24 min read

美元、AI与庞氏:币圈思维炒股指北(上篇)| USD, AI, and Ponzi — A Crypto Native's Guide to Stocks(P1)

Source: http://x.com/i/article/2061909284464861184

* The English version is at the bottom.*

“人类可以没有未来,但绝不会停止掠夺”

法币是人类历史上最有效率、最隐蔽、最能迭代的掠夺工具。

它可以被政权凭空创造 —— 政权只需要持续证明, 这套货币能无条件换得当下维持人类社会运转的核心物料.

问题在于, 人真实的生存门槛只是温饱与繁衍, 低到无法持续证明法币和政权被需要. 所以法币体系必须持续围绕一种"物料"把人的生存成本拉高, 以证明掠夺的合理性. 这个过程, 就是我们说的"发展" —— 不断创造错配刚需, 用新的错配合理化前一轮的错配.

图像

这就是包括美元在内的法币庞氏体系.

这个物料曾经是黄金, 而后是石油. 1974 沙特协议之后, 石油美元支撑美元体系整整 50 年. 但这套机制在 2024 之后开始走向结构性失效 —— 沙特接受人民币结算, OPEC 美元不再自动回流美债, 美元印钞一年比一年快.

它需要一个新物料, 承载未来 50 年, 并且必须赶在石油美元彻底失效之前.

这一次, 它选了 AI. 我称之为**「USD-AI 庞氏」.**

这样的「AI 庞氏论」会被传统美股"正统玩家"嗤之以鼻. 但对于已经熟知

三盘理论

、并在无数次零和交易中,磨练出庞氏嗅觉和对手盘思维的币圈原生玩家 —— 不必妄自菲薄, 更无需放弃自身特长.

本系列旨在帮助这些玩家在不与底层认知冲突的前提下, 快速建立对叙事驱动类股票的理解能力, 以自己最优势的姿势进入股票战壕, 在这场还未完全展开的**「股市币圈化」**中实现利益最大化.

系列分上中下三篇, 本篇为上.

AI上游最终水源 —— 系统性认知美元旁氏

美元本身就是最好的庞氏教科书

“资金需求与预期收益之间的错配产生了一个“缺口”,而这个“缺口”只能通过进一步的错配来填补”

资金需求: 美国财政每年

$2T+

赤字, 政府要持续花钱; 联储要持续印钱满足金融系统流动性.

预期收益: 持有美元资产的人 (债权也好, 现金也好) —— 实际购买力不缩水, 跑得过通胀.

缺口: 政府支出 + 利息复合增长 (赤字一年

$2T+, 累计国债 $36T, 利息支出一年也已经

$1T+

) 跟经济能真实吸收的债务量之间, 长期注定走向不可调和.

短期可以发新债填上 —— 但发新债必须有人买. 国内承接能力有限, 海外结构性减持 (中国持债从 $1.3T 减到 $0.7T), 边际就只剩联储印钱货币化. 而无论是发更多债还是联储印钱, 最终都等于稀释每张美元的购买力 —— 跟持有人"实际回报跑赢通胀"的期待, 天然冲突.

图像

因此美元始终需要一个能承接联储印钞速度的、需求高速增长的"物料".

如果没有物料吸收: 美元推高物价 → 美元资产持有人购买力被稀释 → 抛美元资产 → 抛售规模超过任何买家承接能力 → 联储只能货币化接盘 → 更印钱 → 通胀更失控 → 更多抛售.

美元体系不停制造泡沫、寻找新物料承接美元通胀, 是美元本身的生存刚需.

用三盘结合来看美元: 美债是分红盘. 各国央行、货币基金 (MMF)、美国银行、寿险, 把钱沉进美债, 拿利息. 财政部持续发新债, 利息从税收 + 新债里出 —— 借新还旧, 接力式的入金沉淀. 跟分红盘几何完全一致, 只是这个分红盘有美国军队背书 + 全球结算需求兜底, 撑了 80 年.

联储 - 财政部 - 银行的闭环是互助盘. 联储印钱给银行, 银行买美债, 财政部拿这笔钱发支出, 支出回到银行系统当存款, 银行再买更多美债. A→B→C→A 完整账面循环. 这个互助盘的核心是联储自己的资产负债表 —— 它扩一倍, 循环放大一倍. 2008-2014 扩了 5 倍, 2020 又扩一倍.

