Phreak's Thoughts, Ideas and Opinions

[pahaaadi]

My two cents

1. in last 2-3 years inference cost came down drastically primarily due to software/model improvements. newer hardware is also making per unit token generation cheaper. So net effect is still a better unit economics

2. I personally agree with you on the jobs for most part but the reasons smarter ones say it won’t because they subscribe to this way of thinking jobs. Say an AI agent allows one engineer to do work of five engineers. Now equilibrium isn’t necessarily that 80% of engineers will be unemployed. They believe the boundary of what’s worth building will expands exponentially. The friction we see is a due to mismatch of velocity. The older jobs can be replaced faster than the new ones that are yet to emerge. Just think the new jobs that we see today and how many existed 20 years back.

I think the truth is somewhere in the middle for jobs.

[phreakv6]

Token costs are becoming cheap despite hardware becoming expensive because of improvements in model architectures and also deployment architectures. For eg. gpt-4 was $60/million tokens but today we have much better intelligence available for < $10/million output tokens. Deepseek made some really good improvements with MoE architectures that totally changed things.

I think I can explain this a bit better though its perhaps very technical. I will use examples of local AI since that’s where most of my experience comes from. When a model has say 30 billion parameters, it means 30 billion floating point numbers (weights) and their connections with abstractions like layers. For generating each token (which you can think of as a word), the whole 30 billion params have to be read from memory. This essentially makes the memory bandwidth (how fast can you read 30 billion params which will be maybe 20GB or 40GB roughly depending on the precision - 4-bit vs 8-bit lets say) the bottleneck. So memory bandwidth/model size roughly gives the output tokens/sec (5090 is 1.8 TB/s, 4090 is 1.2 TB/s and 3090 is 900 GB/s which is why you can get 40-100 tokens/sec on one of these 20GB size models in these cards)

So several optimisations came through - one of the earliest being quantisation - which is reducing precision of each of these weights - from 16-bit to 8-bit or 4-bit which cuts size down proportionately with some sacrifice to model quality (16-bit to 4-bit reduces quality maybe 10-15% and not 75% being the key insight). Then MoE architectures dont read all 30 billion params to generate a token - they may have just 3b (3 billion) active params at any point of time - which 3b are active will vary for each token (based on relevance) - this totally speeds up token generations multi-fold on same hardware. Then there is speculative decoding, dflash, mtp (multi-token prediction) which are newer techniques all of which give better performance without much loss of quality.

Another big improvement is batching - essentially since bandwidth is a bottleneck but weights you read are essentially same - how about processing multiple prompts simultaneously, so you read weights once but generate tokens for multiple prompts at once. Continuous batching is one of the key techniques used by data centers to improve performance. This is a typical operating leverage play because serving 20-25 prompts on the GPU at once gives you way better overall perf than serving a small number (local AI mostly does 1 or at most 2 or 3 because of other bottlenecks like memory which is used for kv-cache).

Lot of improvements have been made in deployment architectures too where a single large trillion param model is sharded (split) across a cluster of GPUs, so for generating a token for your prompt, multiple GPUs work together. There are multiple diff types of such parallelism that allows a cluster to serve several prompts at one go.

This is how intelligence is becoming cheap every few months and I think the trend will continue - because our brain is able to achieve so much in just 20W (you are processing so many senses, thinking so many things, while talking and driving a car for eg. even on a casual idle brain).

Does this mean we will not need as many DCs is the natural next question - I think demand is exploding at a higher rate than supply at this point and is likely to continue so for the foreseeable future, esp if intelligence becomes cheap. Similar to how number of devices connected to the internet through WiFi has gone up so much in the last 10 yrs, more and more devices will become intelligent. If I have to draw a parallel to the internet era - we started with dial-ups and then moved to broadband to fiber (speeds have moved up 5-6 orders of magnitude and usage has exploded in the period). We are probably in early 2000s internet at this point (before streaming audio, youtube, e-commerce). We are likely seeing the equivalent of early Google in Anthropic/OpenAI. There is a long, long way to go with a lot of ups and downs.

[Debojit_Kangsa_Banik]

My takeways:

1. Demand/adoption is very very high and so hardware costs are rising even though models are getting better.
2. In the long term, there is no question that AI is here to stay. Companies are adopting even though costs are high now coz they understand this same fact that, slowly token cost will go down. I don’t need the most efficient model for my work, even gpt-4o-mini can do some basic day to day work, the price of which, is already low and will further cool down
3. With more adoption, power theme, memory theme will continue. But eventually I do think more players will emerge especially in memory segment. High margin biz will definitely attract more players into market.
4. We are in volatile/adoption phase when it comes to job market. New jobs will definitely emerge but till then we will likely see shuffling, that may lead to job losses. Only question is how long will it continue till new type of job roles start capturing the new market.

[Aloysius_De_Sa]

On your second point, I feel the widespread roadmap is for software companies to again make a shift from SaaS per user pricing to outcome based pricing. This is not an answer for unemployment as that is another issue altogether where the World has still not yet grasped the effect, but SaaS companies that can migrate to outcome based pricing effectively can thrive as they get rewarded for replacing engineers altogether by doing the jobs themselves.

[sougataG]

I found this take by Palantir CEO Alex Karp very interesting and contrary to popular opinion. I noticed this shift in narrative over the last month or so. If the shift happens even slightly, there might be significant work for the IT services industry.

Palantir (@PalantirTech) on X

[Tom_m]

This trend is not very new. The buzz from IT industry (I used to work there) is that many customers are trying to build native AI capabilities for some months or even years now.

I tried a bit more searching to quantify “many customers”, and it surprised me that consensus seems 50% or even more large to medium IT customers are either already trying it or considering it. Availability of cheaper open source engines and API, IP Ownership, privacy and security, long term thinking etc are driving this trend.

This may also explain recent GCC expansions in India. 64% of all new GCC roles created require AI, data science, or intelligent automation skills.

In Indian stock market context this may mean (it’s very hard to generalise but fools rush in etc.)

1> For IT services it’s a double edged sword. New contracts are being won, but traditional service contracts and manpower based billings are under severe pressure.
2> Datacenters are likely to get more leasing contracts from native AI adopters
3> Companies that provide products and services to Powergrid and datacenters should continue to thrive.

[phreakv6]

This is going to be a rant. Rants are generally useless, so maybe better avoided but I have few thoughts on markets and participants and some trends i’ve seen of late. I dont expect this post to be a big value add. I am writing to clear my head.

The entire market at present seems to be relying too much on technicals. I use technicals to instruct my buy/sell decisions too but never without support from fundamentals. The market currently though only wants stage-2 stocks and even a mild retracement to 50-dma or 20-wma (can be 20% drawdown) is seen as the commencement of stage-4 or whatever. All the gung-ho fundamentals that make their way out when stock is in stage-2 are nowhere to be found all of a sudden.

That brings me to the second gripe - if the objective is to ride simply based on technicals, why even bother to look at fundamentals? All the narratives are forgotten or worse - changed to suit the price action. I believe this is a result of too much reliance on charts and eyes that have trained too much solely on charts. Technofunda approach works great if followed with equal importance to both. Using funda just during stage-2 and ignoring it in a pullback/consolidation is recipe for disaster. I can tell people are losing a lot of money in these fake-outs (as it always happens in a sideways market).

The frequency of narratives roughly match stock hitting all-time highs. If you genuinely want to make alpha in this market, look to buy pullbacks to 50-dma in strong trending stocks or 20-WMA which usually gives a nice 20% discount. If the only thing finfluencers are going to do is talk about stocks that are big movers for the day, it is doing serious disservice to the community. The focus should be on stocks that appear to be turning bearish but that which had “strong fundamentals” 20% higher a month back. Usually nothing changes in fundamentals in a month in most businesses.

When the entire market becomes a trend-following market, things become zero-sum very quickly. No one makes money on average. Everyone’s objective is to sell higher which means they are buying to sell to a greater fool. Whenever my objective is this (that someone would buy it off me higher), there’s a good chance the greater fool is me. Its worth considering.

What makes this worse is the cottage industry on X that uses AI to come up with 5 stocks ideas every week. There’s no follow-up or deep work. Only a quote tweet every time said stock is up 5-10% for the day. This is what enhances all this bad behavior.

One of the things I have changed in the last 3-4 yrs is this. Once you have mastered the art of not losing (knowing what to avoid), it is important to transition towards playing to win (From “The biggest bluff” book)

Taking profits gives happiness in short-term, esp. when they are hard to come by. This is what keeps a sideways market sideways. But even in a sideways market, there are generally lot of stocks that are strongly trending with strong pullbacks generally backed by strong fundamentals. If I take every stock that was a 4x for me in the last 3 yrs, be it Apar, Shilchar, VBL, Wocky, Ceinsys, Shaily - all of them had 5 or 6 pullbacks 20%+, few even had 30/40% pullbacks though they all were 4x in ~1 yr. These are enough to provide enough and more alpha to the portfolio. The old me would have taken a 20% profit in these.

Once you have learnt not to make stupid bets, its important to hold on to good bets and ride them through. One trick I use is to only use monthly charts - all the volatility and 20% drawdowns look routine in those - the noise is tremendously reduced. Whatever happened to cutting the flowers and watering the weeds quote? In a sideways market, weeds are aplenty. In general, its better to buy few businesses many times than buy/sell many businesses once.

[Somnath_Patra]

Oh! It’s a nice post. Have learnt a lot. Learnt how to buy, When to buy.. It will be more encouraging if you describe when to sell too.. Thank you for a lovely post.

[Cshar]

This is real wisdom coming directly from heart @phreakv6 . Very nicely explained. Its a tough market to navigate. One of the best post.

[V_Nagender_Rao_Gunda]

@phreakv6 Best post and very well explained Thankyou

[Sid_Mathew]

Well articulated as usual. I have a bias towards fundamentals but there are traders/short term investors that have built their entire model around technicals and have been quite successful doing so. If their goal is to simply make 25%+ on every trade, then even after paying short term capital gains that’s a fairly good return. In a way it takes a certain kind of discipline to keep fomo aside and keep churning and sticking to that approach as well, knowing that you’re leaving money on the table when the stock eventually moves higher. This sort of behavior will always be there as long as there are scalpers, traders etc.

[Tom_m]

As a professional trader primarily using Index derivatives, I (and many like me) use short to very short term charts based strategies that benefit from periodic swings. As part-time investor, however, I prefer long term bets and like to avoid reactionary entry/exit unless fundamentals change, similar to what this erudite post mentions.

At the same time, I personally know people who invest on short terms and aware of some of the strategies they use. Those methods can be fairly successful, even and especially in sideways market like last 1-2 years. Such strategies are mainly for investors who try to generate regular source of income. This point needs to be understood before deciding on the merit or demerit of this approach.

Which approach is more useful in the long run? In investment it’s very hard to generate proper backtested quantitative data due to broad qualitative definition of strategies and very large number of stocks. As I understand, both have enough merits, and use, to be taken seriously.

[keerthan_bs]

Most of the so called fundamental research is just AI slop. 1 in 20 do actual in depth fundamental work. Natural volatility and trump induced mania makes it easier to catch those 10-15% moves as mispricing becomes more frequent. The technical leaning of market is more to correct these inefficiencies

I have moved to 90% technicals (10,20,50 dma) as I can follow that better than most. Feedback loops are faster and you can improve or refine much quicker.
The flow of returns trickles vs gushes is something everyone has to choose for themselves.

By the time twitter posts you mentioned come up, it’s time to jump ship. I find this more palatable than sticking to one and waiting endlessly only to be proved right/wrong much later.
I believe the large greater fool theory comes to play in the ipo’s and storied themes pricing multiple years of near perfect performance.

I have chosen to avoid mentioning the pump which happens in 2-5% circuit limit stocks, quite insane!

[phreakv6]

Thank you all.

In my view selling is not separate from buying. Its not something done in isolation. While buying you should have a good idea of why you are buying and what will make you sell it. It can be a valuation bridging or a tailwind coming to an end or govt policies changing adversely (which will seriously impact thesis - not all of them do) or serious balance sheet issues or corp governance issues or repeated promises on growth or margins not kept etc.. It most definitely cannot be “someone will buy it from me higher” - I have only lost money on average when my thesis is this. Following this strictly also filters out most of the prospective trades. I cannot insist this enough. Do not take other people for fools. I have analysed my trades over the years and the ones I have lost money on are precisely these - because in all likelihood everyone has the same notion as you and it usually doesn’t end up well. Again, on average - few will do well and make you continue this abomination because its easy to spot such bets and they occupy a lot of mind space since people only talk about such bets on social media in a cacophony of the blind leading the blind.

If you think people will be interested in it higher - ask yourself what will take it higher and ask yourself honestly if that reason is fundamental and if your view is different from the markets and why the market can come around to your view. Any stock moving up purely on technicals without a fundamental backing invariably retraces the whole journey in a sideways/bear market - even if the “chart looks nice”.

Most of my best selling has happened because I have found a better bet and its easier to justify a switch than a sell. This is why a lot of times what I have sold has mostly gone on to double or do well even after a strong run (because the fundamentals are still intact). I still call this best selling because on average I have done well despite this (Its easy to delude yourself if this is not backed by data. Build a portfolio time machine and analyse if your decisions are generating alpha - not single ones but entire set of decisions in a time period. Makes you very grounded and honest, at least to yourself). My objective is never to sell at the top or outthink and outdo my fellow shareholders (again, zero sum game). Without good intentions, its hard to make money consistently in the long run.

Its also good to understand the concept of equilibrium. Any system thats disturbed eventually returns to an equilibrium state. It need not be the same state as it was before the disturbance. I like looking at what an equilibrium state can be when analysing a sector. This makes me not sweat the immediate details. For eg. its easy to guess that data centers are needed and there are going to be several in our 5-10 year future than sweat the details on which models are going to be used or how its funded or how the economy copes with its impact on the environment or society. All these are hard questions with no definite answer and timeline. The easiest one is that there are going to be numerous data centers in the future and no one can refute that - even the most bearish on AI would struggle to come with an argument against this. In Jan, I felt equilibrium for IT stocks was a 12 p/e future, similar to Coal India. Most stocks were trading 20 p/e+ at that time. I felt it might take time to get there though but its happening rather quickly now. Even if there is a bounce, we should understand that these are now in a different equilibrium as compared to the past.

Same way its easier to guess that there is going to be a lot of power required to power these data centers and EVs. Its trivial to understand that generation and consumption happens geographically separately and also intuitive to understand that along with spatial separation, temporal separation is inevitable (power generated in the past, consumed in the future - using BESS). For spatial transmission of so much power which is generated in remote areas the grid must be strengthened (you can put a thermal plant close to areas of consumption but can’t dictate where the sun shines and land is cheap or where the wind blows). (Got this table from X. Credit to the person who created it. I verified and it is based on NEP data)

It is easy to see an equilibrium where we have that 15000 ckm and 32 GW of HVDC transmission lines being added in the medium term. There may be delays in tendering, land acquisition, execution etc - but very few companies can capture that value whether it happens in 5 years or 7 years. I would like not to get stuck on the details (though I go to absurd length to track details, I seldom act on it). So understanding equilibrium states for the stocks and sectors you own can be a comforting thing that helps hold on to the conviction.

I know concentrated vs diversified is a never ending debate. I am a very concentrated investor because I have limits on how many stocks I can really truly understand and appreciate the nuts and bolts of. My general gut feel is that people who are diversified are unable to sort the stocks by conviction and return potential. This is not an easy thing to do and I read a post by some veteran here on VP (most likely @Donald) back in 2018 that mentions this. I was stumped how anyone can do this. Over a long period of time, slowly but steadily I have accumulated this skill with experience. I think no one should force themselves to be a concentrated investor until they develop this skill. You will know when you have it because you cannot sleep at night knowing that your allocation even at 15% in a stock isn’t sufficient and you must scale it (for me it happened with Shilchar and Apar). Stupidly scaling up without having the skill is a recipe for disaster. I mention this because selling needn’t be a complete exit due to fundamental or technical reasons - if you are extremely concentrated, its normal to reduce a bit to bet on other ideas and derisk the position (Another important skill while doing this is to not get bearish after selling a bit and end up selling everything - your brain can play lot of tricks as its a justifying machine). Its not just skill but temperament as well and that varies based on portfolio size. Once the portfolio size has passed the mental threshold for financial stability, it is a lot easier to try out things.

The way to read these posts is not to gain any preset “rules”. Investing is very much like jazz. Jazz works based on a structure but within that structure, players are free to experiment. There’s an underlying key and chord progression but you can play it many ways each time. Sometimes deliberately playing outside the key/time signature but returning back or even changing the key/time totally are acceptable but there is still a structure held by something which is unchanging - either the key, melody, chord progression or time signature/tempo/feel that holds it together. Because the market reacts to any one strategy being abused by punishing it with mediocre returns, it is important to be roughly aware of it. A few years back, everyone was into Buffett and long-term investing and in that paradigm a portfolio that churned constantly made great returns. Now I feel we have gone a bit overboard on this concept, so much so that pretty much no one even thinks of holding anything for a year even. A strategy that plays like counterpoint in jazz does very well in general (plays within that structure with disciplined dissonance).

[Nikhil_Palliyil]

Hey @phreakv6 great to learn and adapt few learning’s from your posts. Just wanted to know your views you seem to be very much optimistic earlier on Apar and now on all of the HVDC value chain eager to know your views on the winding wires players like KSH international which seems to be the largest export player of winding wires and the only approved company to supply CTC for hvdc transformers are you worried about the working capital intensive model of these companies in a copper deficit country??

[Yash_Vyas]

@phreakv6
Hi Bharani,

Thank you for another insightful post.

You framed your best selling as switching into a better bet. I would like your view on the timing gap inside such a switch: specifically, using short-term leverage as a bridge.

The scenario: a holding is marked for exit (for one of the reasons you listed), but it trades well below what my judgement says it’s worth, selling today funds the switch at a poor price. Meanwhile, the target stock is available at an attractive valuation right now. So instead of selling first, use short-term leverage (leverage capped wrt net worth) to buy the target immediately and unwind it when the exit stock hits a preset sell price. The intent is to avoid sacrificing one side of the mispricing for the other.

I can see the risks: the exit stock may never reach my preset price, interest cost eats into the spread and in a drawdown both legs go against me simultaneously.

Since you test decisions against data, Has your trade data ever suggested such a bridge adds alpha over simply selling first? Or does the tail risk kill the idea regardless of the numbers?

[Pawan_Ghosliya]

Hi @phreakv6 ,

Thank you so much for such detailed insights.

Can you throw more light on skills that are essential for concentrated investors. I believe that I am slowly becoming concentrated investor however I am not sure that I am doing enough to manage the incremental risks which come along by being more concentrated.

[phreakv6]

@Nikhil_Palliyil - No views on KSH. I did look at it but in general try to avoid businesses that function on a EBITDA/ton basis. Being the only approved supplier for HVDC transformers though does put it in a unique position. I think with the expanded capacity, they will continue to grow volumes but in general these businesses function with a EBITDA/ton band and there’s generally no big operating leverage. However, consider my knowledge on this company too basic to have a worthy opinion. I have not spent much time on this.

@Yash_Vyas - I don’t like taking leverage. I don’t find the need to optimise so much.

@Pawan_Ghosliya - I think you need to develop a sort of a leaderboard with the stocks you have done work on. Few of the columns there have to be possible returns and a confidence level and downside risk. When downside risk is low with higher possible returns with a great confidence level, it needs to move up and you need to allocate more to it. You can fill these columns with reliable values only with some experience. Its worth striving for though as you must have the most money on your best ideas.

---

Some ideas

Aimtron

I studied Aimtron in some more depth since the last post here where I had a tiny position. It seems interesting on its own, even without the SFP. The company wants to do a 1000 Cr turnover (currently at 300 Cr for FY26) and surprisingly they might have a credible path towards that.

One of the things that stands out (almost like a red flag in fact) with Aimtron is the margins. How does a EMS company make 22% margins? I think the key difference is that what Aimtron does is small batch manufacturing on custom designs. So the design and prototyping cost is apportioned over a small batch during manufacturing which leads to higher margins. It doesn’t look like they are doing high volumes much as of today. Thats why they have so many products in the ppt for a 300 Cr topline company.

Another red flag I noticed is that the company was deriving 55% of its revenues from Aimtron Corp which is the promoter’s other company that’s domiciled in the US (this is FY23 in IPO prospectus). This % though has reduced today to 20%.

Another doubt I had was if the company even had capabilities it is claiming to. It does seem so, going by some of the group companies the promoter has. It looks like Aimtron corp is a 500 Cr business in the US. There is also Aimtron systems which is a ITAR-compliant company serving in defence/aerospace (This company was called “Target corporation” and was purchased in 2016). There is also American Pinball, Inc. which was another group company that made original design pinball machines in the US. This company was bought by JB Vincent LP in Jan '26. Interestingly you will see this as a related party in the IPO docs and most likely Aimtron Electronics was involved in its design and manufacturing (you see a pinball machine in the ppt also under products). So definitely there is some chops here and promoter is a technocrat and first-gen entrepreneur having worked in the EMS space with creative electronics and Sigmatron between 1998-2008 before starting Aimtron in 2009. So my worry that this is a fly-by-night operation is probably misplaced.

Current order book in the standalone entity is ~600 Cr. The guidance for growth in the standalone entity is 40%. I looked at what this company has guided in the past to see if they are reliable and I could see in a past nuvama report the guidance for FY26 was 280 Cr and they have achieved that. They have done a dilution last year to acquire a company called ICS (International Control Services Inc.) by raising 94 Crs. The linked ppt shows what ICS (now called Aimtron International Controls or AIC) does. This seems to be a decision to move from EMS to ODM player. AIC has strong clientele in Caterpillar, John Deere etc.) and also gives them ruggedized electronics capabilities. I feel this is a very good acquisition in terms of capabilities and also in terms of value. This company has $16.9m in revenues (~150 Cr) with 11-12% margins (~16 Cr EBITDA). They paid just 75 Cr for this acquisition (~5x EBITDA) which is great value. This company is operating at 50% util and so has capability to post a 280 Cr topline. FY27 will be the first year with full consolidation of this and this alone should provide a 50% growth to 450 Cr levels. Standalone entity is also guided to grow at 40%, so I think there is good growth. However AIC margins at 11-12% as per latest call is already at 15% (There is a slide with several points target margin improvement and they have already started rationalising) and they expect this to creep up with operating leverage towards where Aimtron electronics today is (20%+) over time.

I noticed one more thing in the Nuvama report.

It looks like the promoter wants to consolidate all his business under the listed entity which will be a get rid of the related party overhang. He can’t do so now because > 25 Cr paid up capital will require a mainboard listing and a mainboard listing can be done only 3 yrs after listing on SME. So my guess is that post June 2027, this business will get reverse-merged into listed entity. If that’s done at a good valuation, I think we are looking at a 1000 Cr topline in pretty short order here.

One thing that gives me hope that it will be done at a good valuation is that during the same time that Aimtron acquired ICS, the group also sold the American Pinball company to JB Vincent. This company was apparently bleeding money and was likely sold at cheap value. He very well could have dumped this onto the listed entity using the pref money but did not. I think what a promoter doesn’t do also tells us more about intentions. I could be wrong but I found it interesting. So overall at ~2500 Cr mcap and ~60x trailing, growing at 50% with visibility for growth over next 2-3 years, I think this company could do well on the mainboard. But do notice that cash conversion, working capital etc. is poor here - I am willing to overlook for the time being. The SFP manufacturing itself will be under Aimtron Mechatronics, a WOS - if and when it happens. The disclosures are really good here considering the company does calls and presentations have so much detail for a SME. I did cross-verify a few details in the ppt but since this post has already gone on so long, I am stopping here.

Asian Energy Services

This seems to me to be a very undervalued business. I first came across it in 2024 and it hasn’t done much since then though business-wise there are several improvements. AESL today has a standalone business that provides exploration services to Oil & Gas companies (seismic surveys, geological studies), development of fields from EPC, facility construction, well drilling/testing and also production (enhancement, O&M of onshore/offshore facilities) and abandonment (plugging, decommissioning).

They acquired a business called Kuiper that provides manpower services (absolutely boring business. Interesting thing here is the price they paid for it - Kuiper does ~500 Cr topline and has 6-7% margins. They acquired this business for just $9.25m or about 3x EBITDA! This business is guided to do $100m in revenue by FY29. Its still unexciting in terms of growth or capabilities but this business alone is probably worth ~450 Cr in listed markets (15x EBITDA multiple).

The standalone business did 492 Cr topline and is guided to grow 40% in FY27 (650 - 700 Cr) with a 16% EBITDA margin. At 15x EBITDA FY27E, this alone should ~1500 Cr in terms of value. Market has somehow not discounted this at all. Maybe because this is an unloved sector and so it may remain a value trap. The standalone business has two very good orders - 865 Cr from Vedanta which would add 150 Cr/yr over 4.75 yrs. Its an interesting model as well where Vedanta reimburses them for the capex and they also get a $/barrel O&M fee. The second is a 459 Cr coal handling plant order from Mahanadi Coalfields which is over 7 yrs. This alone can contribute 200 Cr in FY27. So the 40% growth in FY27 seems to be backed by order book.

The promoter is merging the cash cow business (Oilmax) with the listed entity. Oilmax did 128 Cr revenue in FY25 with a 56% EBITDA margin and has a networth of 290 Cr. This business was valued at 2582 Cr by independent valuer and is the basis for the merger which should be completed by Sept or so. What is interesting is that the management thinks Oilmax can do 800-900 Cr revenue on its own by FY29/30. I did some cursory check to see if that’s feasible.

Oilmax has several assets. In that three of them are currently producing - Amguri (Assam) produces 1850 boepd and has a peak production capability of 3000 boepd by FY29. This is running ahead of estimates because the actual reserves are way above initial estimate. Indrora/Mevad (Gujarat) - Currently doing 200 boepd (was supposed to be 100 boepd as per earlier calls but it has ramped up well too) and peak is 1400 boepd by FY29. Duarmara (Assam). This is not yet producing but has a peak potential of 6100 boepd by FY30. There are a few others like Tiphuk and South Rewa 6 which are either too small or too far away to discount. Duarmara production I think can totally put the thesis in high gear.

Antelopus Selan owns 50% participating interest in Duarmara and it looks like the delay is because they have encountered oil. This is from Q4 ppt

from Antelopus Q3. This seems more like a positive surprise that they have encountered oil and of good quality as well.

I have had bad luck with O&G stocks (HOEC). Lot of things tend to go wrong in these businesses. At least Oilmax is all on-shore fields, so HOEC kind of risk should not happen here hopefully. At least I feel I am not overpaying at current levels as the valuation barely even covers standalone business + Kuiper as of today. I feel merger was done at fair valuation (and not cheap) so we have to see if the non-promoter holders (other than Oilmax) stay on or sell their shares once the new shares get listed (Might happen by Sept).

If my calculations are right, there is likely a 50% upside here based on valuation alone in near-term. Medium/long-term I think can do very well if Duarmara production starts.

Disc: I have recent purchases in Aimtron (from ~1200 levels) and ASEL (~330 levels). I am not qualified to advice. I am a novice and write here to clear my head.

[sumit680]

I have been studying Aimtron and I agree with all your points… But how do you value such businesses… even 1 year forward valuation multiples fail to solve the puzzle..

Any particular mental model here or any case study from past to study and learn on these topics..

Thank you

[Ravi.0104]

in North East, gas evaluation is a challenge. even OIL is constrainted due to this. how OILMAX plans evacuation?