Days like today are brutal, and you can feel the exhaustion across the timeline. I don't have a crystal ball, but I think we are at or very close to the price bottom of this leg.
What we are experiencing right now is a textbook forced liquidation washout. When leveraged momentum assets like gold, silver, Bitcoin (breaking $60k), and AI all get dumped together on a monthly OpEx Friday, that is cross-asset VaR deleveraging and short-gamma dealer hedging, not a sudden collapse in fundamentals. Watching the Nasdaq drop another leg this afternoon on no fresh news while VIX tags 32+ and we sit right on the 20-month SMA is classic mechanical capitulation.
Why I still don't think this is a systemic market crash:
-Look at the equity internals: the Dow is barely red and consumer staples are green. In a 2008-style credit breakdown, defensive stocks don't hold up. This is a violent positioning flush in crowded risk assets, not the economy breaking.
-The macro backdrop is quietly improving underneath the panic: oil is cooling into the low $80s and the 10-year yield just slipped back below 5%.
-The AI demand thesis hasn't broken: GPU rental prices are still rising and cloud backlogs just jumped to $2.5T. High-beta small/mid cap names get hit hardest in liquidity vacuums, which is painful to sit through, even when the underlying balance sheet and pipeline haven't changed overnight.
What to expect over the next couple weeks:
Bottoms are usually a process, not a clean V-shape. Once OpEx clears, we likely get a sharp reflex bounce in liquidity in the coming week or so. From there, expect a choppy retest in early December as peak tax-loss selling collides with positioning ahead of the Dec 9 FOMC and Dec 10 CPI.
The two things that could still break this and trigger a new low: the Fed shocking the market with a 50bps hike (25bps is already priced in) or a major escalation in Iran that spikes oil again for a prolonged period of time. In my view, both would be catastrophic mistakes given the current state of the market.
Barring that, the real tell that the bottom is in won't be one big green day, it will be positive divergence on that early December retest: indices revisiting today's lows while VIX stays lower than 32 and bad news stops taking the leaders down.
As always, stay level-headed and manage your risk.
Replying to @HabeshaTechBro
Today’s gotten the best of many of us incl Serenity. You could feel the stress in his comment earlier today that at these valuations it would be small change for someone to acquire Sive. What’s your take on this?
2
5
1,043
On the $SIVE buyout point: mathematically, @aleabitoreddit is right. At a ~$370M market cap with zero debt and a 100M+ InP laser fab ramping in Glasgow, it is literal petty cash for a hyperscaler or tier-1 optical player looking to secure supply.
That said, I view M&A as a valuation backstop, not a primary thesis. Big players usually prefer locking up capacity through prepayments and LTAs rather than buying an unproven ramp, and management isn't going to sell the company at washout lows after raising at 14.50 & 57 SEK and prepping a US dual-listing.
Honestly, hearing smart bulls start talking about buyouts is usually another classic sign of the peak exhaustion that we’re all experiencing. If they execute the 2027 ramp, the market will re-rate it on its own.
1
3
442
The AI cycle didn't break, it just got flushed out. As a fun exercise, I wanted to share my prediction for how the next 12 months will unfold:
Now-Dec 2026: Max pain. Tax-loss selling collides with peak Fed hawkishness at the December FOMC. High-beta tech finishes washing out and prints its cycle lows.
Jan-Feb 2027: 2027 CapEx confirmation. Hyperscalers report Q4 earnings and lock in another step-up in 2027 spending plans, killing the "peak CapEx" narrative. Strong balance sheet leaders bottom first and start grinding higher.
Late Q1 2027: The macro inflection. Crude oil keeps grinding lower into the $70s as global supply stays in surplus, finally bleeding out of Q1 inflation prints. Rate futures price out summer hikes, the Fed pivots off peak hawkishness, and Treasury yields begin to roll over.
Q2 2027: The violent recovery. Underweight funds sitting on the sidelines are forced to chase back in. Risk appetite rotates aggressively down the cap curve into beaten down, high-beta AI trade.
H2 2027: Peak AI euphoria. Expanding multiples meet accelerating 2027 earnings, and the exact same crowd calling a burst bubble today is back chasing vertical charts at new highs.
The above timeline/sequence of events obviously depend on one another (and most importantly on oil continuing to fall) but the overarching point is that with the AI trade now being severely oversold, the fundamentals are in place and the 2027 bull set up is starting to take shape. All that is missing is the macro green light.