美元拆分盘是把美元需求拆给一个全球性"物料"承接. 美元印出来必须有去处, 不然回流就是通胀. 这个承接物就是开篇讲的物料 —— 历史上是黄金, 而后是石油, 现在轮到 AI.

AI是美元当前最大的拆分盘 这次不一样?这次还是一样

币圈外的人最可能反驳: "AI 真的有用啊, ChatGPT 真的提升生产力, OpenAI 一年 $25B 营收, 怎么就庞氏了?"

兄弟们, 此时此刻, 恰如彼时彼刻: "L2 真的有排序器费啊, ZK 真的能产生收益, 怎么就庞氏了?"

L2 和 ZK 都是 ETH 的拆分盘.

L2 的费用是真的. 但它必须依赖新资产发行不断进行来吸收外部流动性 —— 没有新代币、新协议、新应用涌入, 收入立刻干涸, TVL 被抽走. 失败的 L2 几乎全死于这个.

早在 23 年底我就预言过 ETH 这个最大的灾难 —— 资产发行主舞台被 Solana / BSC 抢走, ETH gas + 销毁机制立刻陷入结构性停滞, 价格相对 BTC 一路新低. 不是 ETH 技术倒退, 是它的拆分盘崩盘了.

这时候可能有人反驳: "但 AI 的下游半导体不一样啊 —— NVDA / TSMC / Marvell 是真造东西, 有真现金流, 总是基本面吧?"

真造东西没错, 现金流也没错. 但半导体公司当前的远期盈利预期本身, 完全建立在"美元体系持续注入流动性 + 云大厂持续

**$300B+**

/年 AI 资本开支"这个前提上.

AI 这条链每年

$1T+

的扩张速度, 远远超过终端 AI 公司自身能创造的真实收入 —— 而这部分真实收入, 绝大多数本质上是用 AI 替代人力, 造成不可逆的就业丧失. 必须依赖外部源源不断的资金注入才撑得起.

一旦美元体系庞氏输入停下 → AI 叙事失速估值退行 → 云大厂资本开支停滞 → NVDA 远期营收下降 → 远期 PE 爆炸 → 估值整体被重新定价 → 半导体重回周期股估值.

这跟 L2 / ZK 异曲同工, 区别只是外部流动性接力来源不一样.

完全可以把半导体当前的"基本面", 跟 23/24 年时 L2 的 TVL、地址数、交易量、主流采用类比 —— 都是推动拆分盘的叙事.

https://x.com/thecryptoskanda/status/1720625080869204032

毕竟那时川普说过要用加密货币化债, AI 作为美元通胀承接又何尝不是化债? 核心都是为了让这场流动性接力更有合理性.

三千世界 —— AI产业链内部庞氏结构拆解 AI 上游本身就是互助盘

讲到这, AI 是美元拆分盘的逻辑清楚了. 但在 AI 内部, 上游 (终端应用 + 云大厂 + 芯片设计) 自己就构成一个互助盘 —— 互为营收, 客观上稳住了各家总市值(也就是互助盘中的“可清算资产”):

  • 微软投 OpenAI $13B + Azure 云服务额度 → OpenAI 在 Azure 上花 $12.4B 算力
  • NVDA 借钱给 CoreWeave 买 NVDA 自家 GPU
  • Google 投 Anthropic, Anthropic 在 Google 云上花
  • Oracle - OpenAI 五年长约捆绑
图像

钱根本没出过这个圈, 但创造了三方共赢的账面: OpenAI 年度营收 $25B + Azure AI 收入同比 50%+ 增长 + Mag-7 股价新高. 跟联储 - 财政部 - 银行的互助盘完全同款几何, 只是量级小一档.

如果你看不清楚 AI 是庞氏, 看微软给 OpenAI 钱让 OpenAI 在 Azure 上花就行了.

这个上游互助盘是 AI CAPEX资本支出的"发起地" —— 它决定 AI 资本开支的方向和量级, 也决定整个下游供应链能拿到多少订单.

