Bars4Dayz

Trade theses, setups, and post-trade reviews. Evidence over hype. Validate your conviction. Not financial advice.

[Bars4Dayz @HTSThesis]

so... i still have fable access?

[Bars4Dayz @HTSThesis]

@NickSung2017 @aleabitoreddit LOL no way there’s a website tracking him.. kinda goated on his part ngl

[Bars4Dayz @HTSThesis]

@retail_mourinho God candle loading

[coffee @coffeedev]

Crazy this was written two years ago, @leopoldasch truly was on point

[Peter H. Diamandis, MD @PeterDiamandis]

Pessimists sound smart. 

Optimists build the world.

[Bars4Dayz @HTSThesis]

@TheProfInvestor $AMSC $PENG

[KawzInvests @KawzInvests]

I don't care about the $SPCX IPO. I want SK Hynix on Nasdaq.

[Bars4Dayz @HTSThesis]

@sunxliao not yet.. i think we go lower towards october. Could be wrong but tbh the chart looks that way

[Bars4Dayz @HTSThesis]

@optionscjp @kevinxu lol hes actually a retard i swear, anybody that follows his plays and pays $200 for his sub is getting rinsed while the losses he incurs don't matter because he gets money from the subs regardless... classic scheme and he's just a fuckin loser lol

[Bars4Dayz @HTSThesis]

$AMSC is one of the more interesting “AI power” adjacent names I’ve looked at, but the thesis has to be framed correctly. This is not a clean cheap-stock setup, and it is not an AI pure-play. The better variant view is that AMSC is a now-profitable grid/defense equipment company with a small but rising data-center power vertical that could force a rerate if the order mix keeps shifting in the right direction. Looking at it as a longer term set up 🚨🚨🚨

AMSC sells into the grid-stability layer: power quality, voltage support, reactive power compensation, renewables integration, transformers through Comtrafo, Windtec systems, and Navy ship-protection systems. That matters because AI infrastructure does not stop at GPUs. The chain is chips → servers → cooling → data-center construction → power delivery → grid interconnection → voltage stability / power quality / transformers. AMSC lives in that last layer. It is not the most direct AI beneficiary, but it sits near a real bottleneck.

The operating trend is real. Q4 FY25 revenue was $86.4M. FY25 revenue was $299.2M, up 34% YoY, with 25% organic growth. Gross margin expanded to 30.5% for the year. Operating income swung to $11.4M from a prior-year operating loss. The company has now produced multiple consecutive quarters of GAAP profitability and ended the year with roughly $147.6M in cash/restricted cash and essentially no long-term debt. This is not a pre-revenue “story stock.” There is an actual business underneath the narrative.

The backlog/orders piece is the key. AMSC exited FY25 with 12-month backlog around $280M, up roughly 40% YoY. Q4 orders were near $100M. Total backlog was above $375M including longer-dated orders. Data-center orders were around 10% of Q4 orders, up from roughly 5% previously. That is the whole thesis in one number. If data centers stay around 10% of orders, AMSC is probably just a richly valued grid equipment company with AI headlines. If that number moves toward 20%+ and holds, the market may need to reclassify the business.

That is the potential rerate path. The market can currently look at AMSC and say: “small-cap grid/defense equipment company, thin operating margins, project-based revenue, premium valuation.” But if data-center power quality becomes a larger recurring order driver, the read changes to: “small-cap AI grid-infrastructure beneficiary with accelerating backlog, net cash, transformer exposure, Navy revenue, and operating leverage.” Same company. Different multiple.

The Comtrafo acquisition matters here too. It brought transformer exposure into the story at the exact time transformers are one of the most visible bottlenecks in grid buildout. Comtrafo also adds Brazil/FX/country risk and purchase-accounting complexity, so it is not a free win, but strategically it makes sense. AMSC is no longer just D-VAR / Windtec / Navy. It now has a broader grid equipment stack.

The Navy piece gives the thesis a second leg outside AI. Grid names can be cyclical and order-driven, but AMSC also has defense-adjacent ship-protection systems. Four of five San Antonio-class LPD ship-protection systems have been delivered, and the first Royal Canadian Navy delivery is expected in FY26 under the Irving Shipbuilding relationship. This does not make AMSC a defense prime, but it gives the business a different demand stream than pure data-center hype.

