The Growth Engine sits at 4 of 7 categories positive. Two of those four — Labor and Global — read 100%, and both are the load-bearing ones.

Both also rest on a single indicator each. One data point apiece, swinging a full seventh of the panel.

Friday’s July employment report came in weaker than expected. Labor is the gauge that can least afford it.

That is the first genuine pressure on this panel since March, and it arrives with the domestic side already hollow underneath: Consumer at 33%, Liquidity at 0%, Housing at 0%. Real incomes, sentiment, the monetary base, building permits — all contracting. The weakness isn’t scattered across the economy. It’s concentrated on one side of it: domestic demand and policy liquidity.

What’s holding the constructive read together, then, isn’t the economy. It’s the price trend. The panel clears its Stage 4 threshold with one category to spare and no single category is pivotal — but a close below the S&P’s ten-month line cuts two stages at once, 100% to 33%, regardless of what the panel says. It’s the only lever with that kind of power, and it’s exactly the one that moved this model in March.

Which is the honest frame for the whole issue: the broader momentum index still reads Expansion at 1.59, but that’s 46% below its 15-year average of 2.98. Positive and slowing, not positive and accelerating. A real expansion, just a cooler and later-cycle one than the headline stage implies.

The three levels

1. The ten-month line — the only lever that matters. Everything else on this panel moves the allocation in single steps. This one moves it two: Stage 4 to Stage 2, 100% to 33%, in one close. Base rate says a break within three months of a Stage 4 read happens ~20% of the time, and from a Stage 4 the model has historically held that stage a median of 8 months. Low probability. But with Consumer, Liquidity and Housing already red, there is nothing underneath to slow the fall if it lands. If you’re mirroring this engine, this is the level to have an actual plan for — not the panel.

2. Diffusion back above 50% — and what a reopen actually buys you. The gate reopens above 50%, and it sits at 40% with three of the six falling votes inside a tenth of a point of flat. That is a genuinely live reopen, not a distant one. Two cautions. First, 44% of stage changes reverse within two months, and this engine already round-tripped cash in March–April. Second — the part worth pricing — a reopen does not return you to where you were. Momentum re-picks between U.S. and ex-US at the reopen, and Geography rotated out of ex-US, not out of the U.S. If you’re sizing a re-entry, size it to whichever sleeve wins the twelve-month momentum contest at that moment, not to the position that was closed.

3. Consumer — the only path to a wider base. Consumer at 33% is two indicators from flipping (real personal income ex-transfers, and sentiment). It is the nearest laggard and the only realistic route to a read that rests on something other than price and payrolls. A flip there doesn’t change the allocation — the panel already clears — but it changes what the allocation is standing on, and that’s the difference between holding this and holding it comfortably.

Atlas Report 2026 08 10 (3)

627KB ∙ PDF file

Download

Download

The Growth Engine holds at Stage 4 Constructive: 100% equity, price above its ten-month line, fourth consecutive month at full allocation. Geography went risk-off, sold its international position, and moved entirely to cash. Global diffusion fell 20 points to 40%.

Held together at 50/50, that’s a portfolio that went from 100% effective equity to 50% in one strike. First disagreement since March.

Two votes closed the gate, and neither was European

Geography’s gate is a vote across ten economies: how many have leading indicators rising month over month. Above 50%, own equities. At or below, step out. Four of ten are rising. That’s the 40% that shut it.

Here’s what’s worth sitting with. For three issues running, every falling economy in this panel was European: Germany, France, the UK, Italy. Four down, six up, 60%, gate open. That configuration held for months. Europe deteriorated the entire time and the model stayed fully invested in international equity right through it.

This month Brazil and China turned down. Two votes. Six to four, 60% to 40%, risk-on to cash.

So, the thing that finally closed this gate wasn’t Europe getting worse. Europe is doing roughly what it was already doing, France is down a tenth of a point over six months, Germany two-tenths. What changed is that the weakness stopped being continental. That’s the difference between a regional soft patch and a global one, and the model doesn’t have a category for “regional.” It counts.

Brazil is the tell. It carries the highest CLI reading in the set at 103.5, up 1.6 points over six months. It sits at the top of the table by level, and it cast a red vote by direction. Diffusion doesn’t ask which economies are in good shape. It asks which ones got better last month.

The other engine is fully invested, and the economy isn’t what’s holding it up

Four of seven gauges are green. Labor and Global are maxed at 100%. Business and Markets sit at 75%.

The three reds are Consumer (33%), Liquidity (0%) and Housing (0%), real incomes, consumer sentiment, the monetary base, building permits. Every one of them is domestic demand or policy liquidity. The weakness isn’t scattered across the economy. It’s stacked in one corner of it.

So, what’s actually carrying the constructive read? Not the panel. The panel clears Stage 4 with a one-category cushion, it could lose a green and hold. The lever with real power isn’t in the panel at all. It’s the ten-month trend line, which cuts two stages at once, 100% straight to 33%, regardless of what the seven gauges say. That’s the lever that dropped this model to Stage 2 in March. Not a panel collapse, a trend break.

Which means the honest description of the current position is this: the model is fully invested because price is holding, not because the economy is broad. With domestic demand entirely red underneath, the market’s own trend is doing the load-bearing work.

The EMI says the same thing in one number. It reads 1.59: positive, consistent with expansion while sitting 46% below its 15-year average of 2.98. Real expansion. Cool one. Positive and slowing rather than positive and accelerating.

One more thing worth noticing

The Growth Engine’s single strongest gauge is Global, at 100%. It resolves from OECD leading-indicator diffusion. Geography’s gate, the one that just shut, resolves from OECD leading-indicator diffusion.

Same family of data. One reads green, one reads red. They’re measuring it differently one against its own long history, the other against last month and that gap is the whole story of this issue in miniature. The global cycle is historically fine and currently deteriorating.

The full Issue 05 report is attached below: the country panel, the seven-gauge attribution, base rates from a Stage 4 reading, the scenario table, the fifteen underlying indicators, and the complete trade log. Everything above is the argument. Everything in the PDF is the evidence.

Atlas Report 2026 08 03

625KB ∙ PDF file

Download

Download

Growth Engine: Stage 4 Constructive, fully allocated, price above its ten-month line. Geography: risk-on, diffusion at 60%, every dollar of equity outside the U.S.

Third issue running with the same reads. For a rules-based model that’s the job working — it doesn’t move because the news moved.

What makes this week different isn’t a signal. It’s the calendar.

Both engines strike at month-end. Month-end is Friday. And almost everything capable of moving a gauge land in the thirty hours before it.

The soft spot is the thing getting refreshed

The panel is four of seven positives. Global and Labor carry it at 100%. Consumer drags at 33% — and Consumer is the nearest laggard to flipping, two indicators from turning positive and giving this read a domestic-demand floor it currently doesn’t have.

Thursday at 8:30 a.m., personal income and outlays lands. That one report carries two of the three indicators inside the Consumer block. Friday at 10:00 a.m., a little over an hour before the strike, U-Mich sentiment closes the third.

The weakest block on the panel gets entirely rewritten in the last day and a half of the month. Then the model acts on it.

One of those three numbers is currently stale

Worth saying plainly, because most models don’t: the sentiment reading in the panel right now is carried forward from May. This month’s print hadn’t landed when the signal was struck.

The model didn’t guess. It didn’t substitute a proxy. It held the last value actually observed and marked it as carried forward. A reading that is late is not the same as a reading that is bad — and pretending otherwise is how models quietly start lying to the people using them.

Friday morning closes that gap, on strike day.

The part that isn’t on the calendar

None of it is what actually holds this allocation up.

The panel clears Stage 4 with a category to spare, so no single gauge is pivotal. The price trend is a different story. An S&P close below its ten-month line cuts two stages at once — 100% equity to 33% — whatever the panel says.

It’s the only lever with that kind of power, and it’s exactly the one that moved this model in March. This week’s data can broaden the read. It can’t break it.

The asymmetry

Lose one positive category and nothing happens — still Stage 4, still 100%. Lose two and it’s Stage 3 at 67%, and categories don’t tend to flip in pairs.

An S&P close below the ten-month line: Stage 2, 33%, immediately. One condition, two stages, no panel involvement.

That’s the entire risk profile in one line. Data improves the read at the margin. Only the trend breaks it.

The four levels

• The ten-month line. Sole downgrade trigger. Fires within three months of a Stage 4 read only 20% of the time — but it fired in March.

• Consumer at 33%. Two indicators from flipping, both refreshing this week. Upside case, not a rescue.

• Diffusion at 60%. Ten points of cushion. A slip to 50% shuts Geography’s gate entirely. The CLI panel runs a month behind, so this one comes with visible lead time.

• The momentum race. International ahead by ~6 points. Geography holds only its winner, so a reversal swings 100% of the position. Six points is not wide.

The scenario that’s underpriced

Not the bear case. The round trip.

44% of stage changes reverse within two months. March to April is the live example — trend break, cut to 33%, recovery, back to 100%. Two trades, no net position change, real cost.

If the trend breaks off Friday’s strike, the base case isn’t “the cycle is turning.” It’s closer to a coin flip on whether the position goes back on by October. Size the reaction accordingly.

The execution date

Signals strike at month-end, execute at the next months open. Friday’s strike is a Monday, August 3 open.

Thursday and Friday’s data set the signal. Nothing trades until Monday. Anyone moving between Friday’s close and Monday’s open is making a discretionary bet and calling it systematic.

Bottom line

Fully invested, fully international, on a four-of-seven panel with a one-category cushion and a six-point momentum edge. Conviction moderate: posture right, support thin.

The trigger still required before anything changes is a single S&P close below the ten-month line. Everything landing this week can broaden this read or leave it alone. Only that one line can break it.

Atlas Report 2026 07 27

626KB ∙ PDF file

Download

Download

Every crude and product tanker pure play in the universe carries a CORE label this week. Median base +0.90, median trend +1.15, median ROIC 20.1%, median revenue growth +77.5%. No other sub-theme in 531 names produces a clean sweep.

That should be a signal. It’s closer to an artifact.

Hormuz has been effectively closed since 2 March. Transits are running near 11% of normal. TD3C has gone from $117k/day in February to $369k/day on 20 July, the highest Middle East VLCC prints in a series that starts in 2005. Two mechanisms, neither requiring incremental oil demand: war-risk premia on vessels that transit, and tonne-miles absorbed by vessels that route around.

Spot economics are up 215% since February. The operators earning those rates are up between −5% and +20%, and four of the seven peaked in May or June and have drifted since, over the exact weeks the prints kept setting records. The FFA curve prices normalisation by Q3.

A fleet earning $300k/day prints +100% revenue growth and 70% gross margins with no change in vessel count, competitive position or cost structure. The score is picking up the freight print, not a franchise. Seven CORE labels are one condition observed seven times.

The tape says it sharper. Split the group on long-run FCF quality: the low-durability basket is up 8.2% since the closure against 2.6% for the structural basket. The two highest RS readings, INSW at 74 and NAT at 72, are both low durability. The lowest, STNG at 51, has the best ten-year FCF quality in the group. A durable re-rating produces the opposite ordering. This is capital buying freight beta and declining to pay up for franchise, a market treating the earnings as rented.

Which means the invalidation isn’t a cycle, it’s a headline. Freight, crude and the equities reprice together in one session, and a stop-loss does not function as risk control on a gap.

Both directions carry risk. Underwrite record freight as a run-rate and the Strait reopens toward $45k/day. Stay absent and the closure runs another two quarters, each throwing off substantial free cash against modest enterprise values, and the names that ignored six months re-rate fast. Congressional analysis recorded a prior consensus that US forces could restore shipping flow. At 142 days, they have not.

The only bloc where base, trend and price setup align at once and the most fragile thesis in the universe. One position, sized for a gap.

Full brief attached: durability routing on all seven, the invalidation table with trigger levels, and the cleanest single expression on the framework’s own terms.

Mmr Tanker Brief 2026 07 21

166KB ∙ PDF file

Download

Download

Prices and figures as of the July 16, 2026 close. AMZN $249.89.

The setup nobody wants to hold

Here is the sentence that defines Amazon right now:

Clean EPS rises every single year from FY25 to FY28 ($6.3 → $7.5 → $9.4 → $11.8) while free cash flow goes to negative $25 billion this year.

That is not a contradiction. It is an accounting fact about a company spending $200 billion on capex it depreciates over five to thirty years. Depreciation is non-cash. Capex is cash. Earnings look through the build; cash flow eats it.

Which means the market is being asked to own an earnings inflection before the cash inflection prints. That is a harder thing to hold than it sounds, and the tape tells you the market hasn’t fully agreed to do it: AMZN is up 10% over twelve months, up 0.6% over three, sitting below its 50 day and 9.1% off its high, while META and AAPL put up +12% in the last month alone.

Amazon leads neither its mega-cap cohort nor the AI trade it is levered to.

That is the whole post. The business is inflecting. The chart is just starting. The debate is timing.

What actually happened at AWS

The number that should have moved the stock and didn’t:

AWS grew 28% year over year in Q1 2026 (the fastest in fifteen quarters) at a 37.7% operating margin.

The five-quarter sequence: 17% → 17% → 20% → 24% → 28%. That is not noise. That is a reacceleration off a $150bn run rate, from the largest cloud in the world at roughly 30% global share. Twenty-eight percent growth on that base adds more absolute revenue dollars than either faster-growing rival does off a smaller one.

Underneath it:

• Custom silicon inflecting. Trainium and Graviton crossed a $20bn run rate. Trainium2 is sold out.

• Contracted, not speculative. AWS remaining performance obligations reached $364 billion. OpenAI committed ~2GW from 2027. Anthropic up to 5GW.

• AI monetization is real revenue. An AI services run rate above $15bn growing triple digits; Bedrock customer spend up 170% quarter over quarter.

• Consolidated operating margin hit 13.1% record!

Management’s framing is explicit and demand-led: they are monetizing capacity as fast as they can install it.

And here is the bill

Free cash flow collapsed to $1.2 billion trailing twelve months, from $38 billion. Capex hit $147bn, up 67%. Long-term debt jumped from $66bn to $119bn in a single quarter, a $53bn raise to fund infrastructure.

Then the harder problem: AWS at +28% is still the slowest-growing of the big three. Azure is running 40%. Google Cloud 63%. Amazon is reaccelerating and still ceding growth-rate leadership, particularly at the AI high end.

So the bear case isn’t that the business is bad. It’s that Amazon is spending the most, growing the slowest of the three, levering up to do it, and asking you to wait until FY27–28 to find out whether the capital earned its return.

Owner earnings: the honest number is neither one you’ll see quoted

Reported FCF says Amazon generates roughly nothing. That overstates the distortion, it charges the entire AI growth build against a business whose existing revenue base doesn’t need it.

The common fix is gross owner earnings: operating cash flow minus D&A. That gets you $78bn today rising to $150bn by FY28, and a headline 19x FY28E. Cheap!

It’s also the optimistic bound, and it cheats twice: it assumes servers need no more replacement than today’s depreciation of a smaller historical base, and it treats $20bn a year of stock-based comp as free.

MMR anchors to the conservative line instead, operating cash flow, minus maintenance capex proxied at 1.3× D&A, minus SBC:

Owner earnings conservative: $37B - $90B

Gross owner earnings (upper bound): $78B-$150B

On enterprise value of $2.84tn (including $119bn net debt, not market cap), conservative owner earnings put AMZN at 41x FY27E and 32x FY28E.

Roughly in line with the earnings multiple. Fair, not cheap.

The reported picture overstates the distortion. The gross picture overstates the value. The line between them is the honest read: a genuine 2.4× inflection over four years that supports an A− business grade and does not make Amazon a bargain.

And one assumption is pivotal for all of it: five-to-six-year server lives. If AI accelerators economically obsolete on Nvidia’s roughly annual cadence — two to three year lives — then lives get cut or silicon gets written down, maintenance capex and depreciation both jump, and conservative owner earnings fall back toward reported FCF. That single assumption is the difference between 32x and un-ownable. It is not a footnote. It is the thesis.

What kills it

Not vague risk language. Specific conditions that end the thesis:

1. Server useful-life reset. AWS margins are flattered by 5–6 year depreciation. If AI silicon obsoletes on a 2–3 year cadence, lives get cut or assets written down margins compress, maintenance capex rises, owner earnings collapse toward reported FCF.

2. Capex climbs with no ceiling. Every guide has stepped up: $83bn → $132bn → $200bn, the last coming in $54bn above Street. Management won’t give an FCF floor. A higher FY27 guide slides the inflection a full year right.

3. Backlog quality cracks. The $364bn RPO leans on OpenAI and Anthropic and Amazon funds Anthropic, which buys AWS and Trainium. That circularity means a renegotiation hits revenue and the equity mark simultaneously.

4. AWS growth stalls back toward 20% or below on Jul 30. Breaks the reacceleration thesis outright and confirms the market’s capex skepticism.

5. A close below the 200-day ($234) ends the multi-year uptrend and moves AMZN out of Stage 2. That is the technical line, and it is not a soft one.

Bottom line

AMZN is a Grade C+ setup on a Grade A− business, and the whole trade is the timing of one inflection.

AWS grew 28% at a 37.7% margin. The $200bn capex peaks this year and is demand-led and contracted. It should convert to AWS revenue roughly doubling by FY28 with margin toward 40% and conservative owner earnings compounding from $37bn to $90bn.

But on enterprise value that’s 32x FY28E owner earnings (fair, not cheap) and it rests on 5 - 6 years server lives holding.

What keeps this at C+ is the tape: up 10% over twelve months, flat over three, below the 50-day, still the slowest-growing of the big three.

This is a high-conviction inflection to stalk for a Stage-2 reclaim confirmed by earnings not to chase in the range.

Amzn Research 2026 07 16

370KB ∙ PDF file

Download

Download

That’s the boring version. Here’s the one worth your time.

The Growth Engine sits at Stage 4 Constructive: 4 of 7 gauges positive. But look at which three are negative: Consumer 33%, Liquidity 0%, Housing 0%. Real incomes, sentiment, the monetary base, building permits. Every one of them sits on the same side of the economy. The weakness isn’t scattered. It’s concentrated in domestic demand.

And the panel isn’t what’s holding the position anyway. At 4 of 7, the model could lose a category and still hold Stage 4. The pivotal lever is the S&P’s ten-month line a close below it cuts the allocation two stages at once, 100% straight to 33%, whatever the panel says. It’s exactly what moved the model in March.

So with domestic demand red, price is doing the load-bearing work.

The second thing: Geography is 100% in international ex-US equity, and that is not a bet against America. The U.S. leading indicator reads 101.0 and is rising, ahead of the ten-country average. The tilt is a price-momentum call, by a margin of about six points. Every falling economy in the ten-country panel is European: Germany, France, the UK, Italy. The drag is one continent.

Third, the quiet number. The Economic Momentum Index reads 1.59 against a 15-year average of 2.98. Positive, so the regime is Expansion, running about 46% below normal. A real expansion, just a cooler and later one than the label suggests.

The backdrop still rewards being invested. It doesn’t reward chasing.

The full report is attached: both engines’ readings, the attribution breakdown, forward base rates from a Stage 4 state, the fifteen underlying indicators, the 50/50 blend, and the complete trade log.

This week’s positioning

The one live wire this week: Tuesday’s Philadelphia Fed non-manufacturing survey (8:30 am ET). It’s the only Atlas indicator with fresh data, feeding the services proxy that’s the lone negative inside an otherwise positive Business block. Also worth watching: the ECB Thursday, given where the diffusion weakness sits, and flash PMIs Friday. Fed blackout runs into the July 29 FOMC.

Everything else is in the report.

Atlas Report 2026 07 20

624KB ∙ PDF file

Download

Download

Impersonal market commentary, not personalized investment advice. Model results reflect historical simulation; past performance does not guarantee future results.

POSITION DISCLOSURE — Buy Stop at $75.25, stop close below 73.60

Read only one row of Williams’ financials and you’d short it: free cash flow fell from $3.4B (FY23) to $0.9B (FY25) a 74% collapse on a ~29x forward multiple and 4.3x net-debt/EBITDA. That’s the bear case, and the arithmetic is right.

The why is wrong. FCF didn’t erode; capex nearly doubled to $5B, and it went into the ground on contracted power and behind-the-meter capacity for AI data centers: Socrates (400 MW, Meta), Atlas, NEO. Operating cash flow kept compounding to $6B. When the buildout peaks and rolls toward maintenance (2028), FCF re-expands toward $4.7B normalized. The number that scares a yield investor today is the one that pays them in three years.

Underneath it, the Q1 2026 print already shows the turn: net income +25%, adjusted EBITDA +14% on roughly flat revenue. That combination is the whole thesis: margin expansion, not a gas-price call. Williams isn’t a pipeline-yield name. It’s a contracted natural-gas demand machine, LNG exports, gas-fired power, and hyperscaler power that skips the grid queue being handed to you while its cash flow is deliberately, temporarily suppressed by growth it has already signed.

The stock sits ~6% below its 52-week high on a tight base above rising 50- and 200-day averages, +23% relative strength, and the gas-infra/AI-power complex is the one group rotating in while Semi’s are pulling back.

The full trade card is attached: Twin-Momentum diagnostics, the grade rubric, valuation, and the catalyst calendar.

The re-rate. On base-case cash flow it’s fairly valued at a 5.5% yield on ~$4.7B normalized FCF ≈ $85B equity, roughly today’s cap. The asymmetry isn’t the DCF; it’s the multiple. If the market values the contracted power EBITDA like infrastructure rather than pipeline throughput, FY27E EBITDA $8.6B at 18–20x puts the stock around $100–115 up 35–50%. Street already leans that way (22 buys, $83.5 mean).

What kills it: a marquee BTM contract cancelled or a hyperscaler pullback; capex rising on uncontracted builds; 2026 EBITDA guide cut below $8B; net-debt/EBITDA above 5x; or a decisive break below the base and the rising 50-day. The single thing that actually governs the thesis is the capex mix: contracted vs. speculative. Watch that above all.

Bottom line: a Grade A− setup on a Grade B+ business. Base case fairly-valued-to-modest-upside; the upside is a re-rating of the power EBITDA.

Wmb Research 2026 07 05

202KB ∙ PDF file

Download

Download

Mythic Market Research · informational & educational only · not investment advice · figures as of 2026-07-05.

Two engines run underneath this report. One reads the U.S. business cycle and decides how much equity to own. The other reads the global cycle, uses it as an on/off gate, and lets price decide where to own it.

This week they agree. The Growth Engine sits at Stage 4 Constructive fully allocated, price above its ten-month line. Geography is risk-on, with global diffusion at 60%, and every dollar of its equity is outside the United States.

That’s the surface. The signal worth carrying into the next few months is quieter.

The framework’s broadest gauge, the Economic Momentum Index, reads 1.59. Positive, so the cycle registers as Expansion. But it’s running roughly 46% below its 15-year average of 2.98. Breadth tells the same story: participation across indicators sits at 61 against a norm near 68.

This is a real expansion. It is a cooler and later-cycle one than the headline suggests. Momentum is positive and slowing, not positive and accelerating.

What’s actually holding the read together

The Growth Engine’s panel is 4 of 7 positive. That clears the Stage 4 line with a single-category cushion.

Carrying it: Labor (100%), Global (100%), Markets (75%), Business (75%).

Dragging it: Consumer (33%), Liquidity (0%), Housing (0%).

Look at where the weakness sits. Real incomes. Consumer sentiment. The monetary base. Building permits. Every one of them lives on the same side of the economy; domestic demand and policy liquidity. The weakness isn’t scattered across the panel. It’s concentrated in one place.

And the factor that matters most isn’t in the panel at all. It’s the price trend. No single category is pivotal; the panel could lose a positive and still hold Stage 4. But if the S&P closes below its ten-month line, the model cuts two stages at once, from 100% equity to 33%, regardless of what the panel says.

That is the only lever with that kind of power. It’s also the one that moved the model in March. That drop to Stage 2 was a trend break, not a panel collapse.

With domestic demand already red, the market’s own trend is doing the load-bearing work of keeping this model invested.

Europe is the entire drag

The global signal is a vote across ten economies. Six are rising. But the vote isn’t evenly spread.

Every rising economy is in the Americas or Asia. Every falling one is in Europe (Germany, France, the United Kingdom, Italy). Brazil leads the set at 103.8; South Korea is the fastest riser month-on-month; China is the only economy still below trend at 98.9, but it’s turning up off a low base.

Here’s the part most people will get backwards: the United States reads 101.0 and rising, actually ahead of the ten-country average once Europe’s decline is folded in.

So, the international tilt is not a bet that the U.S. economy is faltering. It’s a price-momentum decision riding on top of a global backdrop that is broadly healthy and notably un-European. Those are different claims, and the difference matters for how long you’d expect the tilt to last.

The week the calendar comes to the model

Three of the fifteen indicators beneath the Atlas panel print on Friday alone: building permits, industrial production, and consumer sentiment. Consumer: the softest gauge on the panel, and the nearest one to flipping, gets both of its swing readings inside 48 hours.

Wrapped around all of it: Fed Chair Kevin Warsh’s congressional debut, House on Tuesday, Senate on Wednesday, with June CPI landing ninety minutes before he takes his first seat. Last week’s minutes showed a committee that has flipped its dot plot from cuts to hikes, nine of eighteen officials now penciling in at least one increases this year.

That is precisely why Liquidity is pinned at 0%, and why it is likely to stay there.

One thing to watch above all others this week: whether the domestic demand side of the panel finds a floor.

The honest version

History offers comfort without offering a promise. From a Stage 4 reading, the following twelve months have been positive 81% of the time, and the stage has slipped within a quarter only 20% of the time. The constructive read has staying power.

But the same base rates carry a worst column, and conditioning on a good state does not remove tail risk. Geography holds a single market on a 6-point momentum edge, and its trend overlay reverses within two months 44% of the time. Concentration and whipsaw are both live.

The backdrop still rewards being invested. It’s the kind of late-cycle setup that asks you to keep one eye on the exits rather than chase.

The full report below is about exactly which exits to watch.

Atlas Report 2026 07 13 (1)

624KB ∙ PDF file

Download

Download

The AI buildout is not a list of stocks. It is a densely wired supply chain in which the same dollar gets counted at four or five different nodes on its way down, and where a handful of names sit on chokepoints that every path has to cross. Map the connections and one fact jumps out: capital enters wide and exits narrow. It starts spread across five hyperscalers and re-concentrates, again and again, onto a very small number of companies.

Three paths, one funnel

The capex splits into three broad destinations. The first is the chip dollar: hyperscalers buy compute silicon, the chip designers pay the foundry and the memory makers, and every one of those fabs pays the equipment makers. The second is the rack dollar: servers, switches, and neocloud capacity, built out of interconnect and optics components. The third is the electron-and-concrete dollar: the power gear, the utilities selling electricity under long-term contracts, and the contractors physically building the sites.

The chip path is the one that funnels hardest. Roughly 90% of accelerator spending routes through a single company Nvidia, and from there, whether the chip is a Nvidia GPU, an AMD GPU, a Broadcom-designed Google TPU, or a Marvell-designed AWS Trainium, it is manufactured in the same place. That convergence is the core of the whole map.

The chokepoints that own the flow

Three nodes capture value not because they win a competition but because there is no way around them.

TSMC is the deepest chokepoint. It holds roughly 70% of global foundry revenue and produces around 90% of the world’s most advanced chips. Every merchant GPU and every hyperscaler custom ASIC is fabricated there. It doesn’t matter which architecture wins the design war: Nvidia, AMD, Broadcom, and Marvell all hand their designs to the same fab. TSMC’s advanced-packaging lines, CoWoS, are sold out through 2027, and Nvidia has reserved the majority of that capacity. Packaging, not wafer fabrication, is now the binding constraint on how many accelerators can physically ship.

HBM memory sits directly underneath. High-bandwidth memory is the single largest bill-of-materials line in an AI accelerator, and only three companies make it: SK Hynix, Micron, and Samsung. SK Hynix holds well over half the market and roughly two-thirds of Nvidia’s next-generation HBM4 allocation; Micron has reportedly sold out its entire 2026 output. The wiring here is subtle and important — TSMC manufactures the logic base die that sits under the HBM stack, so even the memory layer routes back through the foundry. Memory is where the true scarcity lives.

ASML anchors the bottom. It is the only company on earth that makes EUV lithography machines — 100% of that market — and no fab below the leading edge gets built without it. Every capacity expansion by TSMC, by Samsung, by the memory makers, cascades down into ASML orders. It is the most defensible position in the entire chain precisely because it is the furthest from the end customer: a monopoly on the tools everyone else needs.

The pattern is consistent. The deeper you go and the narrower the field, the more durable the position — and the equipment layer as a whole (ASML, plus Applied Materials, Lam Research, and KLA) captures that durability. The correction worth stating plainly, because it’s easy to get wrong: every memory fab: Micron, SK Hynix, Samsung — and TSMC itself buy from both Lam and Applied Materials, not one or the other. Lam draws roughly 35–45% of its revenue from memory, and Micron is named among its most significant customers in its filings. These are ongoing purchase relationships, not one-off contracts, and HBM4 demand is now an explicit growth driver for them.

The hub that captures the dollars and the duopoly eroding it

Nvidia is where the absolute dollars land. Fiscal-2026 revenue ran around $216 billion, the overwhelming majority from the data center, and its order book stretches into the hundreds of billions. But the more revealing story is defensive: Nvidia has vertically pre-bought the entire chain to protect its own supply. It reserved roughly 60% of CoWoS packaging, secured about two-thirds of HBM4, put a combined $4 billion into Lumentum and Coherent to lock optical supply, and reserved gas-turbine capacity for its customers’ power. It isn’t just selling chips; it’s buying up the chokepoints so no one can starve it.

The countermove is the custom-ASIC duopoly. Broadcom and Marvell design the hyperscalers in-house silicon. Broadcom behind Google’s TPU, Meta’s MTIA, and OpenAI’s chip; Marvell behind AWS’s Trainium and Microsoft’s Maia. This is the part of the map that isn’t “everyone buys Nvidia,” and it’s where the hyperscalers’ governance instinct to avoid single-vendor dependence turns into real revenue. Tellingly, Nvidia itself invested $2 billion in Marvell to integrate its custom silicon with Nvidia’s interconnect. The hub co-opting its challenger rather than fighting it.

The connective tissue

The most interesting names are the ones that bridge two layers, because a single position gives multi-layer exposure and because a shock at one propagates through them.

ARM licenses the CPU cores that sit inside Nvidia’s Grace and Vera chips, Amazon’s Graviton, and Google’s Axion. It’s one name with a foot in nearly every compute program. Amkor and ASE absorb the CoWoS overflow that TSMC can’t handle, bridging packaging and compute. On the rack side, Broadcom appears twice — its switch silicon powers Arista’s networking gear as well as its own custom-ASIC book. As GPU clusters scale past a hundred thousand accelerators, copper runs out of reach and the interconnect-and-optics layer becomes its own bottleneck: Astera Labs and Credo on the electrical side, Coherent and Lumentum on the optical side, Fabrinet assembling the modules, Corning supplying the fiber.

The branch growing fastest

The power and build path never touches a chip, and it’s the branch accelerating hardest as electrons become the true constraint. GE Vernova’s gas-turbine order book is full through 2029 with reservations into 2031; Vertiv and Eaton are carrying multi-year backlogs; Constellation, Vistra, and Talen are signing 20-year nuclear power-purchase agreements directly with hyperscalers; and the construction names Sterling, Quanta, Comfort Systems are posting record backlogs. This branch is flat rather than cascading, because power and concrete don’t have a deep supplier chain the way silicon does. The money fans out and gets spent.

What the interconnectivity actually means

Two conclusions fall out of the map.

First, the biggest beneficiaries are not the same as the biggest winners. Nvidia captures the most absolute dollars, but its share is the one under attack from the ASIC duopoly. The most durable value sits deeper at TSMC, at the three memory makers, and at ASML where there is no second source and every path converges. Those are the nodes that get paid regardless of which chip, which cloud, or which model wins.

Second, and this is the risk hiding inside the opportunity: because the same dollar is counted at multiple nodes, many of these names are not independent bets. Nvidia, TSMC, and the HBM makers are the same flow measured at three points. A shock to CoWoS, to HBM, or to EUV supply hits all of them at once. On the map, that shows up as a dense link bundle and any two names sharing one should be treated as a single position, not two.

The buildout looks like a broad, diversified theme. Traced through its connections, it is a narrow funnel wearing a wide hat.

Structural analysis of publicly reported supply and contract relationships; figures are approximate and drawn from company filings, Goldman Sachs research, and industry reporting. Not investment advice.

Tree 1: the chip dollar. Hyperscalers buy compute silicon; the chip designers pay TSMC, the packaging houses, and the memory makers; and all of those fabs pay the equipment makers at the bottom.

Where each one lands in Tree 1: NVDA and AMD are merchant GPUs the hyperscalers buy off the shelf; AVGO and MRVL are custom-ASIC design partners the hyperscalers co-design with (AVGO does Google’s TPU, Meta’s MTIA, and OpenAI’s chip; MRVL does AWS’s Trainium and Microsoft’s Maia). ARM sits beside them as the CPU-core IP licensor — royalties flow up to ARM from NVDA (Grace/Vera) and the hyperscalers’ own CPUs (AWS Graviton, Google Axion), though not from AMD, which is x86

Tree 2: the rack dollar (AI infrastructure + optics). Hyperscaler capex buys the servers, switches, and neocloud capacity; those systems are in turn built out of the interconnect and optics components on the bottom row.

Where each lands in Tree 2: DELL and SMCI are the branded server OEMs, CLS the contract manufacturer that builds racks and switches, ANET the switch vendor (which itself runs on AVGO silicon), and CRWV the neocloud that buys GPUs wholesale and re-rents them. The bottom row is what physically goes inside those boxes: AVGO’s switch silicon, ALAB’s retimers and CRDO’s active cables that wire GPUs together, and COHR/LITE’s optical transceivers and lasers that carry traffic between racks — with FN doing the optical assembly and GLW the fiber.

Tree 3 — the electron and concrete dollar. This branch never touches a chip. Hyperscaler capex flows directly to three parallel destinations: the power and cooling gear, the utilities selling power under long-term contracts, and the contractors physically building the sites.

Where each lands in Tree 3: GEV sells the gas turbines (generation equipment), VRT the in-hall power distribution and liquid cooling, ETN the electrical switchgear — these are equipment vendors the hyperscalers pay directly. CEG, VST, and TLN are the power producers that sign the long-term PPAs (mostly nuclear) to actually supply the electrons. STRL does site development, PWR builds the grid interconnect and substations, and IESC/MYRG are the electrical construction contractors. Unlike Trees 1 and 2, this one is flat — the money fans out in parallel, it doesn’t cascade, because power and concrete don’t have a deep supplier chain the way silicon does.

So the full flow of capital, in three clean paths: the ~$740B splits into the chip dollar (Tree 1, which cascades four levels deep and re-concentrates on TSMC and the memory makers), the rack dollar (Tree 2, systems built from optics and interconnect), and the power/build dollar (Tree 3, paid straight out to gear, utilities, and contractors).

Where each lands in Tree 3: GEV sells the gas turbines (generation equipment), VRT the in-hall power distribution and liquid cooling, ETN the electrical switchgear — these are equipment vendors the hyperscalers pay directly. CEG, VST, and TLN are the power producers that sign the long-term PPAs (mostly nuclear) to actually supply the electrons. STRL does site development, PWR builds the grid interconnect and substations, and IESC/MYRG are the electrical construction contractors. Unlike Trees 1 and 2, this one is flat — the money fans out in parallel, it doesn’t cascade, because power and concrete don’t have a deep supplier chain the way silicon does.

Position disclosure: Long PENG

Start with the print, because the print is not in dispute.

Penguin Solutions reported FQ3 FY2026 on July 7. Net sales of $478.7 million, up 47.6% year over year. Non-GAAP EPS of $0.84, up 79%, a fourth consecutive beat. Management raised full-year guidance above the prior high end, to roughly +22% sales and $2.60 of non-GAAP EPS. GAAP operating margin went from 3.0% to 10.6% in a single year.

And here is the number that made us look twice: GAAP operating expenses fell 3.4% year over year against 47.6% revenue growth. That is not a rounding artifact. That is a company that grew half again as large while spending less to run itself.

On our eight-quarter Twin-Momentum framework, Penguin’s base score (FBS) is +2.16, the strongest of any AI-infrastructure name we track. Stronger than Micron’s. Stronger than Marvell’s.

The stock closed over +20% on 2.5x volume, at a 52-week high, up 279.6% in three months.

The question is the one that keeps showing up in our work on every parabolic name: what, exactly, are you buying and at what price?

What you are actually buying

Split the revenue line and the story changes.

Segment FQ3’26YoY

1. Integrated Memory: $275.1M +111%

2. Advanced Computing: $137.6M +3.8%

Integrated Memory is now 57% of sales, up from 40% a year ago. Advanced Computing the AI-server business, the segment the entire “AI infrastructure” framing rests on, the segment NVIDIA blessed with an invitation-only AI Factory Specialized Partner designation on June 23 grew 3.8%.

The growth is memory. Penguin buys DRAM and sells modules. When DRAM prices rise, Penguin’s revenue line rises with them and its gross margin thins (right now barely), because it is paying up for the input.

You can watch it happen in the margin rows:

Gross margin is flat-to-down across eight quarters 28.0% to 27.8%, with the company itself guiding FY26 to 26.5%. Operating margin nearly quadrupled over the same window.

Those two lines are the entire thesis. Every point of margin expansion came out of the expense line, not out of price or mix. Micron captures this same DRAM cycle as an 84.9% non-GAAP gross margin. Penguin captures it as revenue.

That distinction is not academic. Cost discipline is real and displayed in there earning release, and it is potentially finite. You cannot cut opex 3.4% a year forever.