上游互投把账面可清算资产 (Mag-7 + AI 公司总市值) 稳在高位,任何单个股东想退出时, 都能接近账面价拿回现金. 正是这种"随时退得出"的预期, 让散户 / ETF / 401k 持续买入

这是 AI 拆分盘能维持

$1T+

/年扩张的核心机制 AI拆分盘内部的双重路径

回到拆分盘定义:

"拆分盘是一种庞氏系统, 其总资金在某一特定时间点保持恒定, 但每单位资金对应的权益或资产数量被成倍增加, 同时新生成的权益或资产的价格按比例降低, 以吸引后续资金流入."

AI 作为美元的拆分盘, 需要吸引资金流入. 这个资金流入分两条路径:

  • 一级市场: 通过资本开支以营收和一级募资的形式, 按产业链一级级传导给下游厂商
  • 二级市场: 不按产业链层级, 而是所有 AI 下游互相抢同一池流动性

讲了上游,那么下游则是AI整体的拆分盘:

  • 半导体配套: Marvell, HBM 厂 (SK Hynix / Samsung / Micron), 光模块 (Lumentum / Coherent), 电源 (Vertiv / Eaton), 数据中心 REIT (Equinix / DLR)
  • 代工 + 设备: TSMC, Samsung Foundry (代工); ASML, Lam, KLA, Applied Materials (设备)
  • 上游能源 + 末梢: SMR 核电 / 电力 PPA / 紫苏叶 / 长尾叙事小盘

下游每一家个体, 盘型上是分红盘 —— 投 fab / 投产能 (沉没成本) → 等上游订单兑现 → 利润分红 / 回购. SK Hynix 投 HBM fab, Lumentum 投光模块产线, TSMC 投晶圆代工, 全是同款盘型

当然这里再次强调,下游的庞氏属性并不来源于其本身,而来自上游AI传导。刨去AI热潮导致的重新定价,这些公司估值会回到周期股级别。这里说的庞氏指的是这个溢价,而不是他们原先的“基本面” 一级 vs 二级, 两种结构

图像

当到这里我们其实只看到了一级市场流动性传导 (营收 + 资本开支). 但二级市场投机资金不受限于供应链, 而是按叙事集群走.

当上一级涨幅不再是 α 而是 β (例如 NVDA 涨完了,并且市场都这么认为), 投机资金会找同一叙事下未被充分定价的候选, 不管它们在供应链上是什么层.

例如 白毛股神 @aleabitoreddit "AI 卡脖子叙事" 集群下:

  • Vertiv (数据中心电力 / 冷却)
  • Marvell (NVDA ASIC 合作)
  • Equinix / DLR (数据中心 REIT)
  • Lumentum (光模块)
  • Lam / KLA (设备)

这些在供应链上是不同层, 但二级市场上是同一集群, 互相争抢同一池"AI 卡脖子叙事流动性".

所以拆分盘"拆分率过高"这个崩盘条件, 在二级市场的具体形态不是"突然出现更多赛道企业", 而是:

这种叙事逻辑被普及, 导致更多更低市值、产业链更下层的标的被快速发现获得注意力, 从而在二级市场造成越来越快的资金轮动 —— 也就是战壕玩家熟悉的"分流"的战术

对应 P 小将的打法可以解释为: 买在底部, 出去喊单建设, 获利 (了不了结另说). 他的紫苏叶理论是一个从逻辑上很有道理、很难证伪的理论, 像极了 @MustStopMurad 当时提出的 meme as cult 理论.

白老师完全正确,但他的交易策略, 最重要的一点其实还是**”买在底部"** —— 比如树莓派和 SIVE.

“战壕式交易”在股市里AI拆分盘的合理性

不必争论炒股应该”价值投资“或者”长持“。事实上即使在美股,从2001年dotcom泡沫之后,叙事的生命周期也在大幅度的缩短:

图像

信息传播速度: 报纸 (天) → 电视 (小时) → 互联网 (分钟) → 社媒 + 算法推荐 (秒)

资本跨境流动度: 尤其是 2014 之后亚洲游资的加入战局. 中资带来的是完全不一样的资本炒作战术体系 —— 加密货币本质上也是跨境亚洲游资流速进化的一个分支, 而这个分支带来了一个完全不同于之前美股机构的逻辑.

机构 HFT 化: 与很多人想象的相反, 目前大多数机构并非"价值投资", 而是高频算法量化 —— 占当前美股交易量 50%. 机构交易团队的算法在竞争中不断趋同, 呈现高度拥挤的"一起起高楼, 一起楼塌了"特征.

散户参与提高: "美股是机构主导"的盖棺定论早已成为过去. 随着各类券商一起推进散户交易, 美股散户交易量已经超过 25%, GME 类 meme 股成为可能. 而散户尤其是年轻人, 在过去十年中注意力不断缩短 —— 这成为他们交易逻辑中的物理性限制.

传统市场管这个叫反身性, 三盘理论中这是"受众决定盘型".

————

说到这里,我的读者相信已经了然:

图像
  • 上游互助盘比下游拆分盘稳; 下游内, 越靠近 AI 卡脖子核心 (TSMC / ASML / HBM) 越稳, 越往末梢 (紫苏叶理论 / 长尾) 越脆 —— 母盘锁住资本和预期的能力, 不沿供应链衰减, 沿二级叙事集群的中心距衰减.
  • 交易的目标是在这些拆分盘呈现出哪怕是短线崩盘信号前撤离

玩盘子, 焚诀只有一个: 买的低, 买的早.

那个给”洛克菲勒擦鞋的小孩“可能是一个天选炒股之子,奈何整个市场在一个具体的时间点,剩余可接盘的量是限定的一个值,十万天兵编制有限。来晚了,他只能成为那个”代价“了,成为洛克菲勒的退出流动性

图像

而如何判断自己买的低、买的早, 处在流动性猎人队列的什么位置, 是每个盘圈人一生的修行.

这跟战壕 P 小将的打法没有实质区别: 你判断的不是基本面, 而是有多少人相信这个"基本面叙事", 然后做先信资本 —— 只不过判断的工具和抓手不一样。

那么, 根据什么去判断?

宏大叙事会骗人, 流动性预期不会.

下一篇讲如何从"出入金"流动性来理解、判断自己入场的相对位置.

------------------------ English Version -------------------------

“Humanity can live without a future. But it will never stop predating.”

Fiat currency is the most efficient, most discreet, and most iteratively perfected predatory tool in human history.

It can be conjured out of thin air by sovereign power. The only thing that power must keep proving is this: the currency can always be exchanged for the core materials that keep human society running in the present.

The problem is that human survival requires very little: food, shelter, and reproduction. That threshold is too low to continuously justify fiat and political power.

So the fiat system must constantly rally around a new “demand sink” and push up the cost of survival around it. That is how predation gets justified.

We call this process “development”: manufacturing new artificial necessities to legitimize the previous round of artificial necessities.

图像

This is the fiat Ponzi system — USD included.

That demand sink was once gold. Then it became oil.

After the 1974 Saudi agreement, the petrodollar carried the USD system for 50 years. But after 2024, that mechanism began moving toward structural failure: Saudi Arabia started accepting RMB settlement, OPEC dollars no longer automatically recycled into Treasuries, and US money printing keeps accelerating year after year.

The system needs a new demand sink to carry the next 50 years — and it must find one before the petrodollar fully fails.

This time, it chose AI.

I call it the "USD-AI Ponzi".

The “AI Ponzi thesis” will be laughed off by orthodox US-equity traders. But crypto natives should not be embarrassed by their own instincts.

You already understand the

Three Ponzi Problem. You have trained your "who do you trade against" thinking through countless zero-sum trenches. That is not a weakness when entering stocks. It is the edge.

This series is written for those players: to map crypto-native Ponzi instincts onto narrative-driven US stocks, and to enter the still-unfolding PVP-ification of equities without abandoning the framework that already made them sharp.

The series has three parts.

This is Part 1.

⸺

I. AI’s Upstream Liquidity - Understanding the USD Ponzi

The Dollar Itself Is the Best Ponzi Textbook

“The mismatch between funding needs and expected returns creates a gap. And that gap can only be filled by more mismatches.”

Funding needs:

The US runs a

$2T+ annual fiscal deficit. The government must keep spending. The Fed must keep printing, directly or indirectly, to satisfy financial-system liquidity.

Expected returns:

Holders of USD assets — whether debt or cash — expect their real purchasing power not to shrink. They expect to beat inflation.

The gap: Government spending plus compounding interest —

$2T+ annual deficit, $36T accumulated debt,

$1T+ annual interest expense — is structurally incompatible with the amount of debt the real economy can absorb.

In the short term, the gap can be plugged with new debt.

But new debt needs buyers.

Domestic absorption is limited. Foreign holdings are in structural decline. China’s Treasury holdings fell from $1.3T to roughly . At the margin, only Fed monetization remains.

Whether through more debt issuance or more Fed printing, the end result is the same: each dollar’s purchasing power gets diluted.

That directly conflicts with the expectation of “real returns beating inflation.”

图像

So the dollar always needs a fast-growing absorber for Fed liquidity.

Without one:

USD pushes up prices → USD asset holders lose purchasing power → they sell USD assets → selling exceeds buyer capacity → the Fed is forced to monetize → more printing → worse inflation → more selling.

The dollar must constantly manufacture bubbles and hunt for new demand sinks to absorb its own inflation.

That is not a policy preference.

It is an existential requirement.

The USD Through the Three Ponzi Framework US Treasuries are a Mining Ponzi.

Foreign central banks, money market funds, US banks, and life insurers sink capital into Treasuries to collect yield.

The Treasury keeps issuing new debt. Interest gets paid through taxes plus more debt.

Borrow new to pay old. Relay-style capital deposition.

The geometry is identical to a Mining Ponzi.

The only difference is that this one is backed by the US military and underwritten by global settlement demand. That is why it has lasted 80 years.

The Fed–Treasury–Bank loop is a Pooling Ponzi.

The Fed prints money to banks. Banks buy Treasuries. The Treasury spends. That spending returns to the banking system as deposits. Banks buy more Treasuries.

A → B → C → A.

A complete accounting loop.

The core asset of this Pooling Ponzi is the Fed’s own balance sheet. Double the balance sheet, double the loop.

From 2008 to 2014, the Fed’s balance sheet expanded 5x. In 2020, it doubled again.

The USD Splitting Ponzi splits global dollar demand onto a worldwide demand sink.

Printed dollars must go somewhere.

If they do not get absorbed elsewhere, they flow back as inflation.

That absorber is the demand sink from the beginning of this piece.

Historically, it was gold.

Then oil.

Now AI.

⸺

II. AI Is the USD’s Largest Splitting Ponzi Right Now

People outside crypto might push back:

“But AI is actually useful. ChatGPT really boosts productivity. OpenAI has $25B ARR. How can this be a Ponzi?”

Brothers, déjà vu.

“But L2s have real sequencer fees. ZK has real yield. How can that be a Ponzi?”

L2s and ZK were both ETH Splitting Ponzis.

The fees were real.

But L2s depended on continuous new asset issuance to absorb external liquidity. Without new tokens, new protocols, and new apps flowing in, revenue dried up and TVL got drained.

Most dead L2s died exactly this way.

Back in late 2023, I already predicted ETH’s biggest disaster: the asset-issuance main stage would be stolen by Solana and BSC.

Once that happened, ETH gas and burn structurally stalled. ETH/BTC kept making new lows.

ETH’s technology did not regress.

Its Splitting Ponzi collapsed.

Now someone may object:

“But AI semiconductors are different. NVDA, TSMC, and Marvell build real things. They have real cash flow. That’s fundamentals, right?”

Yes, they build real things.

Yes, the cash flow is real.

But the current forward earnings expectations of semiconductor companies are built on one premise:

The USD system keeps injecting liquidity, and hyperscalers keep spending

$300B+

per year on AI capex.

The AI value chain is expanding at

$1T+ per year — far beyond the real revenue that end-user AI companies can currently generate.

And much of that “real revenue” comes from replacing human labor with AI, creating irreversible job loss.

The whole structure needs relentless external capital injection to stand.

Once the USD liquidity input stops:

AI narrative momentum slows → valuations retreat → hyperscaler capex stalls → NVDA forward revenue expectations fall → forward PE explodes → the whole structure gets repriced → semis return to cyclical-stock valuations.

This is the same geometry as L2 and ZK.

Only the source of external liquidity is different.

Semis’ current “fundamentals” can be mapped directly onto L2 metrics in 2023–2024: TVL, active addresses, transaction volume, mainstream adoption.

