Plunge, Quantum Threat, VC Exodus: Why the Crypto Market is FUD

比推Xuất bản vào 2026-02-12Cập nhật gần nhất vào 2026-02-12

Tóm tắt

The podcast discusses the recent crypto market crash, quantum computing threats, and the departure of VCs, analyzing the underlying FUD (fear, uncertainty, doubt). Experts attribute the Bitcoin sell-off to broader market risk reduction and leveraged positions unwinding, rather than a single catalyst. Quantum computing is highlighted as a significant long-term risk, potentially forcing institutional intervention (e.g., BlackRock) if not addressed, possibly leading to centralized governance changes in Bitcoin. The conversation critiques the 2025 "junk token" cycle, emphasizing that the future lies in cash-flow-generating infrastructure like DeFi and stablecoins, not speculative tokens. Hyperliquid’s rise and Solana’s derivatives gap are noted, while AI is seen as a transformative but capital-intensive trend, with infrastructure plays favored over model companies.

Podcast: Bits and Bips

Release Date: February 11

Hosts/Guests:

  • Ram Ahluwalia: Lumida CEO & Founder, CFA

  • Austin Campbell: NYU Stern Professor, Zero Knowledge Consulting Founder

  • Christopher Perkins: CoinFund President & Managing Partner

  • Nic Carter: Castle Island Ventures Founding Partner

Podcast Summary: BitpushNews


Foreword: In early February, Bitcoin plunged sharply, falling in sync with high-beta assets like gold, silver, and US stocks. This podcast starts from this cross-asset pullback, delving into market structure, institutionalization, and the governance challenges posed by quantum risks. It also confronts the credibility drain following the "2025 junk token wave": crypto might be shifting from a narrative-driven betting field to something more like real-world financial infrastructure.

Nic Carter pointed out in the program that many of Bitcoin's foundational narratives have gradually weakened over time, and institutional concerns, even if premature, have already slowed Bitcoin's adoption rate. Beyond Bitcoin, he also believes the era of venture capital-driven, flashy tokens is largely over, predicting that the future of cryptocurrency lies in more durable, cash-flow-generating businesses rather than speculative token offerings.

Below is a summary of the podcast content:

Market Plunge: Global De-risking or Crypto-Specific Pressure?

During the sell-off on February 5th, Bitcoin briefly fell towards $60,000, one of the worst cross-asset pullbacks in recent years. Gold, silver, and some stocks also fell. Was it global de-risking, or an internal crypto problem? Market rumors suggest the issue might be related to IBIT (Bitcoin ETF) – for instance, a large IBIT holder blowing up in the options market. Some say a Hong Kong fund blew up; others refute this, saying they haven't seen abnormal IBIT redemptions or signs of the "opaque, uncollateralized lending" leverage seen in past cycles. What's your take?

Rahm: This is more like a unified deleveraging of "high-beta assets," liquidations will surface gradually

Ram Ahluwalia: I tend to think it's over-leveraged funds being forced to liquidate during the decline. Twitter also mentioned Trend Research's Ethereum being liquidated. We discussed months ago: bottoms often coincide with "bodies floating to the surface." You're starting to see signs of that now, and more names might be exposed later.

Also, this wasn't an isolated crypto crash; high-beta assets fell together. That day also saw a pullback in US "Mag 7" stocks, like Microsoft's post-earnings plunge – this looks more like a cross-market risk-off event.

Chris: There might not be a "single perfect culprit," derivatives and margin mechanisms amplify volatility

Christopher Perkins: We might not fully understand the "10/11" crash yet, let alone "2/5." Many things take time to become clear. But what I know is: the market is very nervous now, some market makers liquidate more quickly and decisively when margin is insufficient, this "fast liquidation" atmosphere fuels a downward spiral.

Also, institutional money watches basis trades (arbitrage when futures price is higher than spot, allowing for directionally neutral arbitrage). The entry and exit of basis trades also cause price fluctuations, especially when everyone unwinds arbitrage positions simultaneously, creating downward pressure.

Overall, I think: the market will continue to experience some "violent volatility," but it also seems to be gradually "finding a bottom," slowly rebuilding.

