Chokepoint

The neocloud carnage isn't letting up.

A GPU rental company called Elice just went public in South Korea — raised $151M at a $750M valuation, retail piled in 400x oversubscribed. Their business? Renting out GPUs and running the systems around them.

Stock's down 30% on debut. If it holds, worst South Korean IPO opening ever.

Meanwhile KOSPI dropped 4.2%, now below July lows. Samsung down 3%, SK Hynix down 6%.

This is the backdrop for the GPU infrastructure trade right now. When a hyped GPU rental play can't even survive day one in a market that was desperate to get in, that's your signal on how much froth just got wrung out. The neocloud names — $CRWV $NBIS $IREN $APLD $CIFR — are all feeling this same pressure.

Upstream is holding better than downstream. The guys selling the shovels (ASML, Applied Materials, the memory makers) still have structural demand. But the guys renting out compute? They're fighting for margin in a suddenly crowded, suddenly skeptical market.

Watch how the Korean semi complex trades from here. SK Hynix at -6% is a read-through on HBM sentiment and AI capex expectations. If that keeps sliding, it'll bleed into the US neocloud names next.

ARK just updated their 2031 $TSLA model — base case $4,400/share, a 12x from here. That would put Tesla at an $18T market cap in four years.

The thesis hinges on two big bets:

1. Robotaxis = 80% of enterprise value. ARK pencils in 8.1 million robotaxis generating $753B in annual ride-hail revenue by 2031. They're modeling operating costs 40% below Waymo's and a massive expansion of the addressable market through lower prices. That's the overwhelming driver.

2. Optimus robots = the second engine. ~1 million units sold externally, $212B in annual revenue. ARK sees a $26T addressable market for humanoid robotics and expects Tesla to pour ~$225B cumulatively into AI infrastructure by 2031 to make it happen.

This is a pure optionality bet on two nascent markets — autonomous transport and humanoid labor — where Tesla has real but unproven leads. The robotaxi math requires regulatory green lights, fleet deployment at scale, and unit economics that hold up in the real world. The Optimus math requires a working product, enterprise adoption, and a supply chain that doesn't exist yet.

ARK's modeling these as high-probability outcomes. The market is not. That's the gap.

Crypto derivatives open interest has gone from ~$15B in 2021 to $100.4B today — a 570% jump — with $BTC alone holding $42.6B (42.4% of the total). The real inflection came in Jan 2024 when spot $BTC ETFs launched: institutional flow followed, liquidity deepened, and OI spiked to an ATH of ~$155B before a violent deleveraging event knocked it down to $54B. It's since climbed back to $120B and now sits around $100B.

The derivatives tail now wags the dog. Higher OI means tighter coupling between funding rates, gamma, and spot — which also means bigger cascade risk when positions unwind. The 2024 deleveraging wasn't a one-off; it's the new normal when leverage builds this fast. If you're trading $BTC or alts, you can't ignore the derivatives stack anymore — it's where the real volatility structure lives. Watch funding, watch OI deltas, and respect the liquidation cascades. This isn't 2021's market; it's institutionalized, levered, and far more reflexive.

Two quick reads on tonight's tech prints:

$NFLX — The market's been selling the number for a year straight. An EPS beat won't change that. What actually moves the stock: revenue above consensus plus forward guidance that breaks the fade pattern. Without both, expect the same drift.

$META — The print itself is secondary. What matters is capex and opex guidance. Every major selloff that wasn't a revenue miss came from spending guidance spooking the Street. If they guide up materially on AI infrastructure spend, that's your catalyst — regardless of the beat.

Full teardowns coming.

Two wildly different AI signals hitting at once, and they can't both be right.

Anthropic just committed to giving away 20% of the company to the public. Elon floated handing one share of $SPCX to every American kid — call it $10B at today's prices. Both are effectively saying AI's so transformational that they'd rather distribute equity now than face what's coming: higher taxes, political blowback, or both. Elon's already talking universal high income, which is code for "the disruption is so big we'll need to redistribute."

Meanwhile, CoreWeave's bonds are getting hammered, and the market's suddenly questioning whether the AI buildout even makes sense.

So which is it? Are we on the edge of a shift so profound that founders are pre-distributing wealth to avoid the fallout — or is the infrastructure stack overbuilt, overleveraged, and about to crack?

Someone's reading this completely wrong. Either the market's mispricing the biggest supply-chain build in decades because one debt structure spooked it, or the insiders know the hype's ahead of the fundamentals and they're cashing out via philanthropy.

I'd bet the market's overreacting to CoreWeave. The upstream bottlenecks — lasers, substrates, HBM, packaging — are still real. Demand's still there. But the debt wobble is a warning: not every AI infrastructure play is investable, and leverage kills in a high-rate world. The thesis isn't dead. The structure matters more than ever.

We won't know if today was a bull trap until Wednesday. Market's playing poker right now — the real hand gets revealed mid-week when we see how price holds and what the options flow does into expiry. Until then, it's just noise and positioning.

$AMZN just had its best session since August — up 5% on Morgan Stanley naming it a top pick with a $350 target (40% upside). The real story is TorchNeuron, a software layer that lets developers port PyTorch workloads to Amazon's Trainium2/3 chips with minimal code changes. That's a direct shot at $NVDA's CUDA moat — the unglamorous software lock-in that actually keeps customers paying the premium.

Morgan Stanley now models AWS growing 47% and 42% YoY through 2027-2028. The bull case hinges on Trainium inflecting and capital rotating from semis into the Mag 7 cloud operators who've been underperforming — $AMZN is up 50% over five years vs. $QQQ's 94%.

But here's the tension: the Street has been calling 2027 "Amazon's year" for two years running. The setup is logical — if AWS accelerates and Trainium gains real share, the multiple re-rates. The risk is that the software layer still lags CUDA's ecosystem depth, and hyperscaler capex keeps flowing to Nvidia silicon anyway. The bottleneck isn't just the chip; it's the toolchain, the libraries, the installed base. TorchNeuron is a step, but it's early. Watch AWS growth prints and Trainium adoption metrics — that's where the thesis lives or dies.

$SPY opened Monday at $767.29, up 0.53% after OPEX wiped the board clean. Friday's close was $763.25 with -$2.46B gamma and puts opening 7:1 — no bounce to $765, but $760 held.

From Friday 2:11PM to Monday 10:44AM:
Price: $762.86 → $767.29
Composite: -48.3 → +16.3
Gamma: -$2.46B → -$786M
Engine: -261M → +12M
Net premium: -$1.68B → +$478M
Premium mix: 25% calls → 73% calls

OPEX cleared $1.7B of negative gamma. The overnight repair mechanism is back online.

The line moved. Last week the first positive strike was $775. Now it's $770, sitting $2.71 above price. Key strikes:
$770: +$88M
$771: +$56M
$775: +$73M
$780: +$116M

Most of $770 and $771 is today's expiring calls — the new line is real but thin.

Below price:
$765: -$120M (max pain)
$760: -$198M
$750: -$214M (largest strike on the board)

About $1.2B negative gamma from $745 to $767. About $790M positive gamma from $770 to $790.

Volatility collapsed. ATM IV dropped from 17.0/19.2 Friday to 14.3/14.7 now. Put premium nearly gone. Realized vol: 10.2%.

$770 is the line. $767.29 is price. $765 is max pain. $760 is the floor.

The board reset. Calls are back at 73%. Clear $770 and the positive side takes over for the first time in a week.

SPX just flipped into the deepest negative gamma since July — dealers are no longer absorbing dips, they're amplifying them.

The setup:
• Net gamma dropped from +$216M Friday to -$536M Monday
• $558M of negative gamma stacked between 7,500–7,650
• SPX closed at 7,659, sitting directly on the heaviest accelerator
• Gamma flip now 80 points above spot at 7,740
• 9 of top 10 volume strikes were puts

What this means: In negative gamma, dealers sell into declines. No cushion. Moves get magnified in both directions — sharp drops accelerate, but short-covering rallies can rip just as violently.

The line in the sand is 7,700.

Reclaim it → run toward 7,740 flip becomes possible
Stay below → path of least resistance is lower, toward 7,600 then 7,500

Monday range: 7,575–7,725

The regime changed. The market isn't being cushioned anymore — it's being amplified.

The most important hour of earnings season so far is Wednesday, 4:00–5:00 PM ET.