All of them were real.

All of them were also narrative fuel for a Splitting Ponzi.

After all, Trump once talked about using crypto to monetize US debt.

AI absorbing USD inflation is not fundamentally different.

The purpose is the same: make the liquidity relay look legitimate.

https://x.com/thecryptoskanda/status/1720816703599522266

⸺

III. Anatomy of the Ponzi Structure Inside AI

AI’s Upstream Is Itself a Pooling Ponzi

Now that AI’s role as the USD’s Splitting Ponzi is clear, let’s zoom into AI itself.

Inside AI, the upstream layer — end applications, hyperscalers, and chip designers — already forms its own Pooling Ponzi.

They generate revenue for one another and stabilize each other’s market caps.

Those market caps are the Pooling Ponzi’s liquidatable assets:

  • Microsoft invests $13B in OpenAI, plus Azure cloud credits → OpenAI spends $12.4B on Azure compute
  • NVDA lends money to CoreWeave → CoreWeave buys NVDA GPUs.
  • Google invests in Anthropic → Anthropic spends on Google Cloud.
  • Oracle and OpenAI lock into a five-year long-term contract.

The money barely leaves the circle.

图像

But on paper, everyone wins:

OpenAI reaches $25B ARR. Azure AI revenue grows 50%+ YoY. Mag-7 stocks make new highs.

The geometry is identical to the Fed–Treasury–Bank Pooling Ponzi.

Just one scale smaller.

If you cannot see AI as a Ponzi, just watch Microsoft give OpenAI money so OpenAI can spend it back on Azure.

This upstream Pooling is where AI capex begins.

It decides where capex goes, how large it gets, and how many orders flow into the downstream supply chain.

Upstream cross-investment keeps the liquidatable asset base — Mag-7 and AI-company market caps — marked high.

As long as shareholders believe they can exit near book value, retail, ETFs, and 401(k) flows keep coming in.

That exitability illusion is the engine.

That is how the AI Splitting Ponzi sustains

$1T+ per year of expansion.

⸺

The Dual Paths Inside the AI Splitting Ponzi

Back to the definition of a Splitting Ponzi:

A Ponzi system where total funds remain constant, but the number of equity or asset units tied to each unit of capital is multiplied, while newly created units are priced lower to attract later inflows.

AI, as the USD’s Splitting Ponzi, needs capital inflow.

That inflow follows two paths.

Primary market:

Capex gets transmitted down the value chain as revenue and primary fundraising.

Secondary market:

Liquidity ignores the supply-chain hierarchy. All downstream AI names compete for the same pool of narrative capital.

We have covered the upstream.

The downstream is AI’s broader Splitting Ponzi.

Semiconductor peripherals:

Marvell, HBM makers such as SK Hynix, Samsung, and Micron, optical-module names such as Lumentum and Coherent, power names such as Vertiv and Eaton, and data-center REITs such as Equinix and DLR.

Foundry and equipment:

TSMC and Samsung Foundry on the foundry side; ASML, Lam, KLA, and Applied Materials on the equipment side.

Energy and long-tail narratives:

SMR nuclear, power PPAs, bottleneck picks, and long-tail narrative microcaps.

Each individual downstream company, viewed by Ponzi form, behaves like a Mining Ponzi.

Invest in fabs or capacity → lock in sunk cost → wait for upstream orders → distribute profits through dividends or buybacks.

  • SK Hynix invests in HBM fabs.
  • Lumentum invests in optical lines.
  • TSMC invests in wafer foundry capacity.

Same form across the chain.

Again: the Ponzi quality of these downstream names does not come from their underlying businesses.

It comes from the AI premium transmitted from upstream.

Strip away the AI-frenzy re-rating, and many of these companies revert to cyclical-stock valuations.

The Ponzi is the premium — not the original business.

⸺

Primary vs. Secondary

Two Different Structures

So far, we have only discussed primary-market liquidity transmission: revenue plus capex moving down the supply chain.

But secondary-market speculative capital does not care about supply-chain hierarchy.

图像

It follows narrative clusters.

Once an upper-tier name stops producing alpha and becomes beta — for example, once the market believes NVDA “has already run” — speculative capital starts hunting for underpriced candidates inside the same narrative.

