Alpha³

No.

[Alpha³ @WenfangHan]

@xzzzjpl 小马云真的是马云的私生子?

[ChinoAleman @chinoalemano]

Everyone saw the 1.3 GW plant bankruptcy headline.

Almost nobody knows how to read it.

Let me give you the map, all of it sourced.

This is bullish for $DGXX.

First, anchor the base case: DGXX trades at ~0.6x EV to FY27 revenue TODAY.

With or without this plant.

The $250-300M guidance is built on Alabama, New York and GPUaaS.

Zero of it depends on West Virginia.

Everything that follows is a free option on top of the cheapest name in the sector.

Now the map.

The plant: Pleasants, 1.3 gigawatts.

The asset DGXX signed an LOI on in January.

That was the old management.

Then, in February, management changed.

The creditor side, Tony Robbins' side, took the wheel.

Then, on March 2, one month into the new regime, DGXX publicly extended its due diligence on the deal for an additional 120 days.

In the company's own press release.

Neither regime killed the deal. The door stayed open under both.

The money behind it: Tony Robbins.

Per the WSJ, he put ~$200M into the project and holds a 50% stake.

His entities, TRAG LLC and RG Energy LLC, are the secured creditors per the bankruptcy filings.

The war: the founder, Hodson, lost control in February when the creditor side's forbearance rights installed new management.

In July he tried to buy his way back with a $75.6M payoff check. It was refused.

Ten days later, the new management filed Chapter 11, explicitly to facilitate a sale.

So the side that controls the seller today is the Robbins side.

The side that wants to SELL.

Now the part almost nobody has connected.

Read DGXX's own SEC-filed press release from October 2025, announcing its newest board member:

Ajay Gupta. "Principal of Robbins Gupta Holdings, the exclusive family office for Tony Robbins".

The man who manages Tony Robbins' wealth sits on DigiPower X's board. Amar said it himself on Curzio's podcast: "One of them is on my board, AJ Gupta. And AJ Gupta is a partner with Tony Robbins".

And DGXX's position on the asset: a signed LOI, a load study of up to 1.3 GW, a 200-acre lease framework, zero Omnis equity held, zero capital committed, and a CEO on record refusing to rush: "I cannot rush into a big project that I cannot finance".

A 363 sale is open to any bidder and a Delaware judge supervises every step. The bad-faith fight resolves first. Nobody can promise you the ending.

So here is my call, clearly labeled as my call and not a fact: I believe this process ends with the plant in DGXX's orbit.

The seller's side and the known interested bidder already share one: it runs through a board seat.

The state, owed $50M, needs jobs and a credible future. The plant needs an operator with a real plan. And the cheapest AI infrastructure name in the market needs nothing at all: at 0.6x, the base case pays you to wait while the option resolves.

With the plant: a different company.

Without it: still the mispricing of the sector.

Heads I win big. Tails I win.

Not financial advice. DYOR.

I'm long $DGXX.

[Alpha³ @WenfangHan]

Next. $DGXX @michelamar3
1. ATM is over announcement
2. No future dilution announcement
3. Project financing announcement
4. New contract announcement
5. ...

[Alpha³ @WenfangHan]

@ROIRecruiter @chinoalemano They will deploy llm like Kim K3 in their Token Factory like NBIS. This is the same business like OpenAI and Anthrophic. $DGXX

[ChinoAleman @chinoalemano]

JUST IN: $DGXX is building an AI Token Factory.

Similar to $NBIS.

They revealed it in the last place anyone looks: a job posting.

DigiPowerX just listed a VP of GPUaaS and Token Factory role in Sunnyvale, reporting directly to the CTO.

Read the description carefully, because it contains more strategy than most press releases:

The mission, in their own words: "grow the margin on our power and infrastructure by layering a GPUaaS platform on top of it and steadily improve our tokens per megawatt economics".

Tokens per megawatt.

Write that metric down.

That is the entire thesis of the AI infrastructure era in three words: you own the power, you climb the stack, every megawatt produces tokens, and tokens are what customers pay for.

What the posting reveals about the build:

A Token Factory inference layer running vLLM with an OpenAI compatible API.

An open weight model marketplace and storefront.

Managed Kubernetes, AI workbench, private AI cloud.

A mesh to draw GPU capacity from external marketplaces on top of their own.

That is the $NBIS playbook: full neocloud stack, from infrastructure at the bottom to inference tokens at the top.

Except NBIS trades at $40B+ and rents much of its foundation, while DGXX owns the power underneath and trades at a $325M market cap.

And remember the ladder this climbs: Amar's own numbers put bare metal at $4/hr and GPUaaS at $10-15/hr. This hire IS the climb from $4 to $15.

Same megawatts, triple the revenue.

The posting even names the asset base: 400 MW controlled, "growing above GW+" in their words, with a delivered 40 MW data hall as proof.

Silicon Valley office: real.

Executive hiring: happening.

The strategy the CTO teased with "prefill/decode infrastructures": now a funded org chart.

The market prices $DGXX at 0.6x forward revenue as a construction story.

The job postings say it is becoming a token business.

Not financial advice. DYOR.

[Alpha³ @WenfangHan]

@ShanghaoJin 存储价格是谁推高的？难道不是黄仁勋？

[WquGuru @wquguru]

DeepSeek V4 Flash正式版能力突飞猛进，背后的秘诀是三个字“后训练”，到底后训练、SFT、GRPO、OPD是啥？尽量用浅显易懂的文字来科普一下：

预训练只是让模型读完海量资料打基础。后训练才是精修阶段，专门把它从“会答题”变成“会干活”。

SFT（Supervised Fine-Tuning，监督微调）是用高质量标准问答对手把手教模型，先让它学会正确格式和风格，像老师示范一遍标准答案。

GRPO（Group Relative Policy Optimization，组相对策略优化）是一种强化学习。模型自己反复尝试解题或调用工具，根据结果好坏自动调整。对了就强化，错了就改进，特别适合代码调试和多步Agent任务。DeepSeek还用自研沙箱做大规模真实工具环境练习，让模型在实际执行中成长。

OPD（On-Policy Distillation，在线策略蒸馏）是先分别培养数学、代码、Agent等多个专家模型，再用蒸馏把它们的本事融合进一个模型，既专精又全面。V4还把以前的混合强化学习整段换成了OPD，效率更高。

这���Flash架构和参数完全没变，只重新针对Agent跑了后训练，多项基准就超越了三个月前的V4-Pro预览版。后训练正在以肉眼可见的速度取代单纯堆参数，成为模型真正变强的核心杠杆。

[AB Kuai.Dong @_FORAB]

随着交易的完成，收购 AI 股神 Leopold 剩下股票持仓，捡到便宜筹码的，被曝是 Citadel 城堡对冲基金，它由华尔街老油条 Kenneth Griffin 率领。

戏剧性的是，城堡对冲基金曾发表，7 月美联储意外加息，给市场带来恐慌，间接加速 AI 股票下跌，迫使 Leopold 的 4 倍杠杆基金在低点平仓。

这也让 Citadel 城堡对冲基金，能低价买入 AI 股神 Leopold 基金的剩余持仓，此时它们的基金，已缩水至约 100 亿美元的水平。

2026 年最狠的资本市场故事，年轻的 AI 股神，被华尔街老油条狠狠干��。

[JimInvests @RealMcCoin]

Wow $AMZN just said they plan to spend $220 billion up from $200 billion on capex and it still won’t be enough.

Every single #neocloud MW will get leased.

Don’t get shaken out.

$DGXX $IREN $SHAZ $VIVO

[Alpha³ @WenfangHan]

We will be 10 very soon!! $DGXX

[ChinoAleman @chinoalemano]

$DGXX $NBIS $IREN

[Alpha³ @WenfangHan]

@YungAds_ @ROIRecruiter Are you back?🤪

[ChinoAleman @chinoalemano]

$DGXX could go to $36 and still be cheaper than WYFI and CORZ.

Or flip it to the super bear side: DGXX could report HALF its guided revenue, and even at a 5x multiple it would still be the cheapest name in the sector, at half of what $CIFR and $WULF trade at.

Don't forget:

➟ $0.3B market cap.
�� Zero debt.
➟ $100M capex paid.
➟ $155M cash.
➟ Cerebras deal $1.1B.
➟ Close to adding 1.3GW power plant.

The disaster scenario is still the bargain.

The work: $150M halved guidance at 5x is ~$750M EV, plus $155M cash, against a $325M market cap today.

The failure case, priced at half the peer multiples, roughly triples the stock.

When the super bear case is still the cheapest name on the table, the margin of safety is not in the guidance.

It is in the price.

Not financial advice. DYOR.

I'm long $DGXX.

[ChinoAleman @chinoalemano]

The CEO of $DGXX just posted two videos walking through the live NeoCloudz pods in Alabama.

Watch them twice.

Because this sector has trained you to expect renders, and these are RECEIPTS.

Those racks have been running since May.

Generating AI revenue.

Real customers, real GPUs, real invoices.

The first AI revenue print in company history lands in the Q2 report on August 14.

And here is the part most people still have not processed: this is NOT the Cerebras project.

Same Columbiana campus, two separate businesses:

➟ The Cerebras build: 40 MW of dedicated colocation, $1.1B contract, delivering December 15. Under construction.

➟ The NeoCloudz pods: DGXX's own bare metal GPU cloud, B200/B300 systems, live since May, scaling toward ~10 MW. Already printing.

While the market debates whether December arrives on time, it keeps forgetting the second engine is already running.

You can hear it humming in the CEO's own footage.

Most CEOs in this sector show you slide decks.

Amar walks you through the racks.

The revenue is not coming.

The revenue is here.

December just adds the zeros.

I'm long $DGXX.

Not financial advice. DYOR.

[Alpha³ @WenfangHan]

@michelamar3 Full Speed Ahead! $DGXX

[michel amar @michelamar3]

A quick walk through the floor. Rated-3, running 24/7 Nvidia B300.

[Alpha³ @WenfangHan]

@9andrey9 还能回前高吗哥

[Andrey Gorshkov @9andrey9]

吸筹的时候，分3-5个交易日可以平滑波动，避免遇到单天ICU行情

[Alpha³ @WenfangHan]

@wshngknshji1 @Pathusa 那你还吹 小心欠业债

[0x鸣人 @LuBtc888]

这张图是巴菲特这辈子最经典的笔投资。

今天深度解析这张图，里面藏着三条普通人必须遵守的投资铁律。

第一条，市盈率超过15倍的股票直接规避。

喜诗糖果12.5倍、可口可乐14.7倍、苹果13倍，巴菲特布局这些顶级企业时，市盈率均控制在15倍以内。

任何企业都会出现15倍PE以下的阶段，没有就耐心等待，机会靠等待，不是频繁交易。

当然，并非所有15倍以下的股票都值得买入。

第二条，银行、保险这类反身性极强的行业，想要介入，至少等到0.7倍PB以下。

这类行业存在索罗斯所说的反身性，下跌极易形成恶性循环：股价走低→资本承压→变卖资产→市场恐慌→业绩继续恶化。

地产、券商、持续融资的科技、光伏新能源都属于这类行业，追高风险极大。

巴菲特买入富国银行、盖可保险，都是在市场极度恐慌、股价接���净资产七折时布局。前提是能够看懂金融企业，看不懂切勿盲目抄底。

第三条，坚持逆向投资，反向大众情绪。

回看巴菲特经典交易：盖可保险危机、富国银行储贷风波、美国运通丑闻、华盛顿邮报遭遇负面事件，入场时机全部是市场一片悲观、大家都不看好企业的时候。

人性天然追涨杀跌，行情火热时众人抢筹，利空发酵时集体抛售。

市场无人看好的优质公司，才容易出现性价比极高的买点，市场泡沫大多是从众情绪催生。

最后我总结一下三条：

一、市盈率超过 15 倍 P 的股票一律不要买，牢牢记住这句话。

二、强反身性的金融周期股，不到 0.7 倍 PB 不要买。

三、永远逆向投资，不要跟着大众情绪盲目跟风。

[Terry @RichTerry123]

对于普通人来说大概就这六种：在股市里赚钱最好的方法。
第一种，是等大跌，买指数。
平时什么都不做，该上班上班，该生活生活。
只用闲钱，等到市场出现那种“股灾式”的大跌，再买入宽基ETF。
不用选股，一次行情赚个30%-40%是可能的。

第二种，是专注一两只股票，反复做波段。
找到一两只你认为真正值得投资的股票，永远保留现金。
涨的时候卖一点，跌的时候买一点，来回做，目标是把成本一路做低，甚至做成负数。

第三种，是攒股收股息。
买入那些股息率能到5%左右的国企、央企，在它们跌透时买进去，然后拿着吃分红。
就像有朋友买公路股，本来只想每年拿4%的利息，结果遇上行情，股价还翻了一倍。

第四种，是看懂周期，反向布局。
远离现在炒上天的板块，去找那些没人关���的、低估的板块。
比如2022年，大家都在追光伏、医药、白酒，有朋友那时布局跌了几年的银行股，现在市值已经翻了一倍。

第五种，是价值投资，等好公司打折。
说白了就是“数月亮不数星星”，只买龙头公司。
等它们因为市场情绪被打到极度便宜时，分批重仓买进去。你的目标不是赚几个点，而是等它价值回归，赚两倍、三倍甚至更多的钱。

第六种，是赚企业成长的钱
这是比较有难度的，
就是去投那些市值还小、但前景很优秀未来可能翻几倍甚至几十倍的公司
这需要你对行业有深刻的理解，
就像“数星星”，考验的是你超级的选股能力

这六条路，每一条都能走通，
关键不是知道，
而是选一条适合你自己的，
然后坚持下去