Nic: This wasn't a single catalyst

Nic Carter: I'm not sure there will be a "satisfying single catalyst" to explain this sell-off. We were "spoiled" in 2022 – there were many clear blow-up points to point to and say: that caused it.

But often, market structure becomes inherently unstable at some point, a very small trigger can ignite a chain reaction, so it might not require "some big fund blowing up" to explain a decline.

Will BlackRock Step In to 'Save' Bitcoin from the Quantum Threat?

What are your thoughts on the "quantum skeptics" within the Bitcoin community? How does this affect the willingness of capital to enter?

Nic: The key分歧 (disagreement) is on the "timeline," the worst-case outcome is "centralization"

I think the controversy isn't about whether quantum 'exists', but about the timeline. I worry many people mistakenly assume: when quantum becomes truly dangerous, we'll have a long warning period. I don't see it that way.

The key is: completing a quantum-resistant migration for Bitcoin could take nearly a decade. This isn't a question of 'whether the threat arrives tomorrow', but if you don't start preparing now, by the time you confirm it's 'close enough', it might be too late.

I understand the developers' caution: they don't want to make major changes to Bitcoin for a risk they aren't fully convinced of, because the changes themselves introduce new risks. But I don't think this issue is being taken as seriously as it should be. And I don't think betting the future on 'technology progress slowing down' is reasonable. AI is accelerating engineering and discovery, quantum computing is a physics and engineering problem, AI will push it. Capital is also pouring into quantum – this isn't something to be downplayed.

Even if you personally think quantum is far off, institutions will treat it as a real problem. It will enter every investment committee: 'What about quantum?' This will directly slow institutional adoption.

There are two ways to respond: either actually solve it – e.g., introduce post-quantum signatures (could be gradual, optional); or at least address the 'perception problem' – provide a clear roadmap, milestones, and triggers, let the outside world know you're managing it.

But what we see too much of now is: people raising the risk are called FUD. This doesn't reassure institutions.

If developers continue to do basically nothing, I fear one thing will eventually happen: institutions like BlackRock, which carry huge client assets through the ETF, will be forced to act. You are a fiduciary, the problem is left unaddressed long-term, you have no choice.

So I believe, if not resolved voluntarily, it might ultimately lead to a kind of corporate takeover: institutions will 'fire' the current developers and bring in new development forces to push the upgrade. Then Bitcoin would become more like a more centralized chain – that's the governance outcome I fear most.

Chris: Institutional investment committees do ask about quantum risk, it acts like a "throttle/brake" on adoption speed

I'm more optimistic: because the issue is being raised by people like Nic, at least "the risk is identified."

It won't stop adoption immediately now, but it will slow it down. For example, this conversation will appear in every institutional investment committee: "We want to invest in Bitcoin – wait, what about quantum?" This will slow down the process.

I also think quantum shouldn't be viewed in isolation: three major technologies are accelerating together now – crypto, AI, quantum. AI will accelerate quantum R&D, and also accelerate our defensive capabilities. So our ability to respond is also improving.

Nic: Pinning hopes on technology slowing down is wrong, quantum breakthrough might be only 2~3 orders of magnitude away

I agree AI will accelerate quantum, because quantum is essentially physics and engineering, and AI is very good at those.

I oppose the idea of "betting the future on technology not progressing that fast." Look at AI, it crossed multiple orders of magnitude in less than a decade. The gap from "state-of-the-art" quantum to "sufficient to threaten Bitcoin's elliptic curve" might only require 2 to 3 orders of magnitude, not 6 or 7.

And capital is pouring into quantum. 2025 seems to be the biggest year for quantum funding, with private placements injecting roughly ten-billion-level funds, plus national-level inputs from China etc. If the Bitcoin circle continues to deny it, it worries me.

Rahm: Direction is not timing, many hard techs are "right early, but early is wrong"

I'm skeptical from a "time" perspective: being right on direction doesn't mean it happens short-term. The human genome project, nuclear fusion, etc., all prove: breakthroughs can take a very long time. The "real impact" of quantum on markets might not happen that quickly.

But I also agree with a trading reality: your opinion doesn't matter, the investment committee's opinion does. If institutions slow down due to quantum concerns, then Bitcoin must address those concerns to expand adoption.