People look at charts like this and panic, when in actuality a positioning washout this extreme can be a strong bullish signal.
Prime brokerage data is backward-looking. Seeing the heaviest net selling of global tech in 9 years and net momentum exposure collapse from ~60% down to 8% just explains why the last six weeks felt so brutal. That wasn't a sudden collapse in AI demand. It was forced, mechanical deleveraging as funds cut risk into spiking oil and 5% yields.
When positioning gets flushed this hard, the crowded weak hands are already out. It doesn't guarantee we bottom to the exact day with the December FOMC and tax-loss selling still ahead, but it leaves a massive pool of underweight institutional capital sitting on the sidelines. The moment the macro switch flips and yields roll over, those exact same funds have to chase back in.
It’s wild how few accounts on X understand that earnings only move high-beta names in a risk-on environment. In a risk-off tape like this, a blowout quarter is just exit liquidity for funds looking to de-risk.
Right now, the real value of earnings and supply/demand visibility isn't the next day/week reaction. It’s in helping you pick the right names to own when the macro tide actually turns.
Stop beating your head against the wall wondering why your thesis isn't being validated or buying into the "bubble burst" noise. Start tracking crude oil, the bond market, and Fed policy instead.
1
7
655
Everyone on X: The bubble burst. It’s over. Everything is going to zero.
Me: Quietly watching one of the cleanest macro bull setups I’ve ever seen starting to shape up.
Max fear is everywhere right now, yet WTI crude has quietly dropped nearly 16% over the past month to $83 as rising U.S. supply and Gulf pipeline bypasses to the Red Sea offset the Strait of Hormuz disruption.
Energy has been the primary force keeping inflation sticky, pushing yields above 5%, and forcing the Fed to stay hawkish. As lower oil bleeds into headline CPI over the next couple of months, the bond market gets room to price out future hikes and let yields roll over, creating the exact macro shift that the beaten down high-beta AI trade needs to stop fighting multiple compression.
Every time the AI trade gets washed out, the same crowd comes out declaring the bubble has burst. They completely ignore the underlying demand and miss what is actually driving the tape: monetary policy and the macro risk environment.
Every major AI rally over the last few years has needed the exact same macro backdrop. When the 10-year yield is sitting above 5% and the Fed is actively hiking into an energy shock, institutional capital flows into defensive positions and growth multiples get mechanically compressed. Fundamentals can be accelerating and backlogs can be hitting records, and the space will still struggle to hold a bid until that headwind eases.
The good news is you don't need oil at $65 or the 10-year at 3.5% for the next leg up to start. Markets always front-run the turn. The risk-on switch flips the moment the rate of change reverses, when the narrative shifts from "how many more hikes?" to "the Fed is done."
Here are 5 things to watch over the next several months for signs that the rally is back on:
1. Crude breaking cleanly into the low $80s or $70s: Despite the Hormuz disruption, global supply is already moving back into surplus. As long as the Red Sea stays open and oil keeps grinding lower, the biggest driver of the Fed's hawkishness loses its teeth.
2. Q4 earnings and 2027 CapEx budgets (late January to February): This will be a fundamental trigger. When hyperscalers report Q4 and lock in their official 2027 spending plans, another step-up in CapEx kills the "peak spend" narrative and sets a hard floor under the sector.
3. The oil-to-CPI lag clearing in Q1: It takes 4 to 8 weeks for lower crude to bleed out of headline inflation. The January and February inflation prints (dropping mid-February and mid-March) are where energy finally shifts from an inflationary tailwind to a drag.
4. The 2-year yield and mid-2027 Fed Funds futures: Right now the market is pricing in roughly four hikes through next fall, with December and January already baked in. Don't just watch the 10-year. The week rate futures drop from pricing four hikes down to two and the 2-year yield rolls over is the clearest signal that the rate of change has turned.
5. The March FOMC meeting: Once the market gets past peak hawkishness in December and January, the bar for a dovish surprise drops significantly. If Q1 inflation cools and the Fed signals a pause or softer pace in March, yields peak and the bond market stops fighting the tape.
A surprise diplomatic deal in the Gulf pulls this whole timeline forward overnight. Without one, expect chop through the winter while the bond market finishes digesting peak hawkishness. The AI cycle didn't break. It is just waiting for 2027 CapEx confirmation and rolling yields to unlock the next leg.