It does not care where they sit in the supply chain.

For example, under the “AI chokepoint” narrative popularized by Serenity

@aleabitoreddit, the cluster includes:

  • Vertiv — data-center power and cooling.
  • Marvell — NVDA ASIC collaboration.
  • Equinix and DLR — data-center REITs.
  • Lumentum — optical modules.
  • Lam and KLA — semiconductor equipment.

These companies sit on different layers of the supply chain.

But in the secondary market, they belong to the same narrative cluster.

They all compete for the same “AI chokepoint” liquidity pool.

So the collapse condition of a Splitting Ponzi — “the split rate gets too high” — does not appear in equities as “too many new companies suddenly enter the sector.”

It appears as something more familiar to trench players:

The narrative logic gets popularized.

More low-cap, further-downstream names get discovered faster.

Attention gets fragmented.

Capital rotation accelerates.

In trench language: diversion.

Mapped to the PVP playbook, the strategy is simple:

Buy at the bottom.

Go out and shill/build.

Take profit — whether realized or not is another question.

Her Bottleneck Theory is logically sound and hard to falsify, much like Murad’s “meme as cult” thesis back then.

Serenity is right.

But the most important part of her strategy is still this:

She bought at the bottom.

For example: Raspberry Pi and SIVE.

⸺

IV. Why Trench-Style Trading Makes Sense in the AI Splitting Ponzi

There is no need to argue whether stock trading “should” be value investing or long-term holding.

Even in US equities, narrative lifespans have shortened dramatically since the 2001 dot-com bubble.

Information speed:

Newspapers moved information in days. Television moved it in hours. The internet moved it in minutes. Social media and algorithmic feeds move it in seconds.

Cross-border capital mobility:

Especially after 2014, Asian hot money entered the arena.

Chinese capital brought a completely different speculative playbook.

Crypto itself is essentially one evolutionary branch of cross-border Asian hot-money velocity.

And that branch brought a logic very different from the old US institutional framework.

Institutional HFT-ification:

Contrary to popular belief, most institutions today are not “value investors.”

They are high-frequency, algorithmic, quantitative machines — now accounting for roughly 50% of US equity volume.

As institutional algorithms compete, they converge.

The result is a crowded structure:

They build the tower together.

They collapse together.

Rising retail participation:

The old verdict that “US equities are institution-dominated” is outdated.

As brokers compete for retail flow, US retail trading volume has exceeded 25%.

That is what made GME-class meme stocks possible.

Meanwhile, retail traders — especially younger participants — have shorter attention spans than ever.

That is not a moral judgment.

It is a physical constraint embedded in their trading logic.

Traditional markets call this reflexivity.

In the Three Ponzi framework, this is:

The audience determines the Ponzi form.

⸺

By now, the structure should be clear.

The upstream Pooling Ponzi is more stable than the downstream Splitting Ponzi.

Within the downstream layer, the closer a name sits to the AI chokepoint core — TSMC, ASML, HBM — the more stable it is.

The further it drifts toward the tail — Bottleneck Theory picks, long-tail microcaps, peripheral narratives — the more fragile it becomes.

The parent Ponzi’s ability to lock in capital and expectations does not decay along the supply chain.

It decays with distance from the center of the secondary-market narrative cluster.

The game is not to prove that the fundamentals are real.

The game is to leave before the split shows even its first short-term failure signal.

In the Ponzi game, there is only one secret recipe:

Buy low.

Buy early.

That "Rockefeller shoeshine boy" might have been a born stock-market prodigy.

Unfortunately, at that precise moment, the amount of remaining buying power in the market was already finite.

The heavenly army had limited seats.

He arrived too late.

So he became the price.

He became Rockefeller’s exit liquidity.

图像

Knowing whether you bought low, bought early, and where you stand in the liquidity-hunter queue — that is the lifelong cultivation of every Ponzi-circle player.

This is no different from the trench playbook.

You are not judging fundamentals.

You are judging how many people will believe in the “fundamentals narrative.”

Then you become the ape-in-first capital.

Only the tools and indicators are different.

So what do we judge by?

Narratives lie.

Liquidity expectations do not.

In the next part, we will discuss how to use inflows and outflows to understand your relative entry position.