The bull case, stated fairly

• The operating leverage is genuine. Revenue +47.6%, opex −3.4%. Quarterly ROIC went from 1.0% to 9.1% across the eight-quarter window.

• The base is real. FBS +2.16 on eight clean GAAP quarters is the best structural base in the AI-infra complex, and we compute it the same way for every name we cover.

• Integrated Memory is direct exposure to the AI memory supercycle the same cycle driving Micron, expressed through a $4 billion market cap instead of a trillion-dollar one.

• The NVIDIA designation is a real catalyst landing into a float that is roughly 20.7% short with 3.4 days to cover on average.

• The balance sheet is not the risk. Net debt is roughly $20 million against $440.3 million of cash. Whatever else goes wrong here, it is not a solvency story.

• Street is stale. Consensus was set pre-print, at $1.58B FY26E sales and $2.28 non-GAAP EPS, both below the company’s own raised guide. Sharp upward revisions are coming.

The bear case, which is mostly about the tape

• It is a memory trade, not an AI-server trade. 57% of sales, +111%, versus +3.8% for the segment that justifies the multiple.

• Gross margin is compressing 27.8% versus 29.3% a year ago, guided to 26.5%. Rising DRAM inflates the revenue line and squeezes the margin simultaneously.

• Cash is not following earnings. Receivables +128% and inventories +95% since FY25 year-end, against 48% sales growth. Cash fell from $453.8M to $440.3M over nine months while GAAP net income was $87.4M. That is the single most uncomfortable table in the filing.

• The extension is extreme. 43% above the 50-day. 165% above the 200-day. 279.6% in three months. An 11.7% average daily range and a beta of 3.34.

• Nothing published supports the price. Pre-print consensus target: $39.86. That is 49% below spot. The Street could have been blind to the Memory business.

The rotation problem

Here is the part most people covering this print will skip.

Our AI-infrastructure basket (MU, MRVL, DELL, VRT, SMCI, CLS) equal-weighted is −9.3% over one week and −4.5% over one month against +77.7% over three months. The short window has fallen below the long window. That is the rotating-out signature. Micron is −18.2% this week. Marvell is −21.6%.

Penguin is the exception, +3.0% on the week, on its own catalyst.

And Penguin is not a diversifier from that book. Its 126-day daily-return correlation is 0.66 to the basket and 0.64 to Marvell. It is a higher-beta expression of the same trade. For position-sizing purposes, if you own the AI-infra sleeve, Penguin is not a new position it is more of the one you already have.

What the price requires

Take the raised guide at face value: $1.67B in FY26 sales, $2.60 of non-GAAP EPS. Our FY27 model sits above Street at $1.98B and $3.05 because the guide implies roughly $505M of FQ4 net sales, a $2.0B annualized exit rate, and Street’s FY27 therefore requires a sequential decline from that exit. We do not model that.

But we also do not extrapolate the Q3 growth rate, because of mix. So:

• 25x FY27E MMR EPS → $76. Roughly spot.

• 35x → $107.

• Bear: DRAM rolls, FY27 non-GAAP EPS falls to ~$2.00 → $50 at 25x.

At $78.26, the market is already paying full mid-cycle value for our above Street numbers. There is upside if the multiple expands into a re-rating. There is no margin of safety if it does not.

Note what the bear case is and is not. It is a valuation case, not a solvency case. Penguin at $50 is a company with $440 million of cash and a working business. It is simply a stock that gave back a third of a 280% move.

What would invalidate the thesis

Written before the fact, as always:

1. DRAM price rollover. Integrated Memory is 57% of sales and its 111% growth is substantially price-driven. A downcycle deflates the revenue line and marks down $498.3M of inventory at the same time. Same event, twice.

2. Gross margin breaks below the 26.5% FY26 guide — evidence the memory mix is dilutive, and operating leverage no longer offsets it.

3. Cash conversion fails to normalize. Another quarter of receivables outgrowing sales. This is a big red flag.

4. Advanced Computing stays flat. The AI-infrastructure thesis and the NVIDIA designation requires that segment to inflect. +3.8% does not.

5. Technical mirror: failure to hold the pre-earnings close of $62.71 would say the +20% gap was a squeeze against a 20.7% short float, not a re-rating.

The calendar: FQ4 and FY26 results around October 6. The raised guide implies roughly $505M of Q4 net sales up 50% year over year and about $0.90 of non-GAAP EPS. That is a high bar, and the company set it for itself. Between now and then, the single most important input is DRAM contract pricing.

Bottom line

In our framework, Penguin Solutions is a Grade C+ setup (Now Extended) on a Grade B− business and the split is entirely about price, not quality.

The quarter was excellent. Record sales, a fourth straight beat, a raised guide, operating margin from 3.0% to 10.6% while operating expenses actually fell, and the strongest eight quarter momentum base in the complex. That is real.

Three things temper it. The growth is memory, not AI servers, and it arrives at a thinning gross margin. Cash is not following earnings.

Watch DRAM pricing. Watch whether Advanced Computing inflects behind the NVIDIA designation. Watch whether receivables normalize.

The fundamental base is real. We are watching to see if the fundamental trends continue.

Peng Research 2026 07 08 (1)

203KB ∙ PDF file

Download

Download

Mythic Market Research · for informational and educational purposes only · not investment advice or a solicitation · do your own research.

This is a Mythic Market Research note. For informational and educational purposes only — not investment advice or a solicitation. Do your own research.

There’s a version of the Micron story that reads like a fantasy pitch deck.

Revenue up 345% year over year. Gross margins of 85%, up from 39% a year ago. Diluted quarterly EPS of $24.67, versus $1.68 the same quarter last year. Operating cash flow of $25.4 billion — in a single quarter. Return on equity of 66.6%. A balance sheet with $30 billion in cash and almost no debt. And management guiding next quarter to $50 billion in revenue and $31 in EPS, then more or less shrugging and saying they can only fill about half to two-thirds of the demand in front of them.

You do not see numbers like this from a $1 trillion company. And yet the stock, at ~$938, trades at 12.9x this year’s earnings.

So the interesting question about Micron isn’t “is this a good company?” The answer to that is obvious, and it’s yes. The interesting question is the one the market has been quietly arguing about for weeks: what multiple do you pay for peak earnings in a business that has never once, in its entire history, avoided the downside of the cycle?

The bull case: this time the moat is real

Micron has always been a memory company, and memory has always been a commodity. Boom, bust, repeat. What’s different this cycle is one three letter acronym: HBM — the high-bandwidth memory stacked next to every AI accelerator on earth.

There are exactly three companies on the planet that can make it: Micron, SK Hynix, and Samsung. Micron’s 2026 HBM supply is sold out and largely contracted, and it’s been shipping HBM4 for NVIDIA’s Vera Rubin platform since March. Hyperscalers have earmarked north of $725 billion in AI data-center capex for 2026. When the people building the data centers commit that kind of money, the memory that goes inside them stops looking quite so commodity-like.

The most important structural wrinkle: Micron has signed 16 Strategic Customer Agreements representing about $22 billion in binding volume commitments. That’s roughly half of revenue moving from spot-priced-and-cyclical toward contracted-and-visible. If you believe the cycle is genuinely different this time, that’s the receipt.

The cash-flow transformation is the part that’s hard to argue with. Free cash flow went from negative $6.1 billion in FY23 to $1.7 billion in FY25 to a projected $25 billion this year and $75 billion next. Micron is spending at record levels to build capacity, stated that capex costs are increasing, and it’s still generating enormous cash. That inflection is the whole story.

The bear case: it’s still a memory cyclical

Every prior memory upcycle ended the same way. Micron, SK Hynix, and Samsung all raced to add capacity, supply tipped into surplus, and record margins mean-reverted with brutal speed. The mechanism that’s minting money right now, pricing power from tight supply — is the exact same mechanism that reverses.