The upside case is not hard to understand: data-center order mix moves from ~10% toward 20%+, backlog continues compounding, Comtrafo cross-sell starts showing up in North America, operating margins move from the current low-single-digit level toward high-single-digits, and the market keeps valuing AMSC as a strategic grid-infrastructure beneficiary instead of a normal electrical equipment company. In that scenario, the stock can justify a premium multiple because earnings power would be changing faster than the headline income statement currently shows.

The downside case is also clear. This is not cheap. Around the recent $37–$40 area, the market cap is roughly $1.8B–$1.9B and EV is around $1.7B. On FY25 revenue of $299M, that is close to ~6x sales. Normalized EV/EBITDA is very high because underlying operating income is still only $11.4M. The headline GAAP net income number is misleading because FY25 net income was heavily boosted by a $118.4M non-cash deferred-tax valuation-allowance release. Strip that out and the company is profitable, but not nearly as profitable as the headline P/E suggests.

That is why I would not pitch this as “undervalued” in a traditional sense. The market is already paying for a lot. AMSC is not sitting at 1x sales with no one paying attention. The AI-grid narrative is known by analysts, price targets were raised post-earnings, and the stock still sold off after the print. That price action matters. When a company beats and the stock drops, it usually means expectations were already elevated or guidance/order quality was not enough to support the prior move.

Sentiment is where the setup gets more interesting. Retail attention does not look euphoric. The social read is neutral to cautiously bullish, not crowded. X has some low-quality promo noise, but not much strong organic thesis work. Reddit/StockTwits activity looks muted relative to the fundamentals. The tone is more washed-out post-earnings than euphoric breakout chasing. That matters because the stock may not yet have gone through a full retail “AI power” recognition cycle the way some other grid/generation names have.

So the tension is simple: fundamentals are improving faster than sentiment, but valuation is already demanding. That is usually where the best watchlist setups live. You do not chase because the multiple is rich. You also do not ignore because the business is objectively better than it was a year ago.

What would make me more bullish: data-center orders above 20% of quarterly orders for multiple quarters, backlog continuing to grow, underlying operating margin moving above 6–8%, evidence that Comtrafo is integrating cleanly, and management converting orders into cash without needing another equity raise. If those show up together, the rerate case gets much stronger.

What would break the thesis: data-center order mix stalling around 10%, organic revenue growth falling below the mid-teens, margins failing to expand, backlog quality weakening, another control/accounting issue from Comtrafo, or the stock continuing to trade at premium multiples while the operating leverage does not show up. In that case, AMSC becomes a good company at a bad price.

The insider-sales issue is worth noting but not overstating. CEO/CFO sales in June create some overhang, but much of the selling appears tied to 10b5-1/tax-withholding mechanics. That makes it a soft negative, not a smoking gun. The more important point is that there has been no clear open-market insider buying near the pullback.

My read: $AMSC is a legitimate AI-grid infrastructure watchlist name, not a blind chase. The business has real revenue, real backlog, net cash, defense/grid exposure, and early data-center traction. The problem is that the valuation already assumes a lot of execution. The opportunity is not “this is obviously cheap.” The opportunity is that the market may still be underestimating how large the data-center power-quality vertical can become inside the order book.

For me, the entire thesis comes down to one number: data-center order mix. Around 10%, it is optionality. Above 20% and sustained, it becomes a rerate argument. Below 10% or flat for multiple quarters, the AI-power thesis is probably ahead of the business.

[Bars4Dayz @HTSThesis]

$DRAM $LEU $PENG $ONDS $BWXT $AAOI

[Bars4Dayz @HTSThesis]

$DRAM $LEU $PENG $ONDS $BWXT $AAOI

[Bars4Dayz @HTSThesis]

Two besties, public feud, secret to plan manipulate markets before absolutely pumpin em for the IPO? Fuck it I’ll take it

[The Hormuz Letter @HormuzLetter]

BREAKING: Iran has now placed all of Elon Musk's companies, interests, and assets in its target banks ahead of Trump's announced US attacks tonight, including Starlink ground stations across Qatar, Jordan, the UAE, and Oman, along with SpaceX shareholders, citing US and Israeli military use of Musk's Starlink and X platforms, per Fars.

Iran's Khatam al-Anbiya Central HQ also warns the response to the US attack tonight will be different than usual, threatening to strike Gulf energy infrastructure with "either oil and gas exports are for everyone or they will be available for no one," and warning the war will go "beyond the Middle East."

[Bars4Dayz @HTSThesis]

Exactly. The “intelligence layer” is where the business value shows up, but my second-order angle is the infrastructure constraint underneath that layer.

For SMEs or enterprises to turn raw data into decisions quickly, the system still needs to retrieve, hold, and process context efficiently. So that's where inference becomes memory-heavy. More context, more agents, more retrieval, more concurrent users = more KV-cache / memory pressure.

That is the $PENG angle for me.

Not “copycat GPU infrastructure,” and not “next Nvidia.” More like: if enterprise AI moves from pilots to real workflows, the bottleneck shifts toward memory expansion, CXL, KV-cache infrastructure, and AI/HPC deployment for companies that cannot build like hyperscalers.

Penguin’s Integrated Memory segment grew ~63% YoY in Q2 FY26 and is now the largest part of the business. Their MemoryAI KV Cache Server is directly aimed at that inference memory wall.

The risk is that the stock already re-rated hard while total company revenue has not fully broken out yet, so I’m not calling it cheap. But the second-order angle is real: the intelligence layer needs infrastructure that can actually feed it fast enough.

[Bars4Dayz @HTSThesis]

$PENG + AI memory bottleneck thesis

I’ve been digging into Penguin Solutions ($PENG), and this is one of the more interesting second-order AI infrastructure names on the market. The thesis is not “next Nvidia.” That is lazy. The cleaner thesis is that Penguin sits near a real AI infrastructure constraint: inference memory.

As AI workloads move from training-heavy to inference-heavy, the bottleneck shifts. GPUs still matter, but inference at scale also stresses memory capacity, bandwidth, latency, KV-cache management, CXL expansion, and cluster utilization. If memory cannot keep up, expensive GPU clusters become less efficient. That is where $PENG becomes relevant.

Penguin has three main business lines: Advanced Computing, Integrated Memory, and Optimized LED. Advanced Computing designs, deploys, and manages AI/HPC clusters for enterprises, governments, neoclouds, and customers that cannot build like hyperscalers. Integrated Memory includes SMART Modular / SMARTsemi, DRAM modules, flash/SSD, specialty/rugged memory, CXL memory expansion, and the MemoryAI KV Cache Server. Optimized LED is the Cree LED business, which is mature, cyclical, and not the core of the AI thesis.

The real story is Integrated Memory. In Q2 FY26, Integrated Memory revenue was $171.6M, up from $105.3M in Q2 FY25, or roughly +63% YoY. That segment was about 50% of total Q2 revenue and has become the largest and fastest-growing part of the company. This is what the market is paying for.

But the important nuance: total company revenue is not yet breaking out. Q2 FY26 net sales were $343.0M, down from $365.5M in Q2 FY25. Over the last several quarters, total revenue has mostly stayed inside a ~$324M to ~$366M range. So the stock move is not being driven by broad absolute revenue acceleration yet. It is being driven by mix shift, guidance revisions, AI-memory narrative, and multiple expansion.

That distinction matters. $PENG is not a simple “revenue inflecting everywhere” story. It is a company where the market is re-rating the whole business because the Integrated Memory segment is becoming more strategically important.

+The bull case is that this mix shift is exactly the point. Penguin is moving away from lower-quality / declining pieces and toward higher-value AI memory infrastructure. The company’s MemoryAI KV Cache Server is a CXL-based product designed for enterprise inference workloads and marketed around up to 11TB of disaggregated memory. That directly targets the AI inference memory wall, especially for long-context workloads, RAG, agentic AI, financial/legal document processing, and high-concurrency enterprise inference.
+The second bull angle is AI factory deployment. Penguin is not just selling memory modules. Through Advanced Computing, it also helps design, deploy, manage, and optimize AI/HPC infrastructure. That matters because the next wave of AI infrastructure is not only the hyperscalers. Enterprises, sovereign AI buyers, financial institutions, universities, labs, and specialized cloud providers all need AI infrastructure, but many cannot build like Microsoft, Amazon, Google, or Meta.

