Position: MSFT trimmed 21.3% at $510. Position: CNQ.TO opened at C$70.50, 10% weight. Position: MX.TO opened at C$80.00, 5% weight.

Trimming Microsoft — and why taking profit isn’t the same as losing conviction

Here’s the math first. I bought Microsoft at $392.52. I just sold roughly a fifth of that position at $510, a real, confirmed gain of 29.93% on the portion sold. The stock currently sits at $501.02, meaning even the shares I trimmed were sold close to the actual top of where this position has traded, not a panicked exit into weakness.

I want to be honest about why I did this, because it’s not the reason you might assume. This isn’t me losing confidence in Microsoft. The thesis I built this position on, an elite, wide-moat compounder trading at a genuinely cheap multiple for its growth rate, Azure accelerating, AI revenue run-rate expanding, hasn’t changed. Fair value estimates in the $560 to $600 range still stand, and I still believe in the retained position at that same conviction level.

There’s a mechanical reason for the trim instead. When I built this portfolio, Microsoft was sized at a 10% target weight. Because the stock performed so well, up over 27% from entry, its live weight in the actual book drifted up to 12.7%, a real, meaningful overweight relative to what I’d originally decided was the right amount of conviction to express in a single name. That’s not a flaw in the thesis, that’s exactly what happens when a real winner runs, and if you never trim anything, your portfolio’s actual risk profile slowly stops reflecting the deliberate decisions you made when you built it and starts reflecting whichever stocks happened to go up the most.

So I sold enough to bring the live weight back toward that original 10% target, and I formally reduced the target weight itself to 7.87%, a real, honest acknowledgment that a portion of this position is now permanently realized rather than still fully at risk in the market. The remaining shares still track from the original $392.52 entry, and I still expect them to work.

I think this is worth stating plainly as a real, general principle, not just something specific to Microsoft. A stock going up doesn’t obligate you to keep adding, and it doesn’t obligate you to sell everything either. Real portfolio discipline sometimes just means taking some of the win off the table and letting the rest keep running, exactly what happened here.

Unlock the rest — the complete, real thesis behind both new positions, the exact math behind why Canadian Natural Resources and Methanex are sized differently, and an honest look at how the rest of the portfolio has actually performed since the last update.

Subscribe now

Read more

Here’s the opportunity I actually see. Eight analysts have carried the same Buy consensus on this stock every single month from March through August 2026, unchanged. That kind of sustained, stable conviction is rare, and I wanted to know whether the real numbers actually justified it before trusting the consensus at face value. They do. FY2025 net income grew 103.87% year over year, to $40.51 million. Free cash flow grew 82.98%, to $74.16 million. And the real signal underneath both of those numbers, margins expanded across every single line at once, gross, operating, net, the signature of a business getting genuinely more efficient in multiple places simultaneously, not riding one lucky quarter.

What makes this more than just an improving set of numbers is the real, deliberate strategy behind it. Dexterra is actively shifting toward higher-margin integrated facilities management in the US, a real, structural repositioning rather than incremental tinkering with the existing business. And there’s concrete, verifiable evidence this is actually happening, not just a story in an investor deck. A real, named acquisition is already delivering contract wins under this new model, confirmed directly by management on their most recent earnings call, not just completed and left to sit.

I want to be direct about the real risk sitting alongside this opportunity, because I think that’s what makes analysis useful rather than just promotional. Total debt did grow substantially this past year, and on its own, that number looks concerning. The fuller picture is more reassuring, real, disclosed leverage sits at a genuinely conservative level relative to earnings, with management explicitly guiding for further improvement ahead. But real, actual labor strikes hit some company sites earlier this year, a meaningful share of this year’s earnings comes from items that won’t repeat, and the US expansion this whole thesis leans on is a genuine, live competitive fight, not a done deal.

That combination, real sustained analyst conviction, real confirmed strategic execution, and real risks I’m not going to paper over, is exactly the kind of setup worth digging into properly.

The full report — the complete strategic breakdown, the exact math behind the target price, and every risk laid out in full — is live for premium subscribers.

Subscribe to Premium

Subscribe now

Read more

Two real, opposing forces are running through this coverage universe right now, and August’s results split cleanly along that line. On one side, the AI infrastructure buildout continues pulling real, structural demand into anything touching data centers, power, and electrification. On the other, the Bank of Canada’s rate trajectory is directly, mechanically compressing earnings for any business whose income depends on the spread it earns on cash. Four names reported real, material news this month. Two more have real, dated catalysts still ahead. I want to sort all of it through that single lens, because I think it explains more of what happened than treating each report as its own isolated story.

II. High-Execution Outperformers

Hammond Power Solutions — a real beat, with one number worth reconciling honestly

Record quarterly sales of C$324.8 million, up 44.7% year over year. Adjusted EBITDA of C$53.2 million, up 59.3%. An EPS beat of 29.6% against consensus on the adjusted figures.

Here’s the number I want to walk through carefully rather than skip past, because a sharp reader will notice it immediately. Reported GAAP net income actually fell year over year, from C$13.4 million down to C$9.4 million, with diluted EPS moving to C$0.79. That’s a real divergence sitting directly underneath a genuinely strong adjusted EBITDA number, and I think it’s worth naming exactly why it happened rather than let the adjusted figures stand alone. The gap is driven by non-operating factors, specifically AEG Power Solutions integration costs, real FX headwinds, and share-based compensation, none of which reflect the core business getting weaker, all of which are real, disclosed costs of executing the growth you’re seeing elsewhere in the report.

One more honest detail on the backlog, and I think there’s a real, natural explanation worth adding here rather than just flagging the number in isolation. Up 96.9% year over year is the headline, and it’s real. But it also dipped roughly 6.9% sequentially from Q1, worth noting directly rather than only citing the flattering annual comparison. I don’t think this is a warning sign so much as standard delivery seasonality interacting with a genuinely positive development, Monterrey 4 just reached full capacity, and a facility running at full capacity means Hammond is physically converting backlog into delivered, recognized revenue faster. A record C$324.8 million sales quarter drawing down backlog by high single digits sequentially is close to what you’d expect when a company’s delivery throughput genuinely increases, not a sign demand is cooling. That balanced reporting, showing you the number moved in both directions and explaining why, doesn’t break the structural thesis, it reinforces it. Data centers now represent more than 30% of revenue, real, disclosed exposure to the same AI buildout theme running through this whole market.