Institutionalization Narrative, Reflection on the "Junk Issuance Wave"

Speaking of the "institutionalization" narrative. I see in the 13F filings that the largest IBIT holdings are mostly from hedge funds (like Millennium, Jane Street, etc.), and many are underwater. Traditional institutions typically don't like "averaging down on losing positions." This suggests the so-called "institutional demand" is actually limited. It's more like wealth management channels (RIA) selling the product to end clients, rather than large-scale proprietary allocation by institutions. This also makes the price more fragile – this cycle is different from previous ones.

Chris: This also leads to a point: currently many people view "the entire crypto asset class" as a monolith: Bitcoin = crypto. Quantum risk will also be understood holistically, dragging down other projects. Hopefully, the market will mature in the future, differentiating chains and applications based on "fundamentals/cash flow/governance capability."

Nic: Precisely because of this, if Bitcoin's upgrade coordination is difficult, it might be forced into a more "corporate" path: large institutions have custodial responsibilities and governance preferences, if they deem the problem must be solved, they might ultimately push for a "de facto governance takeover."

Q: You all mentioned "the mood is gloomy." What's the reason? Is it because the 2025 token offerings were too poor quality?

Nic: Yes. Many 2025 token offerings "plummeted badly, and were very crappy." People should be angry about the misalignment of crypto VCs and entrepreneurs, just wanting to quickly issue tokens and cash out.

The real civilization-level innovations are:

  • Decentralized Exchanges (DEX)

  • DeFi Lending: Conducting 'banking business' without banks, without counterparties

But many policy discussions (he mentioned certain bills) don't focus on these core achievements, yet market attention is hijacked by "flipping tokens for arbitrage."

Austin: This is like the "extraction-securitization-re-extraction" cycle in many industries: when the incentive structure encourages short-term extraction, market sentiment sours. Relatively speaking, stablecoins/payments/on-chain financial infrastructure are more exciting because real business is happening there.

Chris: There are indeed "high-achieving students making money" and "idling underachievers." If you look at projects using "fundamentals," you can still see very strong innovation in stablecoins, RWA, on-chain equity, etc.

HYPE / Hyperliquid, Why are Derivatives Key?

Community argument: Kyle Samani criticized Hyperliquid (and HYPE) as "everything wrong with crypto," but his former institution bought a lot of HYPE. What's your view?

Chris: If Solana wants to be the "decentralized NASDAQ," it must solve derivatives

I respect Kyle greatly. The Solana ecosystem is also excellent, its vision is like a "decentralized NASDAQ," designed for high-frequency trading, low latency. But if you want to be "NASDAQ," you must solve derivatives. In traditional markets, "derivatives eat spot," profits and liquidity are stronger.

Hyperliquid's sudden rise and dominance in derivatives is a strategic dilemma for Solana. Solana can make progress in directions like DePIN, on-chain capital markets, but it must catch up on the derivatives piece, otherwise the ecosystem narrative will be challenged.

Regarding Hyperliquid: It's not perfect, mechanisms like ADL (Auto-Deleveraging) need improvement, but it innovates and executes quickly.

Nic: Kyle's "exit posture" is very symbolic – "that old style of token VC play" is ending

Kyle is my friend, I don't want to be too harsh. But his public actions (redeeming/leaving/then publicly talking down related positions) are very "dated." From an industry history perspective, this event is very emblematic. Kyle is important because he is typical: highly focused on liquidity, strong convictions, pays little heed to criticism, deeply involved in the "token game." But now – this "way of working" might no longer exist.

I think: The VC-driven, flashy, launch-an-L1-token model, in its current form, is basically over. There will be tokens in the future, but "that era" is past. What remains is the more boring, more traditional financial infrastructure-like part, and that part can actually create cash flow and long-term value.

Austin: I agree many tokens will die. Those that survive will often either have real cash flow or can be professionally traded and valued like mature markets – this will make "crypto" increasingly resemble "the market itself," rather than a unique narrative bubble.

Japanese Election & Global Risk Assets, AI Bubble

One last macro topic: Japanese election, Japanese market rising, 10-year JGB yield near historical highs. What spillover effects does it have on global risk assets?