This is where Penguin can sit: not as the GPU owner, not as the hyperscaler, but as the infrastructure integrator, memory specialist, and AI deployment layer.

The ecosystem validation is real but needs to be stated accurately. Penguin has Dell ecosystem relevance and was named Dell Technologies Global Alliances Americas AI Partner of the Year. SK Telecom is also strategically involved through a $200M convertible preferred investment. The Celestial AI / Marvell angle is more nuanced. Penguin was an early investor in Celestial AI, which Marvell acquired. Penguin realized a gain from that stake, and the photonic memory appliance development relationship appears to continue through the inherited Celestial AI relationship now under Marvell. But I would not call this a clean formal “Marvell JV” unless the company itself uses that language. The thesis is already strong enough without overstating it.

Now the risk side.

First, valuation. $PENG has already had a major move and recently traded near an all-time high around $73. At current levels around the low-$60s, the stock is no longer undiscovered. The report I reviewed had PENG at roughly 27x FY26 non-GAAP EPS, around 2.2x forward sales, roughly 25x P/FCF, and meaningfully above where SGH/PENG historically traded as a memory-module company. That does not make it uninvestable, but it means the market is already pricing in execution.

+Second, the company still has hardware-cycle exposure. Integrated Memory growth is partly tied to favorable memory pricing. That helps in an upcycle, but it can reverse. If memory pricing rolls over, revenue growth and margin can both get hit. This is not a pure recurring software story.

+Third, margin pressure needs to be watched. The company raised revenue and EPS guidance, but gross margin guidance moved lower. That tells me investors need to focus on the quality of growth, not just the headline revenue number. If Integrated Memory grows but gross margin compresses, the market may start asking how much of this is structural AI demand versus cyclical memory pricing.

+Fourth, Advanced Computing is still weak. Q2 FY26 Advanced Computing revenue was $115.7M, down from $200.2M in Q2 FY25. The bull explanation is that Penguin is moving away from lower-quality work and shifting toward AI infrastructure and memory. That may be true, but until Advanced Computing stabilizes, Integrated Memory has to carry the whole narrative.

+Fifth, governance is a yellow flag. CEO transition earlier in 2026. CFO transition announced June 1. The company stated the CFO departure was not due to disagreement over accounting, controls, financial reporting, or operations. That matters. But two C-suite changes in a short window, while the stock is near highs, still deserves attention. There was also insider selling into the run-up from directors/officers. Insider selling is not automatically bearish, but the timing is not something I would ignore.

+Sixth, dilution. The balance sheet does not look distressed, and the company retired near-term 2026 converts. But SK Telecom’s $200M convertible preferred and the remaining in-the-money convertible notes create potential share-count overhang. Buybacks help offset this, but investors need to watch diluted share count, not just headline EPS.

So the setup is not simple. The business thesis is real. The stock setup is less clean.

The bull case:

AI inference creates a real memory wall
CXL memory expansion becomes more important
KV-cache infrastructure becomes a larger enterprise problem
Integrated Memory is now the largest and fastest-growing segment
Penguin has AI/HPC infrastructure deployment capability
Dell/SKT ecosystem ties add credibility
MemoryAI and photonic memory appliance optionality create upside
Balance sheet is not distressed
Buybacks are offsetting some dilution

The bear case:

Total company revenue has not decisively broken out
Advanced Computing declined sharply YoY
Integrated Memory may be partly riding memory pricing
Gross margin guidance moved lower
Stock already re-rated aggressively
Governance churn is a yellow flag
Insider selling into strength is not ideal
Convertible/preferred dilution needs monitoring
Retail attention is no longer early

The key proof points I want from here:

Total revenue needs to break above the old ~$365M ceiling for more than one quarter.
Integrated Memory growth needs to stay strong without gross margin collapsing.
Management needs to show the growth is product/volume-driven, not just DRAM pricing.
Advanced Computing needs to stabilize and return to YoY growth.
The company needs to name a credible permanent CFO.
Diluted share count needs to stay controlled.
Penguin needs to disclose harder metrics around MemoryAI, CXL, PMA, ClusterWare, AI/HPC bookings, or customer adoption.
Fiscal Q3 FY26 needs to confirm the high-end FY26 guide rather than expose the stock as priced for perfection.

My view: $PENG is a real AI infrastructure bottleneck company, but it is not obviously cheap at current levels. The cleanest thesis is not that Penguin is “the next Nvidia.” The cleanest thesis is that Penguin is a second-order AI memory/inference infrastructure play that the market has started to re-rate before the full financial proof has arrived.

That makes it watchlist-worthy for me

One-line thesis: $PENG is a real AI inference memory-bottleneck play scaling CXL, KV-cache, memory integration, and AI/HPC infrastructure deployment.

One-line risk: $PENG is still a hardware-cycle exposed company that has already re-rated hard while total revenue has not yet decisively broken out.

For me, this is a “real thesis, price matters” name. I want proof before chasing

[Bars4Dayz @HTSThesis]

The $AMPG bull case is interesting, but the bear case has to be said too, it's not automatically a clean “hidden 5G winner” just because the customer slide looks impressive.

If that growth was just a backlog release, if margins do not recover, or if AMPG keeps needing equity financing to fund the story, the stock can still punish holders even if the technology is real.

Bull case: AMPG is becoming a niche RF / 5G / satellite infrastructure supplier.
Bear case: AMPG is still a volatile microcap with weak margins, dilution risk, and an unproven jump from components into larger 5G systems.

[Bars4Dayz @HTSThesis]

$AMPG isn’t a clean “cheap 5G stock” thesis.

It is a high-risk microcap RF-components / 5G-ORAN execution story sitting at the intersection of low-noise amplification, satellite communications, telecom infrastructure, defense, and open-radio access networks.

The bull case is there, the problem is that the bear case is also present 🚨🚨

AmpliTech reported Q3 2025 revenue of $6.09M, up roughly 115% YoY from $2.83M. For the first nine months of 2025, revenue reached $20.72M, up from $7.66M in the prior-year period.

That is not a dead microcap story anymore. It is a company showing real top-line recovery after a brutal FY2024.

FY2024 was ugly. Sales fell from $15.58M in 2023 to $9.51M in 2024, down roughly 39%. Gross margin fell from 46.69% to 36.65%. The company reported a net loss of $11.24M and used $5.30M of cash in operating activities.

So the setup is not “AMPG is cheap.” The setup is “AMPG may be inflecting, but it still has to prove the revenue recovery is durable.”

The part I think people may be underweighting is the product/catalyst stack.

AmpliTech’s growth in Q3 2025 was driven by higher sales in its low-noise amplifier, low-noise block, and 5G product lines. Management specifically tied the increase to stronger demand from telecommunications and satellite communications customers.

That matters because AMPG is not just attached to the word 5G in a generic way. Its core products sit inside the RF signal chain. Low-noise amplifiers are used to improve weak signals before they get degraded by system noise. In satellite, telecom, defense, and high-frequency systems, that can be a real technical requirement.

The newer piece is the 5G/ORAN side.

AmpliTech has been developing a massive MIMO 64T64R ORAN Cat B radio network product. The company said this product was being tested and certified at Northeastern University and expected it to become a flagship product.

That is the upside case: AMPG moves from being a niche RF-component supplier to a more meaningful 5G/ORAN and satcom infrastructure supplier.

But this is where the thesis has to stay honest.

The margin profile is not clean yet. Gross margin compressed hard in FY2024. Operating loss widened to about $8.43M. The company is still not a clean earnings story.

The dilution/capital-raise risk is also real. AMPG generated $21.18M from financing activities in FY2024, mainly from registered direct offerings and debt financing. That helped the company end 2024 with $19.32M in cash and $26.80M in working capital, but it also means the equity story depends on whether revenue can scale faster than the cap table expands.

There was also a $3.25M impairment tied to fraudulent digital-asset transactions in FY2024. That is not part of the operating thesis, but it is a governance/control issue that should not be ignored.

My actual thesis:

$AMPG is not a valuation-discount setup. It is a revenue-inflection and execution setup.

If AMPG keeps converting demand in LNAs, LNBs, satellite communications, and 5G/ORAN products into real GAAP revenue, while recovering gross margin and limiting future dilution, the market may start treating it as more than a forgotten RF microcap.

If the Q3 2025 growth was just a temporary backlog release, margins stay weak, or the company keeps relying on equity raises to fund the story, the stock can compress hard.

What I am watching:

Revenue durability beyond Q3 2025
Gross margin recovery back toward prior levels
Cash burn
Share count / future offerings
5G/ORAN product certification
Customer traction in telecom and satellite communications
Whether the massive MIMO ORAN radio becomes real revenue, not just a product narrative
Any additional governance/control issues
The setup is simple:
Bull case: AMPG becomes a niche RF / 5G / satellite communications supplier with a real revenue inflection, net-cash support, and a product path into ORAN infrastructure.
Bear case: AMPG is a small RF manufacturer with volatile demand, weak margins, recurring dilution risk, and an unproven jump from components into 5G systems.

[Bars4Dayz @HTSThesis]

$ONDS isn't a clean “cheap drone stock” thesis.

It is a high-risk counter-UAS roll-up sitting at the intersection of drones, defense, World Cup security, U.S. prime-contract access, and AI-enabled command/control.

The bull case is real. The problem is that the bear case is also real.

Ondas reported Q1 2026 revenue of $50.1M, up more than 10x YoY, and raised FY2026 revenue guidance to at least $390M. Pro forma backlog reached $457M, up from $68.3M at the end of 2025. The company also ended Q1 with roughly $1.48B in cash, restricted cash, and short-term investments.

That is not a dead microcap story anymore. It is a transformed balance sheet with a very aggressive defense/autonomy acquisition strategy.

The part I think people may be underweighting is the catalyst stack.

Sentrycs was selected to deploy counter-drone protection across multiple 2026 FIFA World Cup venues in North America. This matters because the World Cup is not just another event. It is a live, high-visibility test case for civilian counter-UAS deployment across stadiums, fan zones, and dense urban environments.

Mistral gives Ondas direct U.S. defense prime contractor access. The merger added around $264M of backlog and gives Ondas access to U.S. Army and Special Operations contract vehicles. This is the bridge from “drone tech roll-up” to potential U.S. government program participation.

LADOS is the newer piece. Ondas is launching it as a layered autonomous defense C2 system designed to connect sensors, drones, robotic platforms, counter-UAS assets, ISR, and effectors into one operational layer. This is the company trying to move from hardware portfolio to “system-of-systems” platform.

That is the upside case: ONDS becomes less of a scattered drone roll-up and more of an integrated autonomous defense platform.

But this is where the thesis has to stay honest.

The dilution is brutal. Shares went from roughly 93M at the end of 2024 to roughly 496M by May 2026. There is also a major warrant overhang, including warrants from the October 2025 and January 2026 financings. So the question is not just “can revenue grow?” The question is “can revenue grow faster than dilution?”

The GAAP profit also needs to be ignored. Q1 2026 net income was driven by non-cash warrant-liability accounting and deconsolidation gains, while the actual operating loss was still about $42.7M. This is not yet a clean earnings story.

My actual thesis:

$ONDS is not a valuation-discount setup. It is a catalyst-and-execution setup.

If ONDS converts the $457M backlog, delivers the $390M+ FY2026 revenue guide, holds gross margin near/above 50%, and stops expanding the share count, the market may start treating it as a legitimate autonomous-defense platform instead of a speculative roll-up.

If backlog conversion slips, warrants keep weighing on the cap table, or the company keeps buying revenue with stock, the equity can compress hard.

What I am watching:

Backlog conversion into real GAAP revenue
Organic vs acquisition-driven growth
Share count staying under control
Gross margin durability
World Cup counter-UAS validation
Mistral contract flow
LADOS/SkyWeaver integration becoming more than a slide-deck narrative

The setup is simple:

Bull case: ONDS becomes an integrated counter-UAS / autonomous-defense platform with World Cup validation, U.S. prime access, and a $390M+ revenue run-rate.

Bear case: ONDS is a serially diluting defense roll-up where revenue was bought, profits are accounting noise, and valuation already prices in flawless execution.