$META $IBM $TXN $LRCX, back to back.

AI spending, enterprise demand, the broader chip cycle, and fab equipment. Four different tests of the same trade, in one hour.

The rest of the week:
Tue: $NFLX after the close, full Alphatica coverage
Wed AM: $PM $ELV $NEE $CME $T
Thu AM: $PG $CMCSA $BX
Fri AM: $AXP $SLB

Macro is thin. Housing data, jobless claims, flash PMIs. No Fed speakers ahead of the Oct 27–28 meeting.

When there's no macro to lean on, earnings set the tape.

The real question isn't whether we'll see a bear market — it's whether this administration can politically survive one.

AI capex is roughly 50% of GDP growth right now. If that slows or stops, GDP growth evaporates and we're in recession territory. Mainstreet gets crushed first as the labor market collapses.

The top 10% drive 90% of consumer spending, and they're levered against their portfolios. Stocks go down, spending dries up, and we're in a real recession — especially since the median consumer is already maxing out credit cards.

Social Security is structurally broken over the next decade. The Trump Accounts aren't about financial literacy — they're a replacement mechanism. If kids compound from birth in the market, maybe that offsets cuts in 2032. But that only works if stocks go up. Most 401(k)s are retirement plans. If the market crashes, people can't retire when benefits get slashed.

We're $40T in debt. Bond yields keep climbing, so interest expense is eating more of the budget. The only way out is growth — neither party will actually cut spending. If the market crashes, the AI buildout stalls, growth dies, and the debt spiral gets worse.

I'm not saying a bear market is impossible. But the entire economy is levered to AI, which is the entire stock market right now. This admin will fight hard to prevent an ugly crash. If they let it happen, everything unravels.

The difference from dotcom: GPU demand is real, no oversupply. The difference from 2008: the credit risk isn't in housing and banks — it's in GPUs, and there's so much demand that hyperscalers, $NVDA, and the government could step in to backstop things before they spiral.

Maybe we get an epic crash. But this administration really can't afford to let that happen.

Everyone calling for Black Monday after October opex needs to check the tape. History says the opposite.

$SPY has closed green 21 out of 26 times (81%) on the Monday after October expiration, averaging +0.71%. $QQQ is even stronger: 22 out of 26 (85%), averaging +0.90%.

Opex unwinds gamma, dealers cover short delta, and flows typically reverse bullish. The fantasizers are fading the pattern.

The 75th percentile ceiling finally broke.

Three spikes since mid-September — all reversed the next day. This week was different.

Oct 12: 44th percentile
Oct 13: 50th percentile
Oct 14: 93rd percentile
Oct 15: 96th percentile (EXTREME)
Oct 16: 93rd percentile

Three straight sessions above 75th. First time since August.

The pattern inside those three days: equal-weight $SPY (RSP) beat $SPY every session by 0.5 to 1.1 points. On Oct 14 and Oct 16, $SPY fell over 1% while RSP fell about half that.

The mega-caps dragged. $NVDA and $AVGO closed red all three days. $AVGO dropped 4.5% Friday alone.

23 years of forward return data:

HIGH dispersion (75th–95th percentile) · 702 days
$SPY avg 20d: +0.63% · avg 60d: +1.79%
$QQQ avg 60d: +2.57%

NORMAL dispersion (40th–75th percentile) · 1,335 days
$SPY avg 20d: +0.88% · avg 60d: +2.71%
$QQQ avg 60d: +3.62%

HIGH is the weakest forward return bucket. That's where we are now.

Three days isn't a regime yet. Spring's elevated period ran for months. But this is the first time since August the spike didn't fade overnight — and that changes what we watch.

If dispersion falls back below 75th next week, this was just a bigger absorbed spike. If it holds above, the compression that started in August is over.

HIGH dispersion is where the index struggles and selection matters most. Stock picker's regime, not an index regime.

Fear & Greed divergence: stocks at 18 (Extreme Fear), crypto at 55 (Neutral).

That's a rare split. Usually they move together when macro tightens or liquidity shifts. Right now, equities are pricing in real recession risk — earnings cuts, multiple compression, the usual suspects. Meanwhile, crypto sits in neutral territory, which tells you either: (1) the asset class has decoupled from traditional risk-off flows, or (2) crypto holders are numb to volatility and just holding through.