Rahm: I'm generally bullish on Japan. The Japanese market performed very well last year, Japanese financial stocks are also strong. The bigger theme is: crowded US tech/high-beta assets are being "knocked," capital is spilling over into international markets and value assets. You should watch regions like Japan, Korea, South Africa, South America that are "relatively outperforming."

Rahm: I want to bring the topic to AI: Do you trust Sam Altman? Is OpenAI's valuation real? What about CoreWeave?

Nic: AI is a mega-trend, but model companies are "capital incinerators," private valuations are disconnected from public markets

My trust in Sam is limited, but I'm very optimistic about the AI trend.

The key is: AI capability growth exhibits "super-exponential" characteristics. An important metric is: how much human-equivalent labor (at 50% success rate) can a top model stably complete. This number is jumping rapidly, enough to justify the huge Capex investments. But I still think: OpenAI-type model companies are scary. Competition between models is like a "hot potato," you're strong today, he's strong tomorrow, easily destroying value. Private valuations look like "kind of funny money."

Austin: Could "accounts receivable" from data center companies create bad debt risk?

Model companies have signed massive compute contracts (RPO/performance obligations), but can revenue and cash flow cover them? These contracts might appear as assets/receivables on data center companies' books; if model companies have problems, could they become bad debt? How to view CoreWeave?

Nic: I prefer holding companies in the "AI infrastructure layer": above hardware (NVIDIA) are data centers/compute supply. Model companies are too easily replaced, defeated by user preference changes. Data center companies are relatively more "hedged": even if OpenAI has problems, Anthropic, Google, Grok, etc., might take over capacity. They are closer to an "indexed bet on AI" position than model companies.


Twitter:https://twitter.com/BitpushNewsCN

Bitpush TG Discussion Group:https://t.me/BitPushCommunity

Bitpush TG Subscription: https://t.me/bitpush

Original link:https://www.bitpush.news/articles/7611602

Câu hỏi Liên quan

QWhat were the main reasons discussed for the recent sharp decline in Bitcoin and other high-beta assets?

AThe decline was attributed to a broad-based deleveraging of high-beta assets, forced liquidations of over-leveraged positions, and market nervousness leading to rapid sell-offs by market makers. It was not due to a single catalyst but rather a combination of factors including options market blowups and the unwinding of basis trade positions.

QHow does the quantum computing threat impact institutional adoption of Bitcoin according to the podcast participants?

AInstitutional adoption is slowed as investment committees consistently raise concerns about quantum risk. Even if the threat is not immediate, the lack of a clear mitigation roadmap creates uncertainty. If unresolved, large institutions like BlackRock might eventually force a centralized governance takeover to address the issue.

QWhat criticism did the participants have regarding the token launches in 2025?

AThe 2025 token launches were criticized for being low-quality, speculative, and driven by VCs and founders seeking quick profits rather than genuine innovation. This has damaged the industry's credibility and led to a pessimistic market sentiment.

QWhy is the derivatives market considered crucial for Solana's ecosystem, as per Christopher Perkins?

ADerivatives are essential because traditional markets are dominated by derivatives in terms of liquidity and profitability. For Solana to achieve its vision of a 'decentralized NASDAQ,' it must develop a strong derivatives market, otherwise its ecosystem could face strategic challenges.

QHow do the participants view the investment potential of AI infrastructure companies compared to AI model companies like OpenAI?

AAI infrastructure companies (e.g., data centers, hardware providers) are seen as safer, more durable investments because they provide essential services to multiple AI model companies. In contrast, model companies are highly competitive, capital-intensive, and vulnerable to rapid shifts in user preference and technological obsolescence.

Nội dung Liên quan

Mô hình sinh đã có thể huấn luyện end-to-end? Cốt lõi chỉ là một vòng lặp for

Mô hình tạo sinh (generative models) từ lâu đã phụ thuộc vào các phương pháp đào tạo nhiều giai đoạn (như diffusion, autoregressive), dẫn đến "exposure bias" - sự khác biệt giữa cách huấn luyện và suy luận. Bài báo "Explorative Modeling" (XM) từ UIUC và Harvard đề xuất một giải pháp đơn giản nhưng mạnh mẽ: thay vì tách quá trình tạo mẫu, hãy tách vòng lặp huấn luyện. Trong mỗi bước huấn luyện, mô hình tạo ra K ứng viên, sau đó chỉ chọn ứng viên gần nhất với dữ liệu thực để cập nhật gradient. Vòng lặp `for` đơn giản này giải quyết vấn đề "mode blurring" - xu hướng của hàm mất mát tái tạo đẩy đầu ra về trung bình của các mode hợp lệ, tạo ra kết quả mờ nhạt. Bằng cách khám phá nhiều ứng viên, mô hình học cách bao phủ nhiều mode khác nhau trong phân phối dữ liệu. Khả năng này được gọi là "biểu đạt tạo sinh" (generative expressivity). Thí nghiệm cho thấy "khám phá" (exploration) trở thành một trục mở rộng quy mô hiệu quả thứ ba, bên cạnh quy mô dữ liệu và tham số. Khi quy mô tăng, lợi ích từ việc khám phá càng lớn, cải thiện hiệu suất FLOP, mẫu và tham số. XM đạt được kết quả tạo ảnh trên ImageNet (FID 1.43 không guidance) ngang bằng các mô hình hàng đầu. Quan trọng hơn, XM cho phép huấn luyện mô hình tạo sinh "end-to-end" thực sự. Trong thí nghiệm điều khiển robot, Explorative Policy chỉ cần một lượt truyền thẳng mạng, đạt hiệu suất tương đương Diffusion Policy cần hàng trăm bước suy luận. Điều này chuyển chi phí tính toán từ giai đoạn suy luận sang giai đoạn huấn luyện, mở ra khả năng cho các mô hình tạo sinh nhanh và hiệu quả hơn. Công trình dựa trên lý thuyết "Mode Forcing" trước đó của nhóm tác giả. Hạn chế hiện tại bao gồm chi phí tính toán cho "Forward XM" trên phân phối nhiều mode và thách thức với mô hình ngôn ngữ tự hồi quy thuần túy.

marsbit51 phút trước

Mô hình sinh đã có thể huấn luyện end-to-end? Cốt lõi chỉ là một vòng lặp for

marsbit51 phút trước

Mỗi người tối đa 20 bài? Quy định mới của ICLR bị nhà nghiên cứu DeepMind 'mỉa mai' kiến nghị

Tác giả Google DeepMind Dan Roy đã khởi xướng một kiến nghị trào phúng nhằm phản đối quy định mới của hội nghị AI hàng đầu ICLR, giới hạn mỗi tác giả chỉ được đăng ký tối đa 20 bài báo tham dự ICLR 2027. Quy định này được đưa ra do số lượng bài nộp tăng đột biến 68% vào năm 2026, gây áp lực nặng nề lên hệ thống phản biện. Trong bài đăng của mình, Dan Roy giả vờ ủng hộ việc để AI Agent tạo ra càng nhiều nghiên cứu gia tăng càng tốt, và cho rằng giới hạn này sẽ "làm chậm sự phát triển của AI". Anh còn mỉa mai nhận xét rằng "tất cả công việc phản biện giờ đã được giao cho các mô hình ngôn ngữ lớn xử lý", ám chỉ đến một nghiên cứu trước đó phát hiện khoảng 21% ý kiến phản biện tại ICLR 2026 có khả năng cao là do AI tạo ra. Bài viết chỉ ra rằng thông điệp trào phúng của Roy thực chất phản ánh mối lo ngại về một vòng luẩn quẩn: AI tạo ra bài báo hàng loạt, nhà nghiên cứu gửi đi hàng loạt, và phản biện viên lại dùng AI để đánh giá hàng loạt. Qua đó, tác giả đặt câu hỏi liệu biện pháp giới hạn số lượng cứng nhắc có phải là giải pháp tối ưu cho những thách thức căn bản về chất lượng học thuật trong kỷ nguyên AI này.