And the setup, right now, is stretched. The stock is up roughly 700% in a year, trading around 2.3x its own 200-day moving average, ~25% off its high and visibly digesting a parabolic move. The memory theme as a whole fell 18% in a single week, cooling, not breaking, but not a spot to chase either.

Then there’s the capex itself, which cuts both ways. Record spending builds the capacity that satisfies today’s shortage and if pricing rolls over, that same capex makes free cash flow fall as fast as it climbed. The risk is symmetric.

The debate reduces to one number

Here’s where it all collapses to a single question. FY2026 is essentially locked, three quarters reported plus a Q4 guide landing Micron near $129 billion in revenue and $73 in EPS at record margins. Nobody’s really arguing about this year.

The fight is FY2027. Street models revenue nearly doubling again, to $237 billion, on continued HBM and DRAM tightness. Our house model is more conservative: $200 billion in revenue and $120 in EPS, crediting the contracted agreements and sold-out 2026 supply but allowing for pricing to normalize as all three makers add capacity into 2027–28.

Now watch what the multiple does to that:

• Pay a cyclical 8–10x on $120 of FY27 EPS and you get a stock around $1,200.

• Pay a structural 15–18x and you get something closer to $2,160.

For reference, the Street’s average target sits at $1,486 about 58% above the current price with a high of $2,200. The entire bull-versus-bear argument, hundreds of billions in market value, comes down to whether the market decides to treat Micron’s earnings as durable or as a peak.

Bottom line

Micron is, in our framework, a Grade A business on a Grade B setup. The company is extraordinary; record margins, a sold-out and increasingly contracted order book, one of only three seats in the HBM oligopoly, and $22 billion of agreements that genuinely dampen the historical cyclicality.

But the stock has already run ~700%, sits well above its rising averages, and is digesting a parabola with earnings being the peak, aka new failure leading to the theme cooling. That’s not a knock on the business. It’s a comment on the entry.

What I’m watching from here: a proper higher-low base forming rather than a chase of the parabola; DRAM and NAND contract pricing, which is the classic cycle tell; and the first real FY2027 commentary when Micron reports Q4 on September 23 a print that, with an average daily range near 8%, is going to move and give us more insights into demand and capex costs.

Extraordinary company. Respect the extension and be patient. The high-level information we focus on is in the trade card below.

Mu Research 2026 07 05

198KB ∙ PDF file

Download

Download

Mythic Market Research · for informational and educational purposes only · not investment advice or a solicitation · do your own research.

This week, both say lean in. The Growth Engine is fully invested; Geography is risk-on and holding 100% international — with the entire drag on the global signal coming from Europe, where all four major economies are rolling over while the rest of the world climbs. But underneath the green, the U.S. domestic engine — consumer, housing, liquidity — is entirely red, and the expansion’s momentum is running well below its 15-year norm. A real expansion, just a late, cooler one.

Inside: what’s actually holding the constructive read together (it’s not what you’d guess), the strongest and weakest economies in the OECD panel, the base-rate odds for the next twelve months, and the one signal that would flip the whole thing defensive. Check out the report below!

Atlas Report 2026 07 05

637KB ∙ PDF file

Download

Download

Hey, there!

Check out our August performance!

August turned out to be a bit of a rollercoaster. We started the month on solid footing, quickly gaining 1.62% by August 3rd and pushing a little higher into mid-month, with performance peaking on August 13th. From there, markets grew increasingly cautious ahead of the Jackson Hole announcements. That Friday provided a welcome boost for us, as we captured a positive trend day on the 22nd. But after that, conditions reverted back to choppiness, with tight ranges and limited follow-through, making it difficult for our style of trading.

The Mythic Zoo Portfolio has continued to be the cornerstone of our performance this year, keeping the overall book firmly in positive territory. It’s hard to imagine running without it. Relying solely on Mythic Wand OG would have left us far more exposed to periods of underperformance. To put things in perspective, Wand OG delivered an impressive 84.65% return last year, but on its own it doesn’t provide the same balance or resilience. By contrast, the Zoo’s diversified structure has allowed us to increase our position sizing in each portfolio while cushioning drawdowns.

We’ve also been deliberate about position sizing across the portfolio. Our allocations are aggressive by design, which means we fully expect a higher level of day-to-day volatility. It allows us to capture outsized moves when our systems are in sync with the market, even if it means living with more noise in the short term.

Consolidated Portfolio Performance 1-Year: +31.91%

Consolidated Portfolio Performance YTD: +1.56%

1. Wand OG Performance for YTD: -9.59%

2. Mythic Zoo Performance for YTD: +17.83%

Consolidated Portfolio Performance Since Inception: +67.78

Affiliates

Chart VPS Alpha: The only VPS we trust with our trading, see our customer story!

We love it because it's robust, adaptable, and brutally honest. But it’s not for everyone.

The psychological toll is real. Trend strategies can go through long periods of stagnation, drawdowns, and underperformance—especially in sideways markets. Most people can’t handle that. They bail too soon, tweak too often, or lose faith entirely. That’s why we believe trend following will continue to work: not because it's easy, but because it's hard and most traders simply don’t have the discipline to stick with it.

Core Principles of Trend Following

Trend following is based on the premise that markets tend to move in persistent directions over extended periods. The goal is to identify and ride these upward or downward trends across multiple asset classes. Markets are messy. They’re driven by fear, greed, central banks, geopolitics, memes, and momentum. Trying to predict what happens next is tempting—but often futile. Trend following flips the script: it doesn’t predict, it reacts.

It’s a rules-based strategy that thrives when prices move in sustained directions. Think oil in the 2000s, bonds in the '80s, tech stocks in the 2010s, or Bitcoin in 2020. When markets move, trend followers don’t argue—they ride.

This approach has been around for decades, and despite market changes, it’s still quietly doing its job.

Key Characteristics

• Momentum-based: Trend followers buy assets showing upward momentum and sell those with downward momentum 1.

• Reactive approach: Rather than predicting future price movements, trend followers react to existing market trends 1.

• Diversification: Strategies are applied across many markets simultaneously to enhance consistency 2.

• Risk management: Position sizing and stop-loss orders are crucial for managing downside risk 3.

Common Indicators

Trend followers often use technical indicators to identify and confirm trends:

• Moving Averages: Simple or exponential moving averages help smooth out price data.

• Breakout Systems: Identifying when prices break above or below critical levels.

• Momentum Indicators: Rate of Change, all-time highs/lows, and others to gauge trend strength.

  • Here’s an example of a simple and creative momentum-based trend-following system: https: //www.followingthetrend.com/2014/03/free-trend-following-trading-system-rules/

Historical Development

The trend-following strategy can be traced back to the saying "Cut short your losses, and let your profits run." David Ricardo (19 April 1772 – 11 September 1823) was an English political economist. The trend-following approach gained prominence in the early 20th century, with notable figures like Richard Donchian pioneering systematic approaches in the 1960s.

1970s-1980s: The Turtle Traders

In 1983, famous commodities trader Richard Dennis conducted an experiment to prove that successful trading could be taught. He trained a group of novices, dubbed the "Turtle Traders," in trend following techniques. Their remarkable success brought widespread attention to the strategy 3.

1990s-2000s: Expansion and Sophistication

As financial markets became more accessible and data-driven, trend following strategies evolved:

• Increased computing power: Allowed for more complex algorithms and faster execution.

• New markets: Expansion into emerging markets and new asset classes.

• Risk management: Development of more sophisticated position sizing and portfolio construction techniques.

Performance Across Decades

Trend following has consistently delivered positive returns over decades—often with low correlation to traditional asset classes like stocks and bonds.

To illustrate this, I pulled data from IASG, a free database of Commodity Trading Advisors (CTAs), and built an equal-weighted portfolio using publicly available returns. The result isn’t a backtest or optimized model—it’s a simple demonstration of how diversified trend-following strategies have performed over time.

1970s-1980s: The Golden Age

This period saw exceptional returns for trend followers, particularly in commodities markets. Factors contributing to success included:

• High inflation leading to strong trends in commodities.

• Less efficient markets with fewer participants.

• Limited access to information creating longer-lasting trends.

1990s: Mixed Results

The 1990s presented challenges for trend followers:

• Stock market boom: Equities showed persistent uptrends, benefiting some strategies.

• Reduced commodity volatility: Made trend identification more difficult in these markets.

• Increased competition: More traders adopting similar strategies led to crowded trades.

2000s: Resurgence

The decade saw a revival in trend following performance:

• Dot-com bubble burst: Provided opportunities in downward trends.

• Commodity super-cycle: Strong trends in oil, metals, and agricultural products.

• 2008 Financial Crisis: Trend followers profited from sharp declines across multiple asset classes.

2010s-Present: Adaptation and Evolution

Recent years have presented both challenges and opportunities:

• Low interest rate environment: Reduced trends in fixed income markets.

• Increased market efficiency: Shorter-lived trends in some asset classes.

• New asset classes: Cryptocurrencies have provided new trending markets.

• Machine learning: Integration of AI/ML techniques to enhance trend identification and execution.

Modern Trend Following Tactics

Today's trend-following strategies often incorporate advanced techniques to improve performance and adapt to changing market conditions. Traders analyze trends across multiple timeframes to confirm signals and reduce false positives. For example:

• Long-term trend: Identified on weekly or monthly charts

• Entry timing: Fine-tuned using daily and intraday charts

• Dynamic lookback periods: Adjusting moving average lengths based on market volatility

• Volatility-based position sizing: Increasing or decreasing exposure based on recent market behavior

Crowded Trades

As more capital has flowed into trend-following strategies, some argue that opportunities have diminished due to increased competition. Trend-following can experience extended drawdowns, particularly in choppy or range-bound markets. This can test the patience of investors and lead to strategy abandonment.

Conclusion

Trend following has demonstrated remarkable longevity as a trading strategy, adapting to changing market conditions over decades. While past performance does not guarantee future results, the core principles of trend following remain relevant in today's financial landscape:

1. Riding momentum

2. Managing risk

3. Diversifying across markets

Successful trend following requires discipline, robust risk management, and market evolution. As financial markets continue to change, trend followers who can adapt their tactics while staying true to these fundamental principles are likely to find continued success.

Hey, there!

Check out our July performance!

With July now behind us, we’re releasing this update a bit later than usual. The market backdrop has not shifted meaningfully over the summer: the bullish trend remains intact, volatility is subdued, and daily ranges have stayed compressed. Even today, with the market ( ) down around 70 bps, the move only underscores how rare larger range days have been. This environment continues to challenge our style of trading even though July was a positive month for us.

Consolidated Portfolio Performance 1-Year: 46.17%

Consolidated Portfolio Performance YTD: +8.72%

1. Wand OG Performance for YTD: +3.05%

2. Mythic Zoo Performance for YTD: +16.81%

Consolidated Portfolio Performance Since Inception: +79.62%

Hey, there!

This summer has been characterized by a persistent bullish trend, low volatility, and tight daily trading ranges. Historically, this type of environment tends to be one of our most challenging. Often leading to choppy price action.

Earlier in the year, we capitalized on periods of elevated volatility, particularly on the short side, where we captured significant profits during swift downside moves. That more dynamic environment played to our strengths. The daily uncertainty and news announcements played to our strengths.

Long setups, however, have remained a consistent struggle throughout the year, April 9th being an exception. Even when markets have pushed toward new highs, they’ve frequently failed to hold gains, pulling back intraday and leaving little room for follow-through. This has made it difficult to build sustained momentum on the long side, despite the broader market's upward trajectory.

Mythic Zoo has been the better-performing portfolio YTD thus far, while Wand OG has lagged.

Consolidated Portfolio Performance 1-Year: +62.04

Consolidated Portfolio Performance YTD: +7.34%

1. Wand OG Performance for YTD: +3.12%

2. Mythic Zoo Performance for YTD: +13.12%

Consolidated Portfolio Performance Since Inception: +77.32

Hey, there!

Markets giveth, and markets taketh away — and May was a month of giveback for both Wand and Zoo. Despite the pullback, our portfolios remain firmly ahead: we’re up 11.26% year-to-date, a strong showing in the face of continued market volatility.

What’s particularly worth highlighting is our rolling 1-year performance, which has been exceptional:

• Sharpe Ratio: 1.97

• Sortino Ratio: 3.96

This performance aligns closely with our strategic transition to managing two distinct portfolios. The move to separate Zoo has effectively reduced expected annual drawdowns while preserving upside asymmetry — a key objective in our portfolio architecture. See our performance below:

Consolidated Portfolio Performance 1-Year: +68.94

Consolidated Portfolio Performance YTD: +11.26%

1. Wand OG Performance for YTD: +7.24%

2. Mythic Zoo Performance for YTD: +16.79%

   

Consolidated Portfolio Performance Since Inception: 83.81%

Wand Portfolio Performance Since Inception: 117.00%

Zoo Portfolio Performance Since Inception: 42.04%

Hey, there!

Q2 is off to a hell of a start! With the first two months of the year starting off slowly it has been amazing to have two amazing months. As of the close of April, the Mythic Prop Portfolio is up 19.16%! Our risk-to-return metrics YTD are amazing as well with a daily Sharpe of 1.79 and daily Sortino of 3.73.

Consolidated Portfolio Performance YTD: +19.16%

1. Wand OG Performance for YTD: +13.88%

2. Mythic Zoo Performance for YTD: +26.55%

Consolidated Portfolio Performance Since Inception: 96.86%

Wand Portfolio Performance Since Inception: 130.44%

Zoo Portfolio Performance Since Inception: 53.90%

Hey there!

March was a fantastic month for our team at MMR, achieving a return of +13.93% and bringing us back into the black, with a positive year-to-date return of +8.56%. I don’t want to sound boring or repetitive, but we continue to adhere to our process. We stick to our strategies, understanding that there will be times when they are in favor and times when they are not. As you will see below, we have a beta of 4%, and our daily returns are 3% correlated to the Nasdaq Composite. Our short-term systems are a real diversified return stream. Our greatest assets are patience and time.

Consolidated Portfolio Performance Since Inception: +79.34%

Wand Portfolio Performance Since Inception: +116.81%

Zoo Portfolio Performance Since Inception: +34.48

Consolidated Portfolio Performance YTD: +8.56%

1. Wand OG Performance for YTD: +7.14%

2. Mythic Zoo Performance for YTD: +10.58%

   

Hey there!

Friday, February 21, was our biggest day in 12 months, with a +6.85% gain. Given that, providing a portfolio update feels appropriate. We are a bit behind in our updates, so this review will be a year-to-date overview to catch us up.

Consolidated Portfolio Performance YTD: -1.27%

1. Wand OG Performance for YTD: -5.78%

2. Mythic Zoo Performance for YTD: +5.82

Consolidated Portfolio Performance Since Inception: +64.71%

Wand OG Portfolio Since Inception: +94.73%

Zoo Portfolio Since Inception: +27.77%

Final Thoughts

We believe that we will continue to benefit from market volatility and macroeconomic uncertainty. Last Friday, our systems shorted stock index futures and crude oil, which helped us reach an all-time high in equity and maintain a positive position for the year. Since then, we've experienced a slight pullback. However, our systems thrive on intraday volatility, and intraday trading ranges will likely remain wide for some time. Only time will tell.

Thanks for reading, and happy trading!