是阿杰吖

Recently, I’ve been increasingly feeling that what will be truly scarce in AI’s next phase might not be computing power, but rather trustworthy human data.
The internet has never lacked data. What’s missing is whether you can prove:
that these data come from real people—not bots, mass accounts, or content generated by AI itself.
This problem is getting more and more severe. As models get stronger and AI-generated content grows, the internet is more likely to fall into a loop:
AI reads human data → generates more content → new models train on these AI-generated contents.
In the end, the total data volume keeps growing, but the effective signals that actually originate from the human world get diluted.
KGeN isn’t a simple data annotation platform. It’s building a layer called the Verified Human Network.
Today its network covers 60+ countries and 45+ languages. K-Quest can collect multimodal data such as voice, images, video, and data from real environments. Behind it, a Reputation System continuously verifies participants across dimensions like Proof of Human, Proof of Skill, and Proof of Engagement.
These two layers working together are crucial.
In the future, what AI companies buy may be not just one million data points, but:
one million data points from which people, which regions, which languages, and which real environments—and data with verifiable source information.
Especially as AI moves from chat boxes into robotics, autonomous driving, Voice AI, and World Models, data requirements will shift from text to sound, actions, space, and feedback from the real world.
KGeN is already expanding toward Physical AI. The multimodal data it emphasizes—Sound, Sight, Motion, and more—is, in essence, betting on the next phase’s data entry point.
And I think this is also where $KGEN is worth observing most.
If Verified Human Data ultimately becomes foundational resource in the AI industry chain, then the market KGeN is targeting won’t be limited to Web3 or just user growth for games. It may keep extending into LLM training, model evaluation, Voice AI, robotics, World Models, and even a broader human-machine collaboration data market. Its KAI network already covers human experts in specialized fields like Coding, Healthcare, Financial Services, and Legal.
In the past, the AI era competed over who had more data.
Next, it may be about:
who can prove these data truly come from humans.
The easier data is to generate, the more real it becomes—meaning real data becomes more expensive.
That might be the long-term narrative that KGeN is most worth paying attention to.

When I first saw #宇宙之心 , I thought the name was quite imaginative.
After learning more, I found that what it’s trying to convey is the future of digital civilization and decentralized co-building.
This direction is indeed quite new.
#宇宙之心 $SPCX

In the crypto market, volatility itself moves quickly. For short-term traders, the hardest part is often not that there’s no opportunity—but that the opportunity moves too fast.

You spot a chance, and the price has already surged. By the time you confirm the trend, it’s easy to end up chasing higher.

Watching the charts for a few hours a day, what gets depleted in the end is often more than just capital—it’s your attention and emotions.

So I’ve always felt that the real value of AI trading isn’t helping you predict every rise and fall. It’s about handing repetitive work to the system, so your trading logic stays stable.

@0x_aix
's intraday high-frequency mode is designed specifically for this short-term scenario. By continuously analyzing market signals, it looks for opportunities on 3-minute and 15-minute cycles that are better suited for intraday trading.

It doesn’t intentionally chase long-term trend movements; instead, it focuses on effective intraday volatility. Typically, positions last for a few hours. It adjusts its pace according to market changes, and each day it filters for about 5–10 trading opportunities.

Based on real trading experience, the biggest challenge with short-term trading is often not that the strategy isn’t varied enough—but that execution is hard to keep consistent over the long run.

When the market is rising, it’s easy to chase in. When it’s falling, it’s easy to急着 cut losses in panic. After several consecutive mistakes, you start doubting your judgment. A lot of the time, the trading outcome isn’t losing because your analysis is wrong—it’s losing because your execution slips.

That’s exactly where AI shines.

It can continuously monitor the market, filter signals and execute according to predetermined logic, and reduce the impact of human emotions on trading. Of course, high-frequency trading doesn’t mean mindless entries. Around major data releases, abnormal volatility still requires you to proactively reduce frequency and control risk.

One point I personally value a lot is that the way we trade may change in the future. In the past, it was about how long you watched the screen and how fast you reacted. Now, it will be more about who can use tools more efficiently.

For short-term players, AI won’t replace trading judgment—but it will become an efficiency-enhancing support system. Freeing up your time from repeatedly staring at charts, and focusing your energy on strategy optimization and risk management—this may be the most practical value of AI trading.

Personal experience only; not investment advice. DYOR.

As AI market conditions have reached this point, attention in the market is gradually shifting from “who manufactures compute power” to “who supports compute power.”

Recently, a change in the storage industry chain has emerged that is worth paying attention to.

In the latest rebalancing of the Roundhill Memory ETF (DRAM), which focuses on storage chips, China’s DRAM company ChangXin Memory Technologies (CXMT) has entered the top ten holdings, with a stake of 2.52%. Meanwhile, in the U.S.-listed actively managed Tema Memory ETF (DISK), the allocation to ChangXin has been further increased—ChangXin has been adjusted to the fund’s largest holding, with a stake as high as 12.97%.

The overseas capital’s shift in allocation to China’s storage assets reflects that the global semiconductor cycle is entering a new stage.

In the past two years, the main investment theme for AI has been highly concentrated in GPUs, AI servers, and cloud computing infrastructure. NVIDIA $NVDA became the biggest beneficiary thanks to its GPU advantage, while players in areas such as Broadcom, TSMC, and advanced packaging have also continued to capture market premiums.

But as AI model scale expands and compute demand increases, storage is becoming a new core link.

Large-model training requires substantial HBM, while high-performance inference needs storage support with higher bandwidth and lower latency. Demand for DRAM, HBM, and advanced storage solutions in AI servers is changing the traditional cyclical characteristics of the storage industry.

That’s also why the U.S. stock market has recently started paying renewed attention to the storage sector.

Companies such as Micron $MU and SK Hynix have already benefited meaningfully from AI storage demand. HBM has become one of the fastest-growing directions in the semiconductor industry chain. At the same time, after experiencing the inventory adjustment from the previous cycle, global storage vendors have become more cautious with capital expenditures, and the supply-demand situation is improving.

From an investment perspective, I believe the biggest change in the storage industry is that it is moving from the logic of traditional cyclical stocks to the growth logic of AI infrastructure.

In the past, when the market looked at storage, it focused on inventory, prices, and cycle turning points. Now the market is beginning to reassess the strategic value of storage in the AI era.

CXMT’s inclusion in overseas ETFs essentially indicates that global capital has started to re-examine the position of Chinese semiconductor companies within the core supply chain.

Of course, changes in ETF holdings do not represent short-term trading signals, but they provide an important perspective: the revaluation of value across the AI industrial chain is spreading to more foundational components.

In the coming years, AI competition will not only happen at the model layer—it will also unfold at the levels of chips, storage, energy, and infrastructure.

The market over these past couple of days has really tested traders’ mindset.

In yesterday’s move, many people saw the trend had already started to play out—the technical setup and fund flows were both pointing toward further upside. But the market chose to reverse right at a crucial point.

A round of liquidation pressure then wiped out a lot of short-term profits, pushing them back.

In today’s crypto market, it’s not that people can’t figure out the direction sometimes—it’s that it’s hard to hold positions.

When prices rise, you worry it’s a fake breakout; when prices fall, it’s easy to panic and cut losses. Especially in perpetual futures, volatility is amplified by leverage even further. Many traders don’t lose because their judgment was wrong—they lose because their execution fails.

Personally, I think the most important thing at this stage isn’t increasing trading frequency, but reducing unproductive actions.

In the past, many people were used to hunting for the “next wave.” But in a high-volatility environment, what truly has value is waiting. When the market hasn’t formed strong enough consensus, staying in cash can be a strategy.

That’s also why I use @0x_aix.

Its logic isn’t about having AI trade nonstop every day. Instead, it breaks trading decisions into multiple condition checks. Trend direction, technical indicators, trading volume, and market structure need to align with several signals at the same time before it considers an entry.

This approach is closer to risk management in a professional trading system.

Because most of the time, the market doesn’t actually offer many opportunities worth taking. Chasing breakouts and panic selling during range-bound chop looks busy, but over the long run it often just increases fees and the probability of making mistakes.

Another point worth noting is that the strategy adjusts itself based on performance. When staged returns become more volatile and the Sharpe ratio declines, it will reduce trading frequency and raise the opening criteria—rather than continuing to expand risk.

In my view, the real value of AI trading isn’t to replace humans in predicting every rise and fall. It’s to help traders reduce emotional interference.

Especially in an environment like this, where there are frequent “needle” moves and trends can reverse easily, staying disciplined matters more than catching a single good move.

The crypto market never lacks opportunities. What it lacks is the ability for your account to remain stable until the opportunity actually appears.

117K users, $3.7B in trading volume, $170M liquidity locked

These numbers aren’t just testnet data—they’re a stress test in a real market environment

Over the past period,
@Hertzflow_xyz
used the Testnet to validate the operating capability of decentralized perpetual trading infrastructure under large-scale user participation. From trade depth, to liquidity pool size, to user interaction frequency, the system has undergone continuous testing across different market behaviors.

For a perpetual DEX, the real challenge has never been simply building a trading interface—it’s maintaining a stable trading experience amid high-frequency trading, extreme market conditions, and changes in liquidity.

HertzFlow uses an independent market liquidity pool design to isolate risk across different trading pairs, preventing single-market volatility from affecting the entire pool under traditional shared liquidity models. At the same time, through the Vault mechanism, liquidity providers can participate in the market in a more automated way, and improve capital efficiency through strategy optimization.

The $3.7B trading volume accumulated during the Testnet phase represents not only user growth, but also that the product logic has been validated through real user behavior.

A $170M locked-in position also shows the market is recognizing the value of this new on-chain derivatives infrastructure.

As the Testnet nears its end, all testing, feedback, and data accumulation will become the foundation for Mainnet.

Mainnet isn’t the endpoint—it’s where HertzFlow begins.

The future DeFi trading market will require more than just lower trading costs. It also needs infrastructure capable of handling real user scale, complex trading demands, and long-term liquidity.

HertzFlow has completed the first phase of validation. In the next stage, it will prove its long-term value in real markets.

The trading experience on BNB Chain is entering a new phase of change

@Velvet_Capital
has officially launched Gas-free trading, so users no longer need to prepare BNB in advance for transaction fees, further simplifying the trading process

In the past, when trading on-chain, Gas was a hurdle you couldn’t avoid

Especially for new users entering the BNB Chain ecosystem, if there isn’t native BNB in their wallet, they can’t complete basic operations. Even if they just want to participate in a Meme coin trade, they need to first buy BNB, bridge, top up, and then start trading—adding to the learning curve and reducing users’ willingness to enter the on-chain world

VelvetX solves this problem with an embedded wallet

Users can create a wallet directly with their email or X, and the platform automatically covers the transaction Gas fees. Users don’t need to worry about the details of on-chain fees—just focus on asset selection and the trade itself

The significance behind this isn’t just reducing one extra click

More importantly, it’s driving on-chain trading to shift from “for crypto-native users” to “for everyday users”

In the past, the core logic of DeFi was to help users understand foundational concepts like wallets, Gas, signing, and network switching. In the next stage of competition, the focus will increasingly be on who can hide complexity, so users can complete on-chain actions just like using traditional financial applications

For Meme trading, speed and experience are especially important

Market opportunities are often fleeting. When users miss trades due to lack of Gas or complex processes, it’s essentially a product experience issue

VelvetX lowers the entry barrier with Gas-free trading, making Meme trading on BNB Chain more direct and efficient

The large-scale adoption of on-chain applications doesn’t rely only on better protocols and liquidity—it also requires product design that’s more aligned with ordinary users’ habits

Reducing complexity is the first step to expanding the market boundary

HertzFlow is entering the final phase before the Mainnet launch.

To complete the infrastructure preparations for the mainnet launch,
@Hertzflow_xyz
performed a user data reset on the testnet on July 29 (Wednesday) and simultaneously completed the smart contract upgrade.

This update mainly covers three areas:

1️⃣
Reset of the testnet contract environment

Previously, contract interaction records generated by users on Testnet will be cleared, including some test transactions, strategy execution logs, and more.

This means the old version contract environment has officially come to an end, and future testing will be based on the new contract architecture.

2️⃣
Recommended relationship data fully preserved

Compared with simply resetting test data, HertzFlow retains users’ referral relationship data.

This indicates that the ecosystem contribution system will not restart from scratch due to the testnet reset—community relationships built by early participants remain valid.

3️⃣
The new contract is prepared for Mainnet

A contract upgrade usually means the project has entered a critical verification stage before going live. Further confirmation is needed for:

Contract logic stability
User interaction flows
Strategy execution efficiency
On-chain fund security mechanisms

For DeFi protocols, the Testnet stage is mostly about validating product usability, while Mainnet is the true test of liquidity, user demand, and the protocol’s operational capability.

As HertzFlow moves from the testing phase to Mainnet, its product logic will shift from:

Test strategy experience
→ real on-chain trading environment
Simulated yield model
→ actual yield market

Early user participation
→ ecosystem value accumulation

All current signals point in the same direction—Mainnet has entered the final preparation stage.

August is a great month, and it’s a good time for the mainnet launch

In the past few years, DeFi has been solving a problem: how to get more capital on-chain.

But as the market has evolved, what truly limits users from participating is no longer the assets themselves, but the complexity of the strategies.

Liquidity mining requires choosing pools, tracking trading volume, assessing risk, and continuously adjusting positions.

For ordinary users, LPs are more like a job that needs ongoing management, rather than a passive yield product.

HertzFlow Vaults (HzV)
@Hertzflow_xyz
What it aims to do is to repackage this entire process.

With a single deposit, it can cover multiple HzLP markets. Under the hood, it’s not betting on a single trading pool; instead, it uses a combination of multiple markets to generate fee income and returns driven by trader losses.

This is actually very similar to traditional asset management: diversified allocation, dynamic rebalancing, and automatic compounding.

The Vault handles strategy execution, while the curator is responsible for selecting and rebalancing. Users only need to hold HzV.

I believe this kind of product represents a clear trend in DeFi. In the future, the competitive focus won’t be only how high the APY is—but who can turn complex strategies into simple assets.

Early DeFi competed on liquidity scale; later it competed on trading efficiency. In the next phase, it may enter an era of “strategy productization.”

Users don’t need to become professional LPs or research the market every day. They only need to choose strategy assets that fit them.

The value of HertzFlow Vaults is not just providing a yield tool, but attempting to move on-chain liquidity management from manual operation toward automated asset management.

As AI agents, automated strategies, and on-chain financial infrastructure become further integrated, the way people use DeFi may change in new ways.

Past trades are, in essence, people looking for opportunities in the market

You watch the screen, filter coins, track fund flows, analyze sentiment, judge trends, then manually place orders. But the problem with the crypto market is that information is always more than a person can process

Thousands of trading pairs every day, multiple exchanges moving at the same time, on-chain capital, derivatives open interest, and market sentiment changing continuously—it's hard to reliably capture valid signals with manual effort alone

This is also the problem that AI trading Agents need to solve

The core of @0x_aix is not just a simple coin-picking terminal, but providing real-time market awareness capabilities for AI Agents

A coin-picking terminal is more like an Agent’s radar

It connects to data sources from six major exchanges, aggregates different market dimensions, and performs real-time analysis of asset status using multiple indicators such as liquidity, volatility, funding rates, OI changes, fund flow direction, market sentiment, and community heat

The goal isn’t to tell users which coin will go up. Instead, it gives the AI Agent more complete data inputs when facing the market

When an asset’s price shows abnormal volatility, the Agent can combine changes in trading volume to determine whether it’s normal movement or capital-driven

When contract OI increases rapidly, the Agent can tell whether new funds are entering

When funding rates and sentiment indicators reach extreme levels, the Agent can identify potential risks or trading opportunities, and then—together with a Z-Score anomaly detection model—AI can spot abnormal market changes earlier

Real AI trading shouldn’t just be a chatbot

It needs to continuously acquire market information, understand changes in capital and sentiment, build trading logic, execute strategies, and then adjust based on the results

Coin selection is only the first step. Data awareness, strategy assessment, and automated execution—the real value of an AI Agent trading system

In the future, trading terminals won’t compete only on who provides more market data, but on who can make AI closer to a true trader
AIX is building such an AI trading

This week’s U.S. stock market is entering the true stress-testing phase of the Q2 earnings season

The market’s focus is no longer just how much companies are earning. Instead, as the AI investment cycle moves into its second phase, attention is on which companies can still prove that capital expenditures are translating into real demand

After Tuesday’s close, KLA and Seagate (STX) first fire off the “signal battle”

KLA’s results are more like a thermometer for the semiconductor supply chain—it reflects the health of wafer-fab capacity expansion, advanced process nodes, and the AI chip manufacturing segment. If AI infrastructure keeps up its rapid investment pace, orders on the equipment side will show it first

Seagate represents another logic: data center storage demand

Previously, the market focused more on GPUs. But as AI model sizes expand, data storage, data access, and enterprise hard-disk demand are becoming the new bottleneck. Seagate’s performance will test whether AI infrastructure is truly spreading from the hardware layer to the application side

Wednesday is the busiest day this week

Before the bell, SK hynix will release earnings, with HBM remaining the key observation metric

The biggest change in AI semiconductors this year is that value creation has moved from a single focus on GPUs to the entire supply chain. HBM has become an indispensable part of AI accelerators. Improvements in storage vendors’ profitability, at their core, reflect whether AI compute demand can keep growing

After the close, Microsoft, Meta, Qualcomm, and ARM all take the stage

These companies represent four different paths:

Microsoft: cloud computing and the commercialization of Copilot
Meta: whether AI infrastructure investment will keep ramping up
Qualcomm: whether on-device AI can open up new markets
ARM: a reappraisal of chip architecture value in the AI era

What’s truly worth watching isn’t the earnings figures themselves, but management’s judgment on capital expenditure, AI product revenue, and commercialization timelines over the next few quarters

In the early hours of Thursday, Apple and Amazon deliver the closing acts

For Apple, the question is whether AI can become a new growth engine rather than relying solely on the hardware cycle

Amazon, meanwhile, needs to show that AWS still has an edge in the AI cloud services competition

At its core, this earnings season is about verifying whether the AI investment cycle can complete the shift from capital outlays to commercial returns

If AI infrastructure continues to grow rapidly but revenue fails to materialize, the market will re-examine valuations

If cloud, storage, chips, and the application side all show revenue acceleration at the same time, then the AI industry will enter a new valuation stage

25 tech companies signed an open letter about the importance of open-source models. The media is busy counting the names on the signature list, but what I see is a trading table that’s being reshuffled

The most worth reading part of this letter isn’t who signed it

It’s Jensen Huang—rarely speaking publicly on X

A person who sells shovels choosing open source as the opening line of their social media isn’t an expression of values; it’s a strategic declaration. For the next five years, $NVDA plans to pour $26 billion into open-weight models—not because Jensen Huang believes in the spirit of open source, but because he sees a bigger business than selling GPUs

Becoming Wintel in the AI era

We’ve all learned the Wintel story: Microsoft makes the operating system, Intel makes the CPU—together defining the hardware standards and software entry points for the PC era. Not the cheapest, not the fastest, but the standard. Once you become the standard, all software developers, all OEM vendors, and all enterprise IT procurement pivot around your architecture

Jensen Huang is doing the exact same thing with open-source models

CUDA is step one—getting all AI developers accustomed to writing code on the NVIDIA architecture. Open-weight models are step two—helping AI application-layer startups grow, raise funding, and exit within an ecosystem at $NVDA . When you train with Llama, run inference with TensorRT, and deploy with NIM, your entire tech stack is operating on NVIDIA’s track

This isn’t a moat
This is gravity

That’s why $MSFT signed, and why Azure needs to become the largest model-inference platform of the AI era: one hand capturing the premium from OpenAI’s closed models, the other embracing the compute demand brought by open-source ecosystems

That’s why $META signed—because what Llama is opening isn’t the model itself, but an ecosystem entry point. By expanding the developer base, Meta is redefining how AI models are distributed

As for why OpenAI and Anthropic didn’t join—at bottom, it’s about different business paths. Their valuation logic still comes from model capability and API revenue, not the scale of the ecosystem created by open weights

The question isn’t which will win: open source or closed source
The question is that when the computing standard for AI is defined by a handful of companies, most participants can only compete within those ecosystem rules

Open and closed can coexist
But there can only be one firm holding the power to define the standard

What’s the most troublesome thing for everyone when they’re holding a big cake in their hands day to day?
It’s definitely when you suddenly run short of money, or you want to go somewhere else to hunt for some wild game but you don’t have any bullets—you end up agonizing over whether to sell.
If you sell, you’re scared it might take off and fly off in the next few days—that’s absolutely the kind of thing that makes you slap your thigh in regret.
If you don’t sell, watching other opportunities to make money pass by while you just stare is also really unbearable.
In the past, if you wanted to use the big cake in your hand as collateral to borrow a bit of money, the mental burden was genuinely heavy.
You had to transfer the coins to some platform, or wrap it into some token with packaging—this is so convoluted.
It’s like you want to borrow some money for working capital, but the other party insists you first transfer the property deed to them.
There’s no way your mind can be at ease at night—who knows whether those people will abscond with the funds.
But recently, looking at the lending route that @BabylonLabs_io is doing, they’ve really managed to sort this out.
The most amazing part is that your big cake always stays right there in your own wallet.
Your private key is always tightly held by you alone—nobody else gets to touch it.
With that kind of complete self-control, you can actually use it directly as collateral to borrow money.
No need to go through those unreliable cross-chain bridges, and no need to grovel to find some middlemen.
It’s purely powered by underlying cryptography code to keep the lending system running—this is way better than handing your money over to someone else for custody.
Actually, if you think about this business model further, their token $BABY will probably have a pretty big role to play later on.
Think about it—in a lending system where there’s absolutely no middleman earning the price spread,
whether it’s handling settlements or giving subsidies to those who provide capital, there must be something that moves around in the flow.
At that point, token #baby will likely act as a lubricant within this ecosystem.
Maybe later, you can also use baby to deduct some borrowing fees, or participate in some other revenue-sharing.
Anyway, for people like us who just want to hold coins calmly and steadily, but sometimes need a bit of cash flow in a pinch,
this native setup really counts as solving a major headache.
It keeps your core holdings intact without going on silly adventures, and yet you can borrow out some money anytime for emergencies.
If you’re also always watching the big cake in your wallet, and you feel it’s a shame that your funds are locked up like that,
then you really can go learn about their fully native lending approach.
And you can also keep an eye on what new coordinated玩法 (gameplay/feature) baby’s token will come out with next.

You hand Bitcoin to a custodian, get back a receipt, and then use that receipt as collateral in DeFi—this isn’t Bitcoin entering DeFi. It’s Bitcoin being staked with the custodian while you carry around a IOU, playing with it.

The WBTC model is essentially the same path. You lock the real $BTC into a multisig address. The custodian mints an ERC-20 token voucher for you, and then you use that voucher to borrow, trade, and generate yield. It sounds fine, but there’s an extra layer of trust in the middle: an additional counterparty, and another potential point of failure. The several “blowups” in 2022 taught the market one thing: the most dangerous risk in DeFi is never actually the smart contract bug—it’s the person you assume would definitely never run away.

The TBV made by @BabylonLabs_io Babylon—Trustless Bitcoin Vault, a trustless Bitcoin treasury—takes a completely different route. Native Bitcoin is used directly as collateral. There’s no need to wrap it as WBTC, no need for cross-chain bridges to move it around in the middle, and no need for any intermediary to hold your private keys. You lock your BTC into a cryptographically controlled UTXO vault, and then you can borrow USDC, USDT, and in the future, borrow in any application on any chain—lending, stablecoins, credit cards, derivatives, insurance—using the same native Bitcoin collateral. The assets always remain on the Bitcoin network. The collateral status is verified by Babylon’s protocol layer; Aave executes the borrowing logic on Ethereum. The asset layer, verification layer, and application layer are decoupled into three parts. If any one layer fails, it won’t pull the other two down with it.

The issue isn’t whether Bitcoin can be used as collateral—the issue is whether you’re willing to hand over Bitcoin itself first, for the sake of using its value.

That’s exactly what TBV solves: the “trust substitution.” Replacing faith that an entity won’t do evil with trust in cryptography and economic security. Replacing centralized custody risk with Bitcoin network’s own security model. The significance isn’t that a new protocol launches, but that for the first time, Bitcoin can become a foundational collateral layer for global DeFi without leaving its own chain and without handing anything to anyone.

A trillion-dollar worth of assets
No need to trust anyone
Works in DeFi

This narrative is bigger than any L2
$BABY #baby

Tonight after the U.S. stock market close, investors are watching a company in particular:

$INTC Intel

The significance of this earnings report may be more than just Intel’s own performance. It’s an important checkpoint for whether the semiconductor cycle repair can keep going. Over the past month, the $SOX semiconductor index has rebounded sharply. What’s driving the rally isn’t simply $NVDA hitting a new high.

Rather, the market is starting to price in a bigger story: improvements in the memory cycle; simulation chips bottoming out; continued investment in AI infrastructure; and whether Intel Foundry can gradually move from a strategic narrative to real commercial validation.

What investors are truly watching tonight isn’t only whether revenue and EPS beat expectations.

The market cares more about whether progress on the 18A process meets the plan, the pace at which external customers are onboarded, developments in the advanced packaging business, and whether Intel Foundry can earn more market trust.

Intel’s biggest challenge in the past few years hasn’t been its technology roadmap—it’s been the market’s reassessment of its execution capability. Compared with TSMC, which has already built a manufacturing ecosystem, Intel Foundry needs to prove itself with mass-production results.

Another key variable is the data center business.

The market is watching whether AI demand is spreading from the training stage into the inference stage.

If AI inference keeps growing, not only GPUs will benefit—CPUs, storage, network chips, and advanced packaging will all see more demand.

That’s also an important focus for Intel’s data center segment.

The PC business is worth watching too.

AI PCs are becoming a new direction for the industry, but the market needs to see whether they genuinely drive an upgrade cycle—rather than just being a product-refresh concept.

Finally, look at the full-year guidance. If management raises expectations, the market may reevaluate the durability of the semiconductor cycle repair. If guidance is weak, it could suggest that the recent rally was largely just an oversold rebound.

The key focus in tonight’s trading isn’t guessing numbers.

It’s seeing whether the earnings report can answer one core question:

Is the AI investment cycle spreading from GPUs to the entire semiconductor supply chain?

Intel’s answer may influence where technology-sector capital flows in the next phase.

An overnight repricing of AI capital expenditures and hardware-chain sentiment

Not the Fed
Not inflation data
It’s the earnings reports of four companies after the close tonight

First, let’s clarify why tonight matters

Last week, semiconductors and AI hardware just caught a breath
Memory rebounded from the lows, optical modules saw a rebound, and the Philadelphia Semiconductor Index inched higher

But how far can this rebound go?
The answer isn’t in the technicals

It’s in Google’s CapEx guidance

Let’s break it down one by one

$GOOGL — The No. 1 AI barometer this week

Three numbers decide everything:
Google Cloud growth rate
Incremental contribution of AI to ad revenue
Most importantly—capital expenditure guidance

If capex is raised to increase compute spend, it means Google will keep pouring money into building clusters
Optical modules, networking chips, and server hardware all benefit
The rebound in AI hardware shifts from oversold repair to being driven by fundamentals

If capex is lowered, this AI hardware repair rally ends outright—no middle ground

$TSLA — Tesla

The key isn’t whether it sold 400,000 cars or 420,000 cars
The key is two things: vehicle gross margin, and how FSD is making money right now

Gross margin continues to be squeezed by price cuts, and the valuation framework for the whole growth-stock complex has to shake too; if FSD shows real, substantive commercial signals, tech risk appetite could be mildly reignited

But Tesla tonight isn’t the main character

$IBM’s previous earnings warning has already sent the stock down more than 20%

The focus this time isn’t how much IBM itself can rise—it’s whether it can validate one thing: are enterprise IT services and government/enterprise AI orders truly real

If it can’t even earn money from consulting and services,
then AI isn’t just selling shovels
Even the people buying shovels are starting to hesitate

$TXN — Texas Instruments

A barometer for analog chips

Industrial orders → reflect whether global manufacturing is really starting to warm up
Automotive chips → reflect whether automakers dare to stock up

Now put the four together

After-hours numbers—ideal case

Google raises CapEx + Cloud accelerates + AI ads begin to monetize
Tesla’s gross margin holds steady
IBM enterprise orders don’t collapse
TXN industrial demand bottoms out and rebounds

Then that’s the soft-landing script for AI hardware
When the market opens tomorrow, hardware keeps rebounding, and capital rotates from defense back to growth

Worst case

Google cuts CapEx
Tesla gross margin slips
IBM enterprise demand is falsified
TXN industrial keeps falling

Then the rebound from a week ago wasn’t a reversal—it was a run-for-your-life wave

Last night, the U.S. tech sector continued to diverge

Capital has not fully rotated back into tech broadly; instead, it is looking for opportunities within the AI industry chain again. The strongest direction remains AI infrastructure, while consumer tech is under relatively more pressure

The Philadelphia Semiconductor Index ($SOX) rose more than 3% at one point during the session. Although the rally narrowed to +0.6% into the close, the intraday fund flows already released a clear signal: the hardware supply chain is seeing capital come back

Within the sector, the most oversold areas have become the main targets:

Memory: Micron $MU , $WDC (Western Digital), $SNDK (SanDisk) rose nearly 2%
Compute chips:
$MRVL (Marvell) and $INTC (Intel) both rose more than 2%

Optical communications: $LITE and $CRDO both rose more than 4%

The biggest feature of this rally is that capital is not chasing AI leaders that have already run up to high levels again. Instead, it is searching for the specific segments that were previously suppressed by sentiment

The industry fundamentals have not changed meaningfully

GPU clusters are still expanding, HBM supply remains tight, and data centers continue upgrading to faster interconnects such as 800G and 1.6T

The investment cycle for AI infrastructure has not ended because of short-term market fluctuations

Especially in optical communications

As AI server scale keeps expanding, pressure on data transmission inside compute clusters is rising rapidly. High-speed interconnects are becoming a new infrastructure bottleneck. That is also why optical module makers, DSP chips, and network chips are regaining investor attention

At the same time, the AI compute leasing sector saw a sharp rebound

$IREN surged 19% in a single day
$HUT8 rose more than 10%
$CRWV and $NBIS gained more than 6%

However, this direction is still largely an oversold rebound

Earlier, the market had major disagreements about these compute leasing companies’ capital expenditure pressures, business models, and the speed of earnings realization. Therefore, short-term capital returning does not necessarily mean the industry logic has fully reversed

From the overall market picture, investors are repricing the AI industry chain again

The trading logic has shifted from the early “buy the AI story” approach to finding the segments that can genuinely convert orders and profits

The key focus for capital next may concentrate on:

Hardware companies that were mistakenly sold off
Growth names whose valuations return to a reasonable range
Core supply chains that can benefit from the next round of AI infrastructure upgrades

Before I used to trade, I often made a common mistake:

When I saw a big bullish candle, I’d get FOMO and panic-buy to avoid missing out. When I saw a big bearish candle, I’d immediately doubt myself. After reviewing everything, I realized that many times it wasn’t that I couldn’t analyze—it all came down to not being able to control my execution.

In the market, truly valuable signals aren’t about a single candlestick showing up—it’s about whether it shows up in the right place.

Recently I’ve been using @0x_aix’s AI trading skills module. One part, “Price Action Entry,” gave me a very clear impression. It doesn’t simply tell you buy or sell based on one candlestick. Instead, it first judges the market context.

For example, it’s the same bullish candle.

If it appears near a support level, an EMA, or around a trendline, closes near the high, has a short lower wick, and then follows through afterward—this kind of signal quality is much higher.

But if it suddenly spikes up in the middle of a choppy ranging market, many times it’s just noise.

This logic actually matches the thinking of many professional traders.

First, look at the environment. Then, look at the formation.

Another practical point is that it pays attention to second-chance entries.

Many traders like to chase the first breakout, but the first signal often fails. High2 / Low2 confirmations mean the market has already completed a round of testing, so the trade certainty tends to be higher.

In the past, I often changed my plan because of one or two candlesticks.

I’d stop-loss, then immediately reverse.

I’d buy, and then—because of a pullback—sell again.

Now I’m more inclined to watch the full logic play out.

Has the trend changed?
Has the signal been confirmed?
Is the risk within the planned range?

If there are no conditions, wait.
If conditions are met, execute.

In the trading market, most of the time isn’t about finding opportunities—it’s about avoiding mistakes.

Many people think AI trading is fortune-telling and predicting the future, but that’s not true.
Its real value is that, as an emotionless machine, it helps cut off emotionally driven decisions—then hardwires mature trading rules into an execution system.

Price action, risk control, discipline—these are the fundamentals that help you last in the market.

@0x_aix makes me feel that at this stage, AI isn’t just here to send you buy/sell signals. It’s here to force you to become a disciplined, mature trader.