marsbit57 phút trước

Mỗi người tối đa 20 bài? Quy định mới của ICLR bị nhà nghiên cứu DeepMind 'mỉa mai' kiến nghị

marsbit57 phút trước

OpenAI không bán tiền dựa trên mô hình đắt nhất

OpenAI đã điều chỉnh giá, giảm 80% cho mô hình GPT-5.6 Luna và 20% cho Terra, đồng thời giới thiệu chế độ Fast cho mô hình mạnh nhất Sol. Động thái này, cùng với việc Anthropic giới thiệu Claude Opus 5 với giá bằng một nửa so với flagship Fable 5, đánh dấu sự thay đổi chiến lược quan trọng: các hãng không còn chỉ tập trung bán mô hình mạnh nhất. Thay vào đó, họ khuyến nghị người dùng kết hợp các mô hình theo nhiệm vụ: dùng Sol (đắt, mạnh) để phân tích và lập kế hoạch phức tạp, sau đó giao cho Luna (rẻ, hiệu quả) thực thi. Điều này cho thấy trọng tâm thương mại đang chuyển từ các mô hình flagship (chứng minh năng lực kỹ thuật) sang các mô hình tầm trung, phục vụ cho lượng truy vấn API khổng lồ hàng ngày của doanh nghiệp, nơi hiệu quả chi phí (ROI) và độ ổn định là yếu tố quyết định. Một yếu tố then chốt giúp giảm chi phí là việc chính các mô hình AI (như Sol) tham gia tối ưu hóa mã nguồn và quy trình vận hành của chính chúng, tạo ra một vòng lặp tự thúc đẩy giảm chi phí. Điều này có thể đẩy nhanh hơn nữa đà giảm giá trong tương lai. Cạnh tranh đang chuyển từ "mô hình thông minh nhất" sang "hệ sinh thái hiệu quả nhất". OpenAI muốn biến các mô hình như Luna thành lựa chọn mặc định cho các tác vụ thường ngày, từ đó gia tăng độ phụ thuộc và chi phí chuyển đổi cho doanh nghiệp. Tương tự như ngành điện hay điện toán đám mây, mục tiêu cuối cùng là AI trở thành một tiện ích phổ biến, hoạt động âm thầm và không thể thiếu trong mọi quy trình, thay vì chỉ là một sản phẩm công nghệ đình đám.

marsbit2 giờ trước

OpenAI không bán tiền dựa trên mô hình đắt nhất

marsbit2 giờ trước

Lương trăm nghìn đô la săn thợ điện, Meta vội vàng tự mở trường dạy nghề

Cuộc đua AI đang vấp phải nút thắt mới: thiếu hụt nghiêm trọng lao động lành nghề như thợ điện và công nhân xây dựng để xây dựng các trung tâm dữ liệu quy mô lớn. Tại Mỹ, dự báo cần thêm 130.000 thợ điện và 240.000 công nhân xây dựng từ 2023-2030 cho hạ tầng AI, trong khi mỗi năm có 80.000 vị trí thợ điện không thể lấp đầy. Các công ty AI sẵn sàng trả lương cao (thợ điện giỏi có thể kiếm 240.000-280.000 USD/năm), nhưng sự chậm trễ xây dựng vẫn gây thiệt hại hàng chục triệu USD. Độ phức tạp của các trung tâm dữ liệu AI, với công suất điện khổng lồ, hệ thống phân phối điện và làm mát phức tạp, đòi hỏi lao động có kỹ năng đặc biệt. Để giải quyết, các gã khổng lồ công nghệ như Meta và Google đang đầu tư hàng trăm triệu USD vào đào tạo. Meta mở trường dạy nghề miễn phí, cung cấp cả trợ cấp sinh hoạt, trong khi OpenAI hợp tác với công đoàn xây dựng. Chiến dịch tuyển dụng nhắm đến cả học sinh trung học, góp phần làm tăng mạnh tỷ lệ Gen Z theo học trường nghề. Tuy nhiên, nhu cầu điện cho AI tiếp tục bùng nổ, chiếm phần ngày càng lớn trong tổng tiêu thụ điện và đẩy giá điện lên cao. Một nghịch lý tồn tại: các dự án xây dựng cần hàng nghìn lao động cùng lúc, nhưng khi hoàn thành chỉ cần số ít vận hành, dẫn đến rủi ro dư thừa lao động lành nghề trong tương lai.

marsbit2 giờ trước

Lương trăm nghìn đô la săn thợ điện, Meta vội vàng tự mở trường dạy nghề

marsbit2 giờ trước

Giao dịch

Giao ngay
活动图片