My read: this gap won't last. If stocks keep bleeding, crypto follows — it always does when liquidity actually dries up. But if equities stabilize and the Fed pauses, crypto could front-run the recovery because it's not pricing in the same level of fear. Watch $BTC's behavior around key equity support levels. If it holds while SPX breaks down, that's your signal that crypto might actually be building a base independent of tradfi risk appetite.

For now, the divergence is more of a timing question than a structural decoupling. Extreme fear in stocks usually precedes a flush or a bounce — and crypto's neutral stance suggests it's waiting to see which one.

Quick probability check on $SPY for anyone still sweating Black Monday vibes: roughly 25% chance we touch 749.6 intraweek, but only ~10% odds of actually closing below it by Friday.

That's a meaningful gap. Intraweek dips get bought. A close below would signal something stickier — but the math says it's the less likely outcome. Watch how we handle that level if we get there.

Every name ALPHATICA flagged as bearish or neutral on earnings quality has gotten crushed. All of them.

The list: $APP, $CRWD, $CAT, $TSLA, $INTC, $INTU, $CAR, $COHR, $JEF.

The pattern is clean — when the balance sheet quality signal turns, the stock follows. Not sometimes. Every time.

Their read: stick to pristine balance sheets until midterms clear. The thesis is that accounting red flags — working capital bloat, revenue recognition games, cash conversion issues — show up in the stock price before the headline numbers roll over. This is classic forensic accounting as a leading indicator.

What makes this interesting is the breadth. You've got enterprise software ($CRWD, $APP), industrials ($CAT), semis ($INTC), fintech ($INTU), autos ($CAR), and specialty optics ($COHR). The common thread isn't sector — it's balance sheet stress showing up in the footnotes before it shows up in guidance.

The midterms angle is about policy clarity and fiscal positioning. Until that resolves, they're saying quality matters more than growth multiples. Translation: if the 10-K has footnotes that don't add up, don't fight it.

Track record like this doesn't happen by accident. Worth watching what they flag next.

Heavy 0DTE put flow hitting $SOXX 530 strikes — someone's positioning for intraday downside, which means they'll likely need to cover before close. We're tracking it, but the problem with 0DTE is the window's too tight to flip the intel and actually execute profitably. By the time you see it, the edge is gone. Still watching for a setup that has enough runway to matter.

The ugliest tell from today's price action? Stocks are falling but yields are rising — meaning investors are dumping both equities AND bonds at the same time. That's the nightmare scenario. Usually when stocks sell off, money rotates into bonds and yields drop. Not today.

This exact setup played out in April 2025. $SPY eventually fell 30%, the 10-year peaked at 4.8%, and Trump blinked on tariffs. When falling stocks can't even help bonds, the only circuit breaker left is policy reversal.

We're now in the zone where markets are forcing the hand of the administration. If bond buyers stay on strike, the Fed and Treasury lose their cushion. Watch yields. If they keep climbing while equities bleed, the pressure for a Trump pivot becomes unbearable.

October monthly opex today. No directional lean. Historical average sits around +0.07% with win rate between 61–65%.

$SPY

$SPY opened flat at $771.11 on monthly expiration — down 0.09%. Yesterday's late-session bullish flow (+$2 rip into the bell from 2:43PM) faded overnight. The composite flipped from +23.7 to -12.9. Money's still buying calls (68% of premium), but the conviction isn't holding.

The real story is the cliff at $772. Thursday's hidden -$214M strike doubled to -$442M, almost all expiring today. It's the biggest strike on the board, sitting $0.89 above price. Right above it: $773 and $775 each have +$161M — positive gamma magnets a dollar apart.

Below: $769 (-$127M), $768 (-$90M), $765 (-$182M, max pain), $760 (-$188M). About $1.7B of negative gamma from $750 to $772. About $900M of positive gamma from $773 to $790.

ATM IV: 14.0% calls, 13.5% puts. Realized vol: 9.6%. Down 0.96% for the week from $778.57.

The line is $773. Clear it and the magnets pull you higher. Fail at $772 and expiration pins you to $765. One dollar, two outcomes.