The ongoing AEG integration alongside that Monterrey capacity milestone tells you this is a business actively expanding its own delivery capacity, not just watching demand pile up. High priority for continued coverage.

Propel Holdings — real operating leverage, and the credit metrics that prove it

Read more

Part One: The Current Macro State

The transmission mechanism, traced properly

Go back through the Bank of Canada’s own primary statements this year, not summaries of them, the actual language. March: the war affecting oil prices. April: tariffs and trade uncertainty weighing on exports and business investment. June: the conflict “in its fourth month,” global shipping disrupted, oil reserves reported dwindling in multiple countries. July: inflation rising specifically because of gasoline prices tied to the war, with global GDP growth projected to slow to 2.75% in 2026 before recovering to 3.25% in 2027 and 2028. August: a fresh trade-war escalation layered directly on top of all of it.

Six consecutive rate decisions, and the same conflict named as a material factor in nearly every one. This is a central bank telling you, in its own primary documents, that a single geopolitical event is still actively propagating through inflation data months after most coverage moved on.

Here is the mechanism itself, traced through three distinct, real channels. The energy channel is most direct, conflict near major producing and transit regions constrains supply and disrupts routes, pushing crude prices up, which shows first in the gasoline component of headline inflation. The logistics channel operates independently of the oil price itself, when vessels reroute away from contested chokepoints, the actual distance and time required to move the same cargo increases substantially, a real, measurable concept called tonne-mile demand, the combined effect of volume and distance. Tighter capacity relative to demand pushes freight rates up regardless of what crude itself is doing. The insurance and reinsurance channel is the least visible but genuinely real, war-risk and catastrophe exposure in active conflict zones generates real, disclosed claims activity, a direct cost showing up sector-wide in combined ratios and underwriting income.

Follow those three channels outward and one input produces five genuinely distinct sector-level effects, real cost in reinsurance, real margin compression in fuel distribution, a genuine windfall in tanker shipping, broad sector-wide caution in auto suppliers facing compounding tariff pressure, and a separate but simultaneous capital-strain effect across AI infrastructure spending.

Why the central bank is holding still, and why that’s defensible rather than passive

Look at the actual composition of the data. Headline CPI rose to 3.2% in May, driven specifically by gasoline. Core inflation, the measure stripping out the most volatile, externally-driven components, stayed close to 2% through the same period. That gap is itself a real, sophisticated signal. When headline inflation moves sharply while core stays anchored, that’s the textbook signature of a supply-side, externally-driven shock rather than genuine domestic demand overheating.

The shape of the growth forecast matters as much as the number, a real but time-limited dip tied to a nameable cause. And underneath the visible disruption, real, quiet strength, employment up 75,000 in July, unemployment at a genuine two-year low of 6.4%, USD/CAD on a modest, real strengthening path through year-end. That combination, visible external disruption sitting on top of genuine underlying resilience, is the complete, honest macro state right now.

Part Two: The Eight Equity Patterns This Macro State Is Actually Producing

Read more

Position: ONON. Entry: $30.00. Weight: 5%.
Position: REPX. Entry: $35.60. Weight: 5%.

On Holding (ONON)

Position: ONON. Entry: $30.00. Weight: 5%.

A business earning almost 24% return on equity, sitting on a net cash balance sheet, that just fell over 20% in a single day. I want to walk you through exactly why I think the market got the reaction more wrong than the fundamentals.

What actually happened

On August 11, On Holding reported second quarter EPS of $0.44, beating the $0.41 estimate. A real, clean beat. The stock fell hard the same day, and my entry at $30.00 sits roughly 22.6% below where it traded just before the report. That’s not a modest miss reaction, that’s a genuinely severe move for a company that beat estimates.

The real cause, stated directly: management tempered its full-year sales outlook. Not a miss on what already happened, a real, disclosed caution about what’s ahead. I want to be honest that this is exactly the kind of headline that spooks a market already nervous about consumer discretionary spending broadly, and I think that’s precisely why the reaction was disproportionate to what the actual numbers showed.

Why the underlying business tells a different story than the headline

Net sales grew 21.6% at constant currency to CHF 850 million. Direct-to-consumer sales, the highest-margin, most controllable part of the business, grew 34%. Gross margin hit a record 65.4%. Those are not numbers you get from a business losing its footing, those are numbers from a business that’s genuinely getting more efficient and more in control of its own distribution while it grows.

Return on equity sits at 23.95%. Return on invested capital at 32.30%. Both are exceptional, real figures, not adjusted or one-time boosted, this is a business converting capital into profit at a rate most consumer companies never approach.

The balance sheet, and why it matters for how much conviction I have here

Net cash position of $837.75 million, working out to $2.51 per share. Current ratio of 2.83. Debt-to-equity of just 0.29. A business this cash-rich, growing this fast, with margins this strong, getting punished this hard for cautious guidance rather than an actual deterioration, is exactly the kind of gap I look for. The balance sheet gives real, demonstrated staying power to let the thesis play out, it’s not a company that needs the market’s confidence to survive the next few quarters.

The longer-term context I don’t want to hide

Over the past three years, EPS grew 41% annually while the share price grew just 1% annually. I want to sit with that honestly rather than gloss over it. The market has been broadly skeptical of this name for a while now, this isn’t a story where nobody’s noticed the growth, they’ve noticed and largely discounted it anyway. That’s a real, legitimate pattern worth respecting, and it’s part of why I’m sizing this at 5%, not 10%.

The fresh confirmation that made me act now

Evercore raised its target to $46 from $42 immediately after the drop, not before it, real, live conviction responding directly to the same information that spooked everyone else. Average analyst target sits at $44.89, and against my $30.00 entry, that implies roughly 49.6% upside, essentially half again on top of the position. Forward P/E of just 16.44 times, with a PEG ratio of roughly 0.66, genuinely inexpensive for the actual growth rate.

Why 5%, not a full position

Every full ten percent position in this book has cleared a real, demonstrated moat, checked directly. This one clears real quality, real balance sheet strength, real margin expansion, but the honest, multi-year pattern of the stock lagging its own earnings, and the genuinely high volatility, beta of 2.12, mean I want to see this confirm before I go further. Half-size, with real conviction behind it, not a full bet on a name still proving itself to a skeptical market.

Riley Exploration Permian (REPX)

Position: REPX. Entry: $35.60. Weight: 5%.

A forward P/E of 5.68 times, a real 4.55% dividend yield, and production that just beat guidance. I want to walk you through the actual math, because a number that cheap deserves real scrutiny, not blind acceptance.

The real math, worked through properly

Price of $35.60, implied forward EPS of $6.27, giving a forward P/E of 5.68 times. The dividend runs $1.62 a share, a real 4.55% yield at this entry, on a payout ratio of just 50%, meaning roughly half of earnings support the dividend while the other half funds growth and debt reduction. That’s a genuinely balanced use of cash, not a company stretching to maintain a payout it can’t really afford.

Why that multiple is too low, not just cheap for the sake of being cheap

Energy names with real, confirmed production growth and disciplined capital spending don’t typically trade this cheap unless the market is pricing in real doubt about whether the growth continues. I went looking for evidence that doubt was justified, and I didn’t find it. Q1 oil production beat guidance by 5%, which led management to directly raise full-year production guidance off the back of it, not just talk about optimism. Truist reiterated its Buy rating as recently as August 13, specifically citing sustained capex discipline and balance sheet-backed upside.

The scenario math behind the entry

I built three real scenarios off this before sizing the position, all recalculated against my actual $35.60 entry. If production growth genuinely stalls and the multiple compresses further to 4.5 times, you’re looking at $28.21, a 20.7% loss, my bear case, weighted at 20% probability. If guidance simply holds and the multiple normalizes to 7 times, still below the broader energy sector average, you get $43.89, a 23.3% gain, my base case at 50% probability. If the beats continue and the multiple re-rates to 9 times, you’re at $56.43, a 58.5% gain, my bull case at 30%. Probability-weighted across all three: a 25.0% expected return.

The miss I want to name directly

The most recent quarterly EPS came in at $1.54 against a $1.63 estimate, a real miss. I’m not going to pretend it didn’t happen. But sitting on top of production beating guidance and debt coming down in the same period, I read that as noise inside a genuinely strong trend, not the start of a new one. One number moving the wrong direction while every operational signal moves the right one is a real, useful distinction, and it’s exactly the kind of thing worth digging past the headline for.

The dividend, and why it’s real evidence of quality, not just a nice yield

A real 4.55% yield on a 50% payout ratio, held steady through a quarter with an EPS miss, tells you something concrete about how management thinks about capital allocation. They’re not chasing growth at the expense of shareholder returns, and they’re not stretching the dividend to look generous either. Genuinely balanced, and genuinely sustainable at current levels.

Why 5%, not a full position

The real math here is strong, arguably the strongest single scenario analysis in this whole book right now. But energy remains a genuinely cyclical sector, and I want real confirmation across another quarter or two, particularly on whether that EPS miss stays isolated, before committing full conviction. Half-size lets me participate in what I think is a real, statistically compelling setup, while keeping room to add if the next print confirms the trend holds.

Why both, together, today

Two genuinely uncorrelated names, consumer discretionary and energy production, neither overlapping with anything else currently in this book. Both sized at half-conviction for the same honest reason, real, strong theses with one real, specific thing I want to see confirm before going further, ONON’s ability to actually close the gap with a skeptical market, REPX’s ability to keep the recent EPS miss isolated rather than repeated.

Position: ONON. Entry: $30.00. Weight: 5%.
Position: REPX. Entry: $35.60. Weight: 5%.

This is a tracked model portfolio shared for research and educational purposes. Not investment advice. Do your own research.

— Yonatan

Here’s what pulled me in. Six-month revenue grew 166% to $17.3 million. Income from operations grew 285% to $10.0 million.

Trading at $0.98 right now, I want to walk you through exactly how I got to a target, because I don’t think this is a clean story either way. Net income of $3.9 million included a $4.7 million non-cash accounting loss, gold and silver prices dipped during the measurement window, and the metals sitting on the balance sheet got marked down on paper. Real money didn’t leave the building. Strip that out, and the real, underlying operating profit picture is meaningfully stronger, closer to $8.6 million. I think income from operations is genuinely the more honest number to anchor a valuation on here.

Here’s how I actually built the target. Established royalty companies trade at real premiums, but specifically because they hold 100 to 400-plus diversified assets. Empress doesn’t have that yet, and management says so directly, flagging concentration risk given the small number of investments. But two real, already-announced acquisitions, a North American royalty portfolio and a set of pre-production royalties from Almadex Minerals, are a genuine, concrete step toward exactly that kind of diversification. If they deliver, a real, moderate re-rate is defensible. If they don’t, the current price roughly holds.

The full report — the complete three-scenario valuation model at today’s price, the exact reasoning behind the Speculative Buy rating, and precisely what I’d need to see to change my mind — is live for premium subscribers.

Subscribe to Premium

Subscribe now

Read more

Riley Exploration Permian (REPX)

This is the cleanest name on the list, and I want to actually walk you through the math, not just tell you it’s cheap.

Price sits at $34.47, forward P/E of 5.5 times. Back that out and you get an implied forward EPS of $6.27. The dividend runs $1.62 a share, a real 4.7% yield, on a payout ratio of just 50%, meaning roughly half of earnings support the dividend while the other half funds growth and debt reduction, a genuinely balanced use of cash, not a company stretching to maintain a payout it can’t really afford.

Here’s why that 5.5 times multiple is too low. Energy names with real, confirmed production growth and disciplined capital spending don’t typically trade this cheap unless the market is pricing in real doubt about whether the growth continues. The evidence doesn’t support that doubt. Q1 oil production beat guidance by 5%, which led management to directly raise full-year production guidance off the back of it, not just talk about optimism. Truist reiterated its Buy rating on August 13, specifically citing sustained capex discipline and balance sheet-backed upside.

I built three real scenarios off this. If production growth genuinely stalls and the multiple compresses further to 4.5 times, you’re looking at $28.20, an 18.2% loss, my bear case, weighted at 20% probability. If guidance simply holds and the multiple normalizes to 7 times, still below the broader energy sector average, you get $43.87, a 27.3% gain, my base case at 50% probability. If the beats continue and the multiple re-rates to 9 times, you’re at $56.41, a 63.6% gain, my bull case at 30%. Probability-weighted across all three: a 29.1% expected return.

One honest miss worth naming directly. The most recent quarterly EPS came in at $1.54 against a $1.63 estimate. But sitting on top of production beating guidance and debt coming down in the same period, I read that as noise inside a genuinely strong trend, not the start of a new one.

Read more

Position: VMD. Trimmed 50%. Sold at $9.80. Realized: -18.06%.

A business earning an eighteen percent return on equity, trading at a multiple you’d expect from a company barely growing at all.

Before the AER thesis, a quick, honest update. I trimmed Viemed Healthcare by half this week, sold at $9.80, a real, realized loss of 18.06% on the sold portion. Two consecutive EPS misses with a widening gap, from 33% to 42%, was a genuine pattern I couldn’t ignore, even with revenue still growing and hitting records both quarters. I kept the other half open rather than exiting entirely, since management raised full-year guidance in the same report that showed the miss, and the core revenue thesis hasn’t broken. Full reasoning is in its own post if you want the complete breakdown.

Now, AerCap.

I want to walk you through exactly how I got here, because this wasn’t a name that jumped out immediately. Aircraft leasing doesn’t sound exciting. It sounds like the kind of business you skip past looking for something with a better story. That’s precisely why I went and actually ran the numbers instead of skipping past it.

The real question I asked

Same discipline as always. Not whether AerCap is a good company, plenty of people already have an opinion on that. The real question was whether the earnings being reported reflect something durable, or whether the headline number is doing work it shouldn’t be given credit for. I’ve been burned by that distinction before in this book, and I wasn’t going to skip it here just because the stock looked cheap on the surface.

The full reasoning is for premium subscribers. Get the complete thesis, every initiation and portfolio update, the entire archive, and everything I publish next.

Try premium free for 7 days. Read it all, cancel anytime before the trial ends.

Subscribe now

Read more

BioRem makes biofiltration systems, the actual equipment that removes odors, volatile organic compounds, and hazardous air pollutants from the air coming out of wastewater treatment plants, food processing facilities, and industrial sites. It’s based in Guelph, Ontario. Founded in 1990. Genuinely nobody is writing about this stock, and that’s the entire reason I went looking.

Here’s what I found. Revenue grew 13% for the full year 2025. Then it grew 44% in the first quarter of 2026. Then it grew 66% in the second quarter. That’s not a company slowing down, that’s one accelerating, quarter over quarter, in real, confirmed numbers, not a projection. Each period grew faster than the one before it, and that specific shape matters, it’s the signature of genuinely strengthening demand, not one lucky contract landing in a good quarter.

And the company’s order backlog, essentially the contracted, not-yet-delivered revenue already locked in, hit a record $77 million. Against a trailing revenue run rate of roughly $46 to 48 million, that’s more than a year and a half of revenue already sitting there, waiting to be built and shipped. This backlog has grown at every single checkpoint I could verify going back over a year, which tells you something real, new orders have consistently been outpacing what the company delivers.

I want to be honest with you about something before I go further. The stock has moved since I started building this coverage, up from around $3.09 to $3.75 now, a fresh high. I’m publishing this at today’s real price, not the level where the case looked cleanest, because showing you the number that actually matters right now is the only honest way to do this. The margin of safety here is genuinely thinner than it was even a few weeks ago. I’d rather tell you that directly than let you find out by comparing this to a stale target.

Even accounting for that, this still trades at roughly 10 to 11 times forward earnings, against an industry average closer to 17 to 18 times. That gap is real. Whether it stays that wide, or the market finally catches up to what’s actually been happening in the numbers, is the entire question this report is built around.

I also want to be straight about the real risks, because a growth story this exciting is exactly the kind of thing that’s easy to oversell. There’s a genuine, disclosed securities offering filing I haven’t been able to confirm the exact terms of, and that’s a real dilution risk worth knowing about before you look at any target price. This is also a genuinely tiny, illiquid stock, the kind of thing that can move hard on a single order being delayed. I’m not going to pretend those risks aren’t real just because the growth story is exciting, and at today’s higher price, the bear case in this report actually shows a real loss, not just a smaller gain, if things go wrong.

The full report — the complete valuation model at today’s price, the exact reasoning behind every scenario, and precisely what would change my mind here — is live for premium subscribers.

Subscribe to Premium

Subscribe now

Read more

Savaria (SIS.TO)

What happened: Revenue hit $245.8 million, up 8.4%. Broken down, 6.6% of that was organic growth, the real test of underlying demand, 0.8% came from acquisitions, and 1.0% was currency working in their favor. Accessibility, the elevator and lift business and 78% of revenue, grew 8.7%. Patient Care, medical beds and mobility equipment, grew 7.3%. Gross margin expanded 60 basis points to 39.6%. Adjusted EBITDA hit a best-ever $51.8 million, up 10.7%, at a 21.1% margin.

Why the headline operating income number is misleading, and I want to walk through this carefully. Operating income jumped 34.1%, which sounds extraordinary. Here’s the actual mechanism behind it: last year’s quarter carried $4.6 million of one-time expenses tied to a completed strategic program called Savaria One. Those costs simply didn’t recur this year. That absence of a cost, not a genuine surge in the business, explains roughly 19.7 percentage points of that 34.1% headline number. Strip that out and you get real, solid, but far less dramatic underlying improvement. This is exactly the kind of number that looks incredible in a press release headline and means something much more modest once you actually understand what’s driving it.

Why Patient Care’s margin fell while Accessibility’s rose, and why that matters. Accessibility’s EBITDA margin expanded to 23.6% from 21.9%, genuine operating leverage as the larger, more mature segment scales. Patient Care’s margin actually fell to 18.4% from 20.9%, a real 250 basis point decline sitting inside an otherwise strong quarter. The release doesn’t spell out exactly why, cost pressure or an unfavorable product mix are the likely candidates, but this tells you the business isn’t uniformly strong right now, and it’s worth watching whether this stabilizes or continues.

Why the balance sheet story is genuinely, unambiguously good. Net debt fell to $172.8 million from $191.5 million, with net debt to EBITDA improving to 0.87 times from 1.03 times. That’s real, active deleveraging happening in real time, not a target someone’s hoping to hit. Available funds of $333.4 million give real flexibility for the next acquisition, and Vipal S.p.A, which closed July 1, is already deployed capital working toward the next quarter’s growth.

The valuation, built off management’s own real, dated target. They’re targeting $1.6 billion in revenue and roughly $4.25 of adjusted EBITDA per share by 2030. Current run-rate EBITDA per share sits around $2.74, meaning that target implies an 11.6% annual growth rate from here, a number that’s internally consistent with their stated 12% top-line growth target once you layer in modest margin expansion, not a number picked just to sound impressive.

Running three real scenarios off that 2030 target: if growth stalls below plan and the market keeps the multiple depressed at 10 times, you land around $42.50 by 2030, a modest 5.4% annualized return, the bear case. If the target is hit roughly as guided and the market grants a reasonable 14 times multiple, you land around $59.50, a genuinely strong 14.6% annualized return, the base case. If growth actually exceeds the plan through continued acquisitions like Vipal and the market re-rates the multiple to 18 times, you land around $76.50, a 22% annualized return, the bull case.

Propel Holdings (PRL.TO)

What happened: Nearly every single metric hit a record simultaneously. Revenue of $179.6 million, up 26%. Adjusted EBITDA of $43.7 million, up 24%. Adjusted net income of $24.8 million, up 29%. Adjusted diluted EPS up 28% to $0.58 USD. Adjusted return on equity of 35% annualized.

Why so many records landed in the same quarter, and what actually drove it. Total originations funded hit a quarterly record of $243.4 million, up 25%. New customer originations grew 34% on their own, and 43% once you include their newer Lending-as-a-Service business. That combination, new customer growth accelerating faster than the existing book, is the actual mechanism behind revenue, EBITDA, and net income all setting records together, you’re not just growing what already exists, you’re adding meaningful fresh volume on top of it.

Why the stable credit performance matters more than the growth numbers themselves. Here’s the real risk in any fast-growing lending business: you can juice short-term growth by approving worse borrowers, and it looks fantastic right up until the losses show up a couple of quarters later. Provision for loan losses held at exactly 50% of revenue, unchanged from a year ago. Net charge-offs held at 12% of average loan balances, also unchanged. That means they grew originations 25% without loosening underwriting standards to do it. That’s the real, concrete evidence this growth is sustainable rather than borrowed from the future.

Why the falling cost of debt is real, meaningful money, not a rounding error. Cost of debt fell to 10.2% from 11.4%, driven by both improving company-specific credit quality and a more favorable rate environment. A drop of over a full percentage point on hundreds of millions of borrowings flows straight to the bottom line.

Why this business can grow fast and raise its dividend twelve straight quarters at the same time. A 35% adjusted return on equity is the actual engine behind this. When a business earns that much on retained capital, every dollar not paid out compounds at a genuinely exceptional rate. Assuming a conservative 30% payout ratio, consistent with the real dividend growth pattern, the retained-earnings-driven compounding works out to roughly 24.5% annually. That’s the real, mathematical reason Propel doesn’t have to choose between growth and shareholder returns, the return on capital is high enough to fund both simultaneously.

Unlock the rest — Kits’ real trade-off explained through the actual analyst exchange, why Calian’s backlog matters more than its quarter, the real reason Pulse Seismic’s 80% revenue drop isn’t what it looks like, the Waymo risk quantified into an actual return range for Uber, and why Viemed’s widening miss is the one number that actually needs watching. Free trial running now.

Subscribe now

Read more

Active Buy — the core of the coverage universe, and it worked

Thirteen positions. A 76.9% win rate. Average return of +10.10%. Median of +4.35%.

This is the tier that matters most, because it’s where the bulk of real conviction calls live, and ten of thirteen names are positive. Savaria (+38.6%), Badger Infrastructure (+36.6%), Calian Group (+23.8%), Propel Holdings (+20.6%), and BQE Water (+19.0%) form a genuinely strong core. Exco Technologies (+13.7%) and Cipher Pharmaceuticals (+4.3%) add real, further depth. Richelieu, Total Energy, and BioSyent sit closer to flat, real, live positions still developing. Only three names in this entire tier, Viemed, Vitalhub, and Zedcor, are actually negative, and none by a dramatic margin relative to the tier’s overall strength.

A near 77% win rate on the highest-frequency conviction rating is a real, meaningful validation of the process behind these calls, not a coincidence.

Strong Buy — small sample, but exactly what you’d want to see

One position. Hammond Power Solutions, +36.0%. Reserved for the highest conviction, and it delivered. Too small a sample to draw a full statistical conclusion from alone, but directionally, exactly right.

Speculative Buy — appropriately modest, exactly as the label promises

One position. Olympia Financial Group, +2.9%. Precisely what a Speculative Buy rating should look like in practice, not a loss, but not the same conviction-driven outperformance as the Active Buy tier either. The rating told readers to expect real uncertainty, and the actual result matched that framing honestly.

Watching — a wider net by design, and it caught two of the best calls in the entire coverage universe

Fourteen positions. This tier exists for a specific reason, it’s where names get real, ongoing attention before they’ve earned full conviction, genuinely earlier-stage, wider-range territory than Active Buy by design. A 21.4% win rate and a -6.67% median here isn’t a failure of the tier, it’s the tier doing exactly what a lower-conviction category is supposed to do, taking on real uncertainty so Active Buy doesn’t have to.

And here’s what that willingness to watch things early actually bought. Data Communications Management returned +69.7%. ADF Group returned +56.0%. Those aren’t just the two best performers in the Watching tier, they’re the two best performers in the entire coverage universe, ahead of every single Active Buy and Strong Buy name on the board. That’s the real, honest value of keeping a Watching category at all, it’s where the earliest, highest-upside names get flagged before the evidence is strong enough yet to call them a full Buy.

It’s also worth noting the tier’s average return still came in positive, at +0.72%, even with more losers than winners inside it. That’s these two names doing real, disproportionate work, exactly the kind of asymmetric outcome a well-run watchlist is supposed to produce, a lot of small, contained misses, and a couple of genuinely oversized wins that more than cover them.

The honest, useful question this raises: did the Watching rating undersell DCM and ADF at initiation, or did real, subsequent developments justify a move that a Watching rating, by design, doesn’t fully anticipate? Either way, both have earned a real conversation about upgrading to a higher-conviction tier now that the evidence has caught up.

The bottom line

A 76.9% win rate at Active Buy against a 21.4% win rate at Watching is a real, statistically meaningful gap, exactly what you’d want a working conviction framework to show. But the Watching tier’s real job isn’t to win most of the time, it’s to catch the names before anyone else is convinced, and this time, it caught the two best performers in the whole coverage universe. That’s the tier doing its actual job, not falling short of Active Buy’s.

What actually happened: revenue of $90.01 billion, up 18%, against $87.62 billion expected. Non-GAAP EPS of $4.74 versus $4.24 consensus, though 27 cents of that came from a $3.2 billion Anthropic investment gain and lower retirement costs, real money, but not repeatable operating income. The core story was Azure crossing $100 billion in annual revenue for the first time, growing 43% in the quarter against a 39-40% guided range.

Why it’s good: the beat was broad, not manufactured by one line item. Azure accelerated above guidance rather than just meeting it, and Productivity and Business Processes ($37.85B, +14.3%) beat its own consensus too. Microsoft 365 Copilot went from 20 million to over 30 million paid seats in a single quarter, real, monetizing adoption, not vanity usage numbers. Forward Azure guidance of ~45% is above the prior 41.4% consensus, meaning management is signaling acceleration, not just holding steady.

Why there’s real risk sitting inside it: free cash flow fell 23%, and capital spending hit a record $41 billion in the quarter alone, with FY2027 guidance even higher. Running the rough math off this quarter’s non-GAAP run-rate, the stock now trades around 22.8x forward earnings post-move, not expensive for this growth, but no longer the deep discount it was when the AI-spending selloff had it near a 52-week low. The bet is now fully underwritten on execution continuing to outrun the spend, with less room for disappointment than existed a few months ago.

The honest verdict: this was the cleanest, most convincing report of the week, and the stock’s 8.9% move was earned, not sentiment-driven. The comparison to Alphabet and Meta, both of whom fell the same week on similar spending stories, is the real tell: the market is pricing Microsoft’s execution as genuinely superior right now, not just riding the same AI wave everyone else is on.

American Express (AXP)

What actually happened: EPS of $4.53 beat a $4.40-4.45 estimate. The stock fell nearly 6% anyway.

Why the underlying business is genuinely strong: net write-off rate held flat at 2.0%. Thirty-day delinquency sat at 1.4%, below the company’s own 20-year average of 1.5%, meaning credit quality right now is better than the long-run historical baseline, not just “fine.” Management released credit reserves rather than building them, a real signal that internal risk models see improving, not deteriorating, conditions. Card spending grew at its strongest pace in three years, and 65% of new consumer accounts came from Millennial and Gen Z customers, a durable, structural signal, not a one-quarter blip.

Why it fell despite all that: revenue missed the specific consensus dollar figure by a modest amount even while growing a real 19.4%. Expenses rose 12% to $14.5 billion, driven by variable engagement costs tied to a Platinum Card refresh, a real, disclosed, largely one-time driver. Management reaffirmed full-year guidance instead of raising it, and on a stock that was already priced for continued perfection, “reaffirmed, not raised” reads as disappointment even when nothing actually broke.

The honest verdict: at roughly 18x forward guided earnings after the drop, against fresh analyst targets in the $380-400 range, this looks like the market punishing a temporary expense bump on an otherwise excellent quarter, not a fundamental problem. The real test is whether those Platinum-refresh costs stabilize by the next report, if they do, this was a buying opportunity disguised as a bad print.

Unlock the full article — Weyerhaeuser’s real recovery, SK Hynix’s live 6x-vs-19x valuation gap, Constellium’s turnaround confirmed twice now, and two names almost nobody else touched this week: Hammond Power’s near-doubling backlog and Badger’s real unit economics story. Free trial running now.

Subscribe now

Read more

SK Hynix — a record quarter the market still doesn’t trust

Here’s a stock that just posted the best quarter in its history and fell anyway.

SK Hynix reported revenue of 79.3 trillion won, up 257% year over year, and operating profit of 60.5 trillion won, up 557%. A 76% operating margin. Both numbers are all-time records for the company. Net profit included a large one-time gain from an asset disposal, so I stripped that out to see the real, recurring number, and even the adjusted figure is genuinely strong. HBM4, the next generation of high-bandwidth memory that’s become the critical bottleneck component for AI infrastructure, just began mass production shipments. SK Hynix holds the number one global market share in HBM at 56.4%.

The stock fell anyway. Not once, either, it’s been a repeated pattern over the past month, including another drop the same day this record quarter was reported.

Here’s the actual debate happening in the market right now, and it’s worth understanding precisely because it’s the same question hanging over every AI infrastructure name. One research firm makes a specific, checkable argument that the selloff has wrongly turned a company-specific pass-through timing issue into a false narrative about industry-wide demand collapsing. Their evidence: Korean customs export data shows memory exports rising from 39.8 billion dollars in Q1 to 62.3 billion dollars in Q2, with June the strongest month on record. HBM contracts get negotiated over long periods, so reported revenue can lag real spot demand increases. That’s a real, technical, falsifiable claim, not just optimism.

What actually got my attention is the valuation. Forward P/E sits at just 6 times, against a 19 times average for the broader South Korean semiconductor industry. Even after everything, that gap is enormous. Management is guiding for DRAM supply constraints to persist all the way to 2030, and has locked in significant HBM orders from Nvidia specifically.

I want to be honest about the risk here too. This is a genuinely volatile, cyclical business, and if the bear case is right, if this really is a demand air pocket and not just a reporting lag, cheap can get a lot cheaper before it’s actually cheap. I don’t think this is a name for a modest position. I think it’s a name where you decide which side of a specific, real debate you believe, and size accordingly.

Running the actual scenario math, using the real 6x-vs-19x gap as the anchor: if the bear case plays out, Tessara’s export-data thesis proves wrong and the market keeps treating this as genuine demand collapse, pushing the multiple toward historic trough levels, you’re looking at roughly -25% downside. If demand proves durable and the multiple simply normalizes partially, not fully, toward peer levels, that’s +50%. If the HBM4 and Nvidia-locked-in supply story plays out as management guided through 2030, the bull case runs +117%. Weighting those by how likely I think each is, the probability-weighted case comes out to roughly +44%, genuinely one of the widest, most binary setups in this piece, which is exactly why I framed it as picking a side of a real debate, not handing you a target.

The full breakdown continues below — Constellium’s already-delivered turnaround, the real reason American Express fell despite beating, and the small, quantifiable discount at Global Indemnity that almost nobody’s talking about.

Subscribe now

Read more

A stock the market treats like a taxi app with no real advantage, sitting on the one thing that actually is one.

I want to walk you through exactly how I got here, because this wasn’t an easy add. Everyone has an opinion on Uber already, and most of those opinions were formed years ago and never updated. I went and actually tested the specific claim, does this business have a real, defensible moat, or is it just a big app that happens to work, using the same rigor I’d apply to anything else in this book. The answer surprised me.

The test I actually ran

I didn’t ask whether Uber is a good company. Everyone already has a take on that. I asked something narrower and more useful: is the competitive position defended by something real, happening right now, or by a story about what might happen eventually. That distinction matters enormously, because a lot of “moats” people talk about in this market are really just hopes wearing a business-strategy costume.

The full reasoning is for premium subscribers.

Get the complete thesis, every initiation and earnings review, the entire portfolio archive, and everything I publish next.

Try premium free for 7 days. Read it all, cancel anytime before the trial ends.

Read more

It showed up.

Current EBITDA margin sits at 10.97%, essentially at that target already. Net income jumped from $14.41 million in Q1 to $23.24 million in Q2, a real, sequential 61% improvement. Management’s own words on the call: a stronger second half is coming, acquisition activity continues, and they’re actively investing in expanding the sales force.

This is a business that’s been quietly compounding for decades the unglamorous way, buying smaller, well-run specialty hardware distributors across Canada and the US, one at a time. They just completed their hundredth acquisition. That’s not a marketing line. That’s a genuine, demonstrated track record.

I want to be honest about what’s still uncertain. One strong quarter isn’t a settled trend, and the next report in October is the real test of whether this holds. The tariff and currency exposure that caused the original miss hasn’t gone away, it’s structural. But I told you exactly what I needed to see before I’d call this a real opportunity, and instead of moving the goalposts, I’m telling you it arrived.

The full report — the complete valuation model, the exact scenarios, and precisely what would change my mind — is live for premium subscribers.

Subscribe to Premium

Subscribe now

Read more

The thing everyone’s missing: two separate stories just became one story

For most of this year, there have been two distinct narratives running in parallel. One was geopolitical, Iran, then the ceasefire, then the ceasefire breaking, oil swinging around. The other was about AI infrastructure spending, how much the hyperscalers are pouring into data centers, and whether it’s paying off. I’ve been tracking both, but treating them as separate.

They’re not separate anymore. Last week, an analyst put something plainly that I think is the single most important framing for the next quarter: elevated oil prices and heavy AI infrastructure spending are simultaneously stressing the bond market. Both are pushing rates higher, at the same time, for different reasons. Oil-driven inflation pressure on one side, “heavy credit demand associated with funding monumental technological ambitions” on the other. When two separate forces push the same direction on rates, that’s not a coincidence you can ignore, it’s the backdrop for everything else this quarter.

Here’s the oil piece in full. The Iran ceasefire held for about three weeks after June 17, oil settled around $69-70. Then it broke on July 8, oil spiked to $78. It cooled some by mid-July. And then, days ago, a second, distinct shock landed, Houthi attacks on tankers in the Red Sea, sending Brent above $100 a barrel for the first time since May. That’s not the same story repeating, it’s a second flashpoint layering on top of the first. If you’re underwriting a “rates ease gradually through 2026” thesis, that thesis needs oil to cooperate, and right now oil is not cooperating.

The AI capex story just got a lot more concrete

This is the part I think deserves the most attention this quarter, because it stopped being theoretical.

Alphabet just reported the first negative quarterly free cash flow in its history as a public company. Not a slowdown, not a miss, negative. Capex hit $44.9 billion, more than double a year earlier, and full-year guidance got raised again, to $195-205 billion, with 2027 expected to go higher still. They’re funding the gap with real, disclosed debt and equity issuance, $20 billion in new debt, an $84.75 billion equity raise in phases. That’s a company telling you directly the spend has outgrown what operations alone can cover.

Oracle is arguably in a more precarious spot. Its dividend hasn’t been covered by free cash flow since 2024, not one bad quarter, an ongoing structural issue. Capex is guided up to $95 billion for fiscal 2027, funded partly by a planned $40 billion capital raise. The backlog is genuinely enormous, $638 billion in contracted revenue including a $300 billion OpenAI deal, but there’s real customer concentration risk sitting inside that number, and one specific, credible scenario, GPU utilization falling below 70% for two straight quarters, could push the stock materially lower if a major customer shifts workloads elsewhere.

Tesla’s Q2 showed the same capital-intensity pressure from a different angle. Operating margin collapsed to 1.4% from 4.1% a year earlier, free cash flow went negative, and management quietly stepped back from volume production targets on three separate new products in the same report, the Cybercab, the Semi, and Megapack 3. That’s not one soft quarter, that’s a concrete retreat from stated ambitions on multiple fronts at once.

Here’s the comparison that actually matters for how I’m positioned. Microsoft, which I hold, is in this same capex race, capex grew 84% last quarter too. But Microsoft’s operating cash flow is still running about one and a half times higher than its capex. It hasn’t crossed the line Alphabet and Oracle have already crossed. Its AI business alone is doing over $37 billion a year now, covering something like 30% of what it spends on capex every quarter, a real, measurable number, not a talking point. That ratio, cash flow to capex, is the single thing I’m watching most closely into its report this week. If it holds, the thesis I built stays intact. If it narrows meaningfully, that’s a real, different conversation.

The honest, sector-wide read: this isn’t one company’s story anymore. It’s now visible, in hard numbers, across the group. Some of these companies can fund it comfortably. At least two, right now, cannot.

The other side of this: where the real opportunities have actually been sitting, away from the AI headlines

Read more

A P/E that scares people off, sitting on top of a business that’s actually one of the cheaper things I’ve looked at all year.

New position going into the portfolio, and this one’s a little different from the others, because the number everyone sees first is exactly the wrong number.

Viemed Healthcare trades around 31 times earnings. Glance at that and your instinct says walk away, and it’s priced for perfection or nothing to see here. I get it, I’d have the same reaction if I stopped looking there. I didn’t stop there, and here’s why.

What Viemed actually does

Home-based respiratory care. Ventilators, sleep therapy equipment, oxygen, the kind of durable medical equipment that lets patients with chronic respiratory conditions manage their care at home instead of in a hospital. Viemed’s respiratory therapists work directly with patients and physicians, managing the equipment and monitoring compliance, which is both the real clinical value here and the reason the relationship with payers stays sticky. It’s a recurring, needed service in a corner of healthcare that’s genuinely growing, as more care shifts out of expensive hospital settings into the home.

The full reasoning is for premium subscribers. Get the complete thesis, every initiation and earnings review, the full portfolio reasoning, and the entire archive, plus everything I publish next.

Try premium free for 7 days.

Subscribe now

Read more

Olympia entered into what’s called a Reorganization Agreement. In plain terms, they’re terminating the existing management agreements they had with a company called Tarman ATM Inc., and with some of their own senior executives, and paying them out in cash and shares worth about $49.7 million total. It still needs shareholder approval, so this isn’t finalized yet, but it’s a real, formal thing, not a rumour.

Here’s the number that actually matters if you own this stock. Olympia’s market cap was sitting around $276 million before this. Issuing $49.7 million in new shares against that is close to 18% dilution. That’s not small. Every number I published in the initiation, the DCF, the EV/EBITDA range, the dividend model, the whole valuation triangulation, gets pulled down by roughly that much on a per-share basis. I built all of that before this was public, so I want to say plainly: it’s not current anymore, at least not without adjusting for this.

Now, is this just bad news? I don’t think it’s that simple.

The shares involved are locked up in escrow for a few years, and the executives’ portion actually gets forfeited if they quit or get fired for cause. That’s not nothing, it’s real retention structure, the kind of thing you’d actually want to see if a company’s trying to keep people around through a transition, not just handing out a payday.

And there’s a plausible reason this is happening at all. Olympia’s been quietly working toward becoming a federally regulated trust company for a while now, which would let them finally operate in Ontario, the biggest financial market in the country, and one they currently can’t touch. Cleaning up management agreements like the Tarman one is usually exactly the kind of thing regulators want to see cleared up before they’ll approve something like that. So maybe this dilution is the cost of unlocking something much bigger down the road.

I don’t know that for sure yet. I want to be upfront about that instead of pretending I’ve got it all figured out.

So where does this leave things? Rating stays Speculative Buy. The core reason I liked this, the real asymmetry between what has to go wrong versus what has to go right, hasn’t changed. But that $120 target I published needs to be treated as under review until I can properly work the dilution into it, and until there’s more clarity on whether this Ontario angle is actually real or just something I’m reading into it.

I’ll come back to this once the shareholder vote happens or there’s more to go on. Wanted you to have this now rather than find out later that I knew and sat on it.

Not investment advice.

— Yonatan

I want to walk you through something I don’t always do, showing you all the math, not just the conclusion.

Olympia looked, on the surface, like exactly the kind of underfollowed small cap I go hunting for. Through its subsidiary Olympia Trust, it administers self-directed registered accounts, the RRSPs and TFSAs that hold private mortgages and other non-traditional assets, a business the major banks have no real interest in building infrastructure for. A return on equity historically above 40%. Covered by essentially one analyst.

Then I found the real, current numbers. 2025 revenue fell 3.9%. Full-year earnings fell 17%. Trailing revenue is running below even that reduced figure. And in the last few days, Olympia cut its monthly dividend roughly 17%, from $0.60 to $0.50.

So instead of picking one valuation method and calling it a day, I built three, independently, using different math entirely.

A discounted cash flow model, projecting real cash generation forward and discounting it back. An EV/EBITDA model, comparing the business to what similar companies actually trade for. A dividend discount model, since this is a real, dividend-paying business, not a speculative story stock.

Here’s what happened: all three, at their base-case assumptions, landed within a few dollars of each other, clustering close to where the stock trades right now. When three different methods built on three different sets of assumptions converge like that, it’s a real signal, not a coincidence.

The honest, boring answer sitting underneath all that work: Olympia is roughly fairly priced today, not obviously cheap.

But I didn’t stop at fair value, because there’s a real asymmetry worth understanding. The scenario where this gets meaningfully cheaper requires two separate things to go wrong. The scenario where it gets meaningfully more valuable requires only one thing, and it’s already the consensus economic forecast among major bank economists: interest rates simply holding where they are instead of falling further.

The full report — every model shown in detail, the exact target and exactly how I built it, and the specific numbers that would move this to a plain Buy or a Sell — is live for premium subscribers.

Subscribe to Premium

Subscribe now

Read more

Buying the best-run capital allocator on the market at a multiple that only looks expensive if you’re reading the wrong number.

New position going into the book, and I want to walk you through this one carefully, because on the surface it’s going to look like a strange fit.

Brookfield Corporation reports a P/E of roughly 84 times earnings. If you just saw that number cold, you’d assume I’d lost my mind putting 10% of the portfolio into something trading at 84x. I get it. Stick with me, because that number is one of the most misleading you’ll find anywhere in the market, and once you see why, I think you’ll understand exactly why I bought it.

Here’s what Brookfield actually is. It’s the parent holding company for one of the largest alternative asset managers in the world, infrastructure, renewable power, real estate, private equity, credit, all of it. It’s run by people who are, genuinely, some of the best capital allocators alive. 2025 was a record year for the business: $112 billion of new money came into the asset management arm, $91 billion got monetized out of mature investments, $126 billion got deployed into new ones, and the company bought back over $1 billion of its own stock along the way. It’s currently sitting on $188 billion of dry powder, ready to deploy when opportunities show up.

That’s the business. Now here’s why the 84x P/E lies to you about the price you’re actually paying for it.

The full reasoning is for premium subscribers. Get the complete thesis, the exact number I use instead of GAAP earnings, every initiation and earnings review, and the entire archive, plus everything I publish next.

Try premium free for 7 days. Read it all, cancel anytime before the trial ends.

Subscribe now

Read more