Nvidia’s ($NVDA) Q3 call just confirmed the optical bull thesis: 1.6T pluggables are booming while CPO arrived a year early. What it means for $LITE, $COHR, $AAOI, $SIVE, $AEHR, and $AXTI:
-Networking & The Vera Rubin Velocity ($LITE, $COHR, $AAOI): Networking crossed $21B in Q3 ($84B+ run rate), while Vera Rubin hit ~$20B in its first quarter. Hyperscalers brought Rubin racks online in one month (half Grace Blackwell's time), compressing commissioning and pulling forward optical transceiver deployments. Colette noted Meta is expanding Rubin after its Muse agent hit 10M setups in six weeks. With $AAOI anchored by Meta and Microsoft, 2027 demand for 800G and 1.6T pluggables has a massive tailwind alongside Tier-1s $LITE and $COHR.
-The Agentic East-West Traffic Explosion ($AAOI, $LITE, $COHR, $SIVE): Weekly token volume on OpenRouter surged 35x YoY and 3.5x in three months, with agents driving 80% of tokens and Vera CPU revenue doubling in FY28. While intra-rack traffic rides NVLink, multi-node agents coordinating across racks drive exponential east-west scale-out optical bandwidth. Add frontier labs (15% of DC revenue) building their own clusters and a 2GW Australia sovereign buildout by late 2027, and 1.6T optical demand is expanding well beyond legacy clouds.
-CPO Scale-Out Is in Volume Production ($LITE, $COHR, $SIVE, $AEHR): Colette confirmed CPO scale-out is in volume production today, anchored by Lambda deploying Nvidia Photonics CPO switches across 10,000+ GB300 GPUs ahead of Rubin Ultra (2027) and Feynman (2028). CPO switches delivered 5x better power efficiency, 15x higher reliability, and 10% lower latency than pluggables, killing the 2029 science-project bear thesis. As Tier-1 CW laser capacity ($LITE, $COHR) tightens, this creates spillover for merchant CW-DFB laser arrays ($SIVE) and accelerates wafer-level burn-in ($AEHR), as laser failure in sealed CPO packages cannot be tolerated. For $AAOI, while pluggables rule 2027, internal Sugar Land and Taiwan InP fabs provide a vital pivot into upstream laser supply as CPO scales in 2028+.
-$390B in Purchase Commitments & Upstream Scarcity ($AXTI, $LITE, $COHR, $SIVE, $AAOI): Nvidia raised FY28 growth guidance from 70% to 75% YoY, cloud backlog jumped $500B in 90 days to $2.5T, and purchase commitments hit $390B. Alongside buybacks, Nvidia's #1 capital priority is supplier prepayments to expand capacity after absorbing Q4 cost inflation. While most dollars lock up CoWoS and HBM, multi-year networking reservations tighten global InP laser supply, trickling down to InP substrates ($AXTI) and boosting vertically integrated US and Taiwan InP fabs ($LITE, $COHR, $AAOI).
-Modular DSX Pods & Time-to-First-Token ($LITE, $COHR, $AAOI, $SIVE, $AEHR): Modular DSX Pods cut cluster deployment from months to weeks. With a 1GW AI factory generating ~$50B in annual revenue, bringing a cluster online one quarter early unlocks $10B+ in revenue. That math drives hyperscalers to sign multi-year optical LTAs, fund second-source qualifications ($SIVE, $AAOI), and use wafer-level burn-in ($AEHR) rather than let an interconnect shortage delay a $50B cluster.
While the headlines fixate on the C-suite transition, the 2027-2029 optical setup is crystal clear: 800G and 1.6T pluggables carry the near-term surge, CPO takes over the multi-year roadmap, and every path runs through InP, CW lasers, and wafer burn-in.
Everyone is understandably frustrated watching $AAOI file a $550M ATM shelf 12 days after CFO Stefan Murry said on the Q3 call that they planned to rely more on debt and internal cash flow. Management deserves the hit to their credibility for how they telegraphed this.
Once you get past the communication fumble and open the Q3 10-Q, though, the numbers explain what is actually happening behind the scenes:
-The Unfunded Commitments in Note 18: The headline Q3 release showed $621.5M in cash and a new $200M bank loan ($821.5M total liquidity). What most people missed is Note 18 of the 10-Q, which shows $AAOI is carrying $928.8M in binding equipment and facility purchase commitments for Sugar Land Phase 2 and Taiwan. They were already over $100M short of their committed CapEx bill before funding a single dollar of new working capital.
-Funding a $5B+ LTA Without Customer Prepayments: Unlike $LITE, which received upfront customer prepayments to build capacity,$AAOI's first $1B to $1.5B 1.6T contract included zero upfront capital from the hyperscaler. On the Q3 call, Thompson Lin stated a second LTA is in the final stages, weeks away, and at least 5x larger than the first ($5B to $7.5B). Scaling toward $471M/month in revenue by late 2027 means $AAOI has to prepay 1.6T DSP suppliers and front hundreds of millions in inventory and receivables on its own balance sheet while waiting 60 to 90 days for hyperscaler invoices to clear.
-Passing the Counterparty Audit & Selling Into Strength: No Tier 1 hyperscaler risk committee is signing off on a $5B+ contract with a $4B market cap supplier unless that supplier can prove it has the liquidity to survive the working capital ramp. Filing the ATM right after the 10-Q gives $AAOI ~$1.37B in total liquidity capacity to check that box for the customer. It also puts the legal shelf in place now so Raymond James and Needham can quietly sell shares into the massive volume spike after the 5x LTA hits the tape (just like they executed the prior $600M ATM at a $105.36 average price), rather than killing the rally by filing for dilution the day after the announcement.
-The Dilution Runway Through Late 2027: With $1.37B in combined cash, debt, and ATM capacity, the entire $928.8M CapEx commitment is covered with over $440M left for working capital, and Q4 Non-GAAP net income is guided to jump to $46M to $60M in a single quarter. Unless they land yet another massive LTA on top of this upcoming one that requires expanding capacity beyond 1.2M units/month, this should be their last ATM through the end of 2027.
The ATM creates near-term overhead supply and management handled the messaging poorly. Even so, factoring in 10% to 12% dilution and a 50% execution haircut to their late 2027 run rate, $AAOI is trading around 10x buffered 2027 earnings with a $5B+ contract weeks away. Selling in the low $40s right before Q4 earnings inflect misses what this balance sheet is being built to support. I’m holding.
Great 13F recap from @JonahLupton. Looking across all 85+ funds, six major institutional trends stand out:
1. Mag 7 Bifurcation: Funds are treating $GOOGL as a source of cash (~20 top sales vs. only 4 buys) and trimming $TSLA to pile heavily into $MSFT (~27 top buys) and $META (~22 top buys).
2. Semi Rotation: Institutions dumped wafer fab equipment ($KLAC, $AMAT, $LRCX, $ASML) and took profits on AVGO (8 top sales, 0 top buys) to rotate into $NVDA, $AMD, and a contrarian $INTC turnaround (9 hedge fund top buys, 0 top sales).
3. Memory Split: Aggressive hedge fund accumulation in flash and HBM ($SNDK, $SKHY), a two-way tug-of-war in $MU, and wholesale dumping of legacy hard drives ($WDC).
4. Macro Hedging: Multi-strats (Citadel, Millennium, Elliott, D.E. Shaw) loaded up on index and credit puts ($SPY, $QQQ, $IWM, $HYG) while monetizing and closing out their semiconductor put hedges ($SMH puts).
5. Software & Event-Driven Clustering: $PLTR and $ORCL saw heavy mainstream institutional buying, WBD became a crowded event-driven trade across 6 funds, and healthcare managers rotated out OF $UNH into medtech and generics ($DHR, $TMO, $MDT, $TEVA).
6. Real Assets & Private Secondaries: Dedicated REIT funds dumped apartments and offices to buy data center and tower infrastructure ($DLR, $EQIX, $AMT, $SBAC), while SpaceX ($SPCX) saw massive two-way secondary clearing.
Big money isn't leaving the market, it is just hedging macro index risk while rotating aggressively out of crowded winners into specific compute, memory, and physical infrastructure plays.
Here's a recap of some 13F's:
A16z:
Top Buys: SPCX, BRVE, FRNM, LIME, SCTX
Top Sales: META, FIG, RVMD, ERAS, OMDA
Adage:
Top Buys: TEVA, MDT, MRK, MOD, HESM
Top Sales: TSLA, MU, AVGO, LRCX, AON
AllianceBernstein:
Top Buys: MSFT, MU, AMD, TXN. ORLY
Top Sales: KLAC, BKNG, NFLX, VUG, SYK
Alyeska:
Top Buys: AMZN, HOOD, MU, ORCL, INTC
Top Sales: IVV, CRWV, CRM, ANET, ISRG
Ameriprise:
Top Buys: MSFT, AAPL, TMO, PLTR, HPE
