVL
Log in to explore more content