Big Tech Pushback,Google Slump,Buffett Move

● BigTech-Squeeze,Google-Drag,Buffett-Bet

Big Tech Tightens Pressure on Memory Semiconductors, Alphabet Shares Lag, and Buffett-Style New Bets in One View

The key issue is not simply that “AI semiconductors are attractive,” but that the bargaining power between Big Tech and memory semiconductor suppliers is shifting again.

Also relevant are the reasons Alphabet shares have been lagging the broader market and how individual investors should interpret Warren Buffett-style bets.

This report also separates out issues that are often glossed over in other coverage: the real bottleneck in the AI investment cycle, the structural reasons Alphabet’s valuation is under pressure, and the risks of simply following Buffett’s purchases.

1. Big Tech Begins to Push Back on Memory Suppliers: The Power Balance Is Shifting

The center of AI infrastructure investment initially began with Nvidia GPUs, but the bottleneck is increasingly expanding into memory semiconductors and advanced packaging.

In particular, high-bandwidth memory, or HBM, has become a critical component that determines AI server performance.

SK hynix, Samsung Electronics, and Micron have returned to the market’s focus for this reason.

Past structure: Memory semiconductors were a highly cyclical supply-driven industry with large price swings.
Current structure: As AI server demand has surged, companies that secure sufficient HBM capacity have gained stronger negotiating power.
New shift: Big Tech is increasingly unwilling to accept supplier pricing and volume terms as given.

Accordingly, it is too simplistic to conclude that memory suppliers are the clear winners.

In the short term, memory vendors benefit from HBM shortages.

However, over the medium to long term, Microsoft, Google, Amazon, and Meta are moving to tighten control over supply chains directly.

2. How Big Tech Is Pressuring Memory Suppliers

Big Tech’s objective is clear.

It seeks to lower AI infrastructure costs and reduce dependence on any single semiconductor or memory supplier.

First, in-house AI chip development.
Google is developing TPU, Amazon Trainium, and Meta MTIA.
These efforts are designed to reduce dependence on Nvidia GPUs and to restructure memory sourcing around Big Tech’s own requirements.
Second, a multi-vendor strategy.
Rather than relying on HBM from one company, large buyers are increasingly validating multiple memory suppliers.
Even if SK hynix remains ahead, the entrance of Samsung Electronics and Micron can shift pricing power back toward Big Tech.
Third, long-term supply agreements and pre-orders.
Big Tech seeks to secure required volumes in advance while locking in pricing over time.
This provides suppliers with stable revenue, but may limit further upside in excess profitability.
Fourth, optimization of total AI data center costs.
Big Tech is not focused only on chip prices, but on the total cost structure, including power, cooling, networking, and depreciation.
In the end, the key metric is not “the fastest chip,” but “the lowest cost per token.”

3. The Key Metrics That Matter for Memory Semiconductors

When evaluating memory semiconductor stocks, it is not enough to focus only on AI demand.

Investors need to track pricing, volume, costs, customer diversification, and inventory cycles together.

HBM average selling price:
The key issue is whether HBM pricing can remain at elevated levels.
If prices hold, earnings improvement could accelerate.
Yield and mass production capability:
HBM is more difficult to manufacture than standard DRAM.
If yields remain weak, revenue growth may not translate into proportional margin expansion.
Customer qualification:
The speed at which suppliers are approved by Nvidia, AMD, and Big Tech custom-chip programs is important.
Delays in qualification can weigh directly on share prices.
CAPEX expansion pace:
If memory suppliers expand investment too quickly, concerns about oversupply may resurface 1 to 2 years later.
Recovery in DRAM and NAND:
HBM strength alone is not sufficient to fully restore earnings.
Demand recovery in PCs, smartphones, and server memory must also continue.

4. Who Wins: Short Term vs. Long Term

In the short term, the winners are memory semiconductor companies with strong HBM supply positions.

As long as AI server capacity continues to expand, HBM shortages are unlikely to disappear quickly.

Over the long term, however, Big Tech is likely to emerge stronger.

That is because it controls end demand, data, cloud platforms, custom chip design capability, and capital resources.

Time frame	More favorable side	Reason
Short term	Memory semiconductor companies	HBM shortages and continued AI server investment
Medium term	Big Tech and select leading memory suppliers	Long-term supply contracts, technology qualification, customer lock-in
Long term	Big Tech	In-house AI chips, multi-vendor strategy, data center cost control

From an investment perspective, the more important question is not whether memory semiconductors are attractive, but who can retain pricing power for the longest period.

5. Alphabet Share Weakness: Is This Really a “Dow’s Curse”?

When Alphabet shares lag other large-cap technology names, markets often offer multiple explanations.

One symbolic explanation is a so-called “Dow’s curse.”

However, treating Alphabet’s weakness as a mere market superstition misses the core drivers.

There are four main reasons Alphabet shares have been under pressure.

First, AI-driven search disruption risk.
Generative AI could alter search behavior and weaken the profitability of the existing search advertising model.
Alphabet remains the leader in search, but investors continue to question whether AI answers could reduce ad clicks.
Second, AI investment spending.
Alphabet is also investing heavily in AI data centers and semiconductor infrastructure.
If capital expenditure is rising faster than revenue growth, the stock may face valuation pressure.
Third, regulatory risk.
In the United States and Europe, scrutiny over search, advertising, app stores, and data concentration remains elevated.
Regulation is a persistent valuation discount factor for large technology platforms.
Fourth, cloud competition.
Google Cloud is growing, but its market position remains weaker than AWS and Microsoft Azure.
Profitability in AI cloud services is a key variable.

6. Key Metrics to Watch for Alphabet

Alphabet should not be assessed only on the basis of a low P/E ratio.

The market is now focused less on whether large-cap technology is cheap or expensive, and more on whether it can preserve margins through AI.

Search advertising growth:
This remains Alphabet’s core cash-generation engine.
Stable search ad growth would likely support the share price.
YouTube advertising and subscription revenue:
YouTube is Alphabet’s second growth pillar.
Shorts, connected TV, and premium subscriptions are important to monitor.
Google Cloud operating margin:
Margin expansion is now more important than revenue growth.
The key question is whether AI demand translates into stronger cloud profitability.
Speed of AI search monetization:
The key issue is whether AI answer services remain a cost center or become a new advertising product.
Share repurchases:
Alphabet can continue buybacks supported by strong free cash flow.
This is an important support for per-share earnings over time.

7. Should Alphabet Share Weakness Be Taken Seriously?

The answer is yes.

However, that does not mean investors should avoid the stock entirely.

Alphabet still owns one of the world’s most powerful advertising platforms and data assets.

But in the AI era, the market is pricing in a transition in search advertising, rising data center costs, and higher regulatory risk.

For that reason, Alphabet is better approached as a long-term investment based on valuation and cash flow rather than as a short-term momentum trade.

Rate-cut expectations may help growth stocks overall, but Alphabet will likely require evidence of AI monetization before a stronger rerating occurs.

8. The “Oracle of Omaha” Is Making a New Bet: Should Investors Follow?

Whenever Warren Buffett or Berkshire Hathaway discloses a new position, market attention rises sharply.

Buffett’s investment philosophy is straightforward.

Buy good businesses at reasonable prices, and allow time to compound value.

However, copying Buffett’s purchases directly can be risky for individual investors.

First, there is a reporting lag.
Berkshire holdings are disclosed through quarterly filings.
By the time investors see them, the buying may already be complete.
Second, the entry price is unknown.
Buffett’s average purchase price may differ materially from that of individual investors.
Third, the portfolio context is different.
Buffett invests with the full context of cash, insurance float, bonds, Apple, energy, and financial holdings.
A single-stock copytrade by retail investors carries a very different risk profile.
Fourth, the real message is the investment framework, not the ticker.
The key concepts are pricing power, cash flow, low leverage, buybacks, and durable competitive advantage.

9. What the Market Should Read from Buffett-Style Bets

More important than the specific stock Buffett bought is the type of industries he favors and the assets he avoids.

Recent Buffett-style allocations send several recurring signals.

Cash has become more valuable.
In a high-rate environment, short-term Treasury bills can still generate meaningful returns.
This increases the ability to wait rather than chase expensive growth stocks.
Insurance and financial cash flows remain important.
Insurance companies collect premiums first and pay claims later, creating valuable float.
When managed effectively, this can enhance long-term compounding.
Price still matters, even in AI.
Buffett does not reject technology innovation.
He simply does not view even excellent businesses as attractive if the valuation is too high.
Recession risk remains part of the framework.
When markets are optimistic, he prefers to build cash and wait for quality companies to become cheaper.

10. The Most Important Point Missing from Much of the Coverage

The first key point is that AI semiconductor competition is shifting from chip performance to total cost competition.

Nvidia GPUs, HBM, networking equipment, power, cooling, and data center depreciation all matter in determining how cheaply AI models can run.

That is why Big Tech has an incentive to push back against memory suppliers.

The second key point is that Alphabet’s risk is not technological weakness, but the cost of business model transition.

Alphabet is not weak in AI technology.

The issue is whether AI search can generate margins comparable to the existing search advertising model.

Even strong technology can be discounted by the market when the monetization model is changing.

The third key point is that Buffett’s new bets should not be interpreted as simple stock recommendations.

Buffett increases cash when he believes the market is expensive.

For individual investors, the more useful message is that good businesses should not be bought at excessive prices.

The fourth key point is that rate cuts will not lift all growth stocks equally.

Lower rates can support valuation multiples for growth stocks.

However, companies with excessive AI spending, uncertain monetization, or elevated regulatory risk are likely to remain differentiated.

11. Investor Response Strategy

For memory semiconductors, focus on earnings momentum while watching for oversupply.
HBM demand remains strong, but memory remains a cyclical industry.
If CAPEX accelerates too much, the market may discount the risk in advance.
For Alphabet, monitor AI monetization rather than short-term price action.
Search advertising growth, cloud margins, and AI search monetization are the key variables.
A low valuation alone is not sufficient.
For Buffett-style investing, follow the framework, not the ticker.
Focus on cash flow, pricing power, capital allocation, leverage, and shareholder returns.
Chasing a famous investor’s disclosed purchase is not a sound strategy.
In the AI era, the likely winners are the companies that control the full supply chain.
Chip makers matter, but the power of Big Tech remains strong because it controls end customers, data, and cloud platforms.

< Summary >

The core of the AI semiconductor market has expanded beyond GPUs to HBM and the broader memory supply chain.

In the short term, memory suppliers benefit from HBM shortages, but over the medium to long term, Big Tech is likely to regain bargaining power through in-house chips and multi-vendor sourcing.

Alphabet’s share weakness is not simply a market quirk; it reflects AI search transition risk, advertising profitability concerns, cloud competition, and regulatory pressure.

Buffett-style new bets should be read as a statement of investment discipline rather than a direct stock tip.

In the current market, investors need to assess AI growth, cost structure, rate-cut expectations, Big Tech capital expenditure, and U.S. equity valuations together.

[Related Articles…]

*Source: [ 소수몽키 ]

– 빅테크의 본격 메모리 견제 시작? 뒤바뀐 갑을관계의 승자는/ 다우의 저주에 걸린 구글? 주가 부진 염두에 둬야할까/주식의 신이 찍었다? 새로운 베팅 적중할까

● Bitcoin Winter, AI Power War, Stablecoin Revolution

Bitcoin’s Period of Weakness, the AI Infrastructure Race, and the Stablecoin Payments Shift

The key issue is not simply whether Bitcoin will rise or fall.

Bitcoin price is being influenced by interest rates, liquidity, regulatory developments, institutional flows, and the ongoing concentration of capital in AI-related assets.

Added to this is the debate over whether quantum computing could eventually threaten Bitcoin’s security model.

However, a more important issue lies elsewhere.

Before the AI bubble debate, the more relevant focus is the shifting bottleneck across GPUs, HBM, data centers, power infrastructure, and SMRs.

As physical AI and on-device AI expand, stablecoins may evolve from a crypto asset into a payments infrastructure for the AI era.

In other words, the outlook should be viewed not through Bitcoin alone, but through an interconnected structure involving the crypto market, AI semiconductors, data centers, power shortages, stablecoins, and U.S. crypto policy.

1. Why the Bitcoin market is being described as a winter phase

The current challenge for Bitcoin and the broader crypto market is supply-demand imbalance.

Institutional buying has weakened, and retail participation has declined materially.

Some exchanges have reported trading volumes falling by as much as 90% from prior levels.

Lower volume indicates weaker attention, liquidity, and short-term buying power.

At the same time, investor capital has shifted more aggressively toward AI-related sectors.

AI semiconductors, data centers, power infrastructure, and large-cap technology stocks are attracting capital because earnings visibility and growth expectations remain clearer.

By contrast, Bitcoin remains vulnerable when expectations for rate cuts weaken and U.S. Treasury yields stay elevated.

Weaker institutional buying
Lower retail trading volume
Capital rotation into AI-related sectors
Reduced expectations for rate cuts
Disappointment over delayed crypto legislation

Bitcoin should therefore be viewed not only as a price correction, but also as part of a broader liquidity contraction across the market.

2. What the halving cycle suggests about the timing of a bottom

Bitcoin’s most recent halving occurred in April 2024.

Historically, prior cycles showed strong performance for roughly six months to 18 months after a halving, followed by the onset of a winter phase.

Applying that pattern to the current cycle suggests that the downturn may have entered a more visible phase in the second half of 2025.

Bitcoin winter periods have historically lasted two to three years.

However, this cycle is more complex because U.S. crypto policy, stablecoin legislation, the direction of interest rates, and capital rotation into AI are all interacting simultaneously.

The discussion suggested that a bottom could emerge as early as Q4, or potentially continue into the first half of the following year.

For new entrants, a staged accumulation approach may be more appropriate than deploying capital at once.

That said, this should be understood as a risk-management approach rather than investment advice.

3. Trump’s crypto policy: why the legal framework matters more than near-term price

Many investors expect Trump’s pro-crypto stance to directly lift Bitcoin prices.

However, the more important view is that his objective may be less about short-term price appreciation and more about securing leadership over the crypto ecosystem.

For the U.S., the key issue is not Bitcoin’s price alone, but the restructuring of global crypto finance around dollar-based stablecoins.

Bitcoin is a flagship asset, but the broader strategic objective appears to be the reinforcement of U.S. financial influence through stablecoin infrastructure.

If Bitcoin were to surge too quickly, the ecosystem could become more speculative, slowing the move toward institutional adoption and regulatory clarity.

Trump has positioned himself as pro-crypto
However, institutional integration may matter more than near-term Bitcoin appreciation
The U.S. focus appears to be stablecoins and the preservation of dollar dominance
Delays in legislation are weighing on sentiment
Renewed legislative progress could support a price rebound

Accordingly, investors should focus less on statements and more on the actual passage of crypto-related legislation.

Concrete regulatory reform could restore confidence across Bitcoin and the broader crypto market.

Continued delays, by contrast, could prolong the current period of weakness.

4. Can quantum computing break Bitcoin?

Some analysts argue that if quantum computing advances sufficiently while Bitcoin retains its current security architecture, Bitcoin’s cryptographic model could eventually be compromised.

Current quantum systems developed by firms such as Google and IBM are generally described as being at roughly the 100-qubit level.

However, if stable systems with tens of thousands or even hundreds of thousands of qubits become feasible, existing cryptographic methods could face material risk.

The critical point is that Bitcoin is unlikely to remain static.

Although Bitcoin is decentralized and has no central operator, the network can upgrade its security through community consensus via hard forks or soft forks.

Post-quantum cryptography, designed to resist quantum attacks, is already being discussed across the global security industry and within the Bitcoin community.

The correct framing is therefore not that quantum computing will definitively destroy Bitcoin.

Rather, the issue is that faster quantum progress would force Bitcoin to upgrade its security model.

Current security assumptions could face long-term risk if unchanged
Quantum computing would also threaten broader financial systems
Post-quantum cryptography is under active development
Bitcoin can change its security architecture through consensus
Investors should focus on resilience and adaptation, not short-term panic

5. The AI bubble debate: this looks more like an early infrastructure build-out

One of the most common questions is whether AI is in a bubble.

The discussion argued that AI is difficult to classify as a bubble at this stage.

The reason is straightforward.

AI is already demonstrating measurable productivity gains.

It is being used in report writing, coding, translation, design, research, customer support, and content creation.

Unlike the dot-com era, when expectations often ran ahead of execution, AI is already changing workflows across businesses and households.

However, the sector remains in the early phase of a larger cycle.

This was described as AI Stage 1.

Stage 1 is the phase in which AI learns from human knowledge and existing datasets.

That phase requires GPUs, HBM, data centers, and power supply.

AI has not yet fully penetrated manufacturing and services
The market is still focused on model training and infrastructure build-out
This explains the surge in demand for AI semiconductors and HBM
Physical AI and AI agents are not yet fully scaled
The correct focus is infrastructure bottlenecks, not a bubble narrative

From an economic perspective, AI is not just a market theme. It is a long-duration industrial cycle.

Short-term stock performance alone is therefore insufficient to judge the trend.

6. The AI bottleneck is shifting from GPUs to data centers and power

To understand AI, the key is to identify where the bottleneck is moving.

The first bottleneck was GPUs.

Training AI models required Nvidia GPUs, and demand surged.

The next bottleneck became HBM.

GPUs require high-bandwidth memory to function effectively, which brought companies such as SK hynix and Samsung Electronics into focus.

The bottleneck then moved to data centers.

Even if GPUs and HBM are available, AI deployment remains constrained without enough capacity to install and operate them.

Global data centers are currently estimated at more than 10,000 facilities, and some forecasts suggest the number could nearly double by 2030.

The problem is that more data centers require much more power.

The next bottleneck is therefore power infrastructure.

Primary bottleneck: GPUs
Secondary bottleneck: HBM
Third bottleneck: data centers
Fourth bottleneck: power infrastructure
Fifth bottleneck: cooling technology and energy efficiency

This structure is also important for investors.

Capital initially flows into AI semiconductors, but over time it may shift toward data centers, power grids, cooling systems, SMRs, and offshore data center concepts.

7. Why data center power shortages and SMRs are gaining attention

Data centers consume large amounts of electricity.

Power demand comes not only from servers, but also from the cooling systems required to manage heat output.

In some U.S. states, efforts to attract data centers have also increased concern about local power shortages.

Power infrastructure cannot be expanded quickly.

Even a thermal power plant can take five years or more to build, while a nuclear plant may require close to a decade.

By contrast, data center demand can rise sharply within one to two years.

That timing gap creates the bottleneck.

As a result, SMRs, or small modular reactors, are emerging as a key energy alternative for the AI era.

SMRs may be built faster than traditional nuclear plants and could serve as distributed power sources near data centers.

Regulatory, safety, social acceptance, and economic considerations remain significant.

Still, interest in SMRs is likely to remain elevated as AI infrastructure expands.

Offshore data centers have also been discussed as another option.

Locating servers underwater or in marine environments could help leverage natural cooling.

8. Why Jensen Huang is focused on Korea

Nvidia CEO Jensen Huang’s engagement with Korea is not simply a ceremonial event.

Korea holds important parts of the value chain needed for physical AI and on-device AI.

Nvidia makes GPUs, but it does not build smartphones, appliances, vehicles, robots, telecom networks, or power infrastructure end to end.

That is why cooperation with Korean companies matters.

Samsung Electronics: HBM, smartphones, appliances, AI PCs, on-device AI
SK hynix: core HBM supply chain
Hyundai Motor: vehicles, robotics, physical AI through Boston Dynamics
Naver: digital twins, AI platforms, cloud services
LG Electronics: appliances, AI PCs, on-device AI expansion
Doosan Robotics: robotics value chain
SK Telecom: telecom infrastructure linked to AI services
Doosan Enerbility: power infrastructure and SMR potential

Excluding GPUs, Korea has substantial capabilities across AI hardware, manufacturing, appliances, robotics, telecommunications, and power infrastructure.

That is what makes Korea an important partner for Nvidia.

Huang’s visit therefore reflects more than a GPU sales strategy. It also reflects an effort to build a broader physical AI ecosystem with Korean partners.

9. The current state of physical AI and on-device AI

Many people view robot vacuums and service robots as evidence that the physical AI era has already arrived.

In reality, many of these systems are still closer to algorithmic automation than to advanced AI.

Physical AI must integrate language, vision, and action.

For example, if a user says, “Pick up the apple in front of you,” the robot must understand the command, identify the apple visually, and determine how to handle it appropriately as a food item.

That is the level required for true physical AI.

Some logistics warehouses are already beginning to deploy humanoid robots in actual labor roles.

Robots such as Figure AI’s humanoids are moving toward tasks comparable to human workers.

Hyundai Motor’s Atlas platform from Boston Dynamics also demonstrates potential for manufacturing applications.

Going forward, the key will not be movement alone, but how much real-world data these systems can learn from.

That is why Korea, as a manufacturing-heavy economy, is becoming strategically important in physical AI.

10. As AI agents expand, stablecoins may become the payment layer

In an AI agent economy, individuals may no longer handle every payment manually.

AI systems could interpret user intent, select products or services, and execute payments automatically.

For example, an AI refrigerator could detect a shortage of eggs and place an order automatically.

An AI printer could detect low toner levels and reorder supplies on its own.

An AI kiosk could remember a user’s purchasing history and proactively recommend the usual coffee order.

Such systems require a fast, low-cost, cross-border digital payment mechanism.

This is where stablecoins become relevant.

Stablecoins offer lower fees and faster settlement than traditional card payments or cross-border transfers, making them suitable for machine-to-machine transactions.

They may therefore evolve from a crypto asset into the payment rail for the AI era.

However, this area remains shaped by regulation, the international monetary order, dollar dominance, and central bank policy.

The stablecoin race is therefore both a technology competition and a financial power competition.

11. What human competitiveness means in the AI era

AI can now generate knowledge, write text, produce images, code software, and create music.

In this environment, simply having more information is no longer a strong competitive advantage.

The key is not to reject AI, but to use it to expand productivity.

The process of adapting to AI was compared to the five stages of coping with death.

Stage 1: Denial — assuming AI will not replace one’s work
Stage 2: Anger — feeling that AI is undermining professional value
Stage 3: Bargaining — starting to use AI for selected tasks
Stage 4: Depression — becoming concerned that AI performs too well
Stage 5: Acceptance — treating AI as a tool and redefining one’s role

Competitiveness in the AI era does not mean outperforming AI in every task.

It means knowing what to delegate to AI and where human judgment, verification, creativity, and context remain essential.

Prompting skills, validation skills, problem structuring, and contextual reasoning will become more important.

To be competitive above AI does not mean controlling AI. It means learning how to use AI as a lever for work and life.

12. The most important points that are often overlooked in mainstream coverage

First, U.S. crypto ecosystem design matters more than Bitcoin price alone.

Investors should focus on how the U.S. intends to expand dollar influence through stablecoins, rather than on whether Trump will lift Bitcoin in the short term.

Bitcoin is a flagship asset within that system, but the deeper objective may be control over payment infrastructure and financial order.

Second, the focus should be on bottleneck migration rather than the AI bubble debate.

Capital is likely to move from GPUs to HBM, from HBM to data centers, and from data centers to power infrastructure.

Understanding that sequence is essential to reading the AI investment cycle.

Third, power shortages are both a constraint and an investment opportunity.

Power constraints may slow AI deployment, but they also create demand for SMRs, grid infrastructure, cooling systems, and offshore data centers.

Fourth, Korea has a more important position in the physical AI value chain than many recognize.

Nvidia’s interest in Korea is not only about semiconductors.

Korea has the manufacturing, appliance, automotive, robotics, telecommunications, and power infrastructure needed for physical AI.

Fifth, stablecoins may become the transaction language of the AI agent era.

As AI systems begin ordering and paying on behalf of users, the market will require a fast, low-cost digital settlement mechanism.

Stablecoins could occupy that role.

< Summary >

Bitcoin may face a prolonged winter phase due to weaker institutional demand, lower retail volume, fading rate-cut expectations, and capital rotation into AI.

The main catalysts for recovery are U.S. crypto legislation, stablecoin policy, and a rebound in liquidity.

Quantum computing may pose a long-term security challenge to Bitcoin, but Bitcoin could also upgrade toward post-quantum cryptography.

AI appears to be in an early infrastructure build-out phase rather than a bubble phase.

The key bottlenecks are moving from GPUs to HBM, then to data centers, and then to power infrastructure.

Power shortages are increasing interest in SMRs, offshore data centers, and cooling technology as potential beneficiaries of the AI build-out.

Nvidia’s interest in Korea reflects Korea’s importance across manufacturing, semiconductors, appliances, vehicles, robotics, telecommunications, and power infrastructure.

As AI agents and physical AI expand, stablecoins may become an important payment layer in the AI economy.

Human competitiveness in the AI era will depend on the ability to use, verify, and integrate AI into work and decision-making.

[Related Articles…]

*Source: [ 경제 읽어주는 남자(김광석TV) ]

– [풀버전] “고통의 시간이 더 길어질 수 있습니다” 비트코인 투자자가 봐야 할 변수 | 경읽남과 토론합시다 | 김상윤 교수

NextGenInsight.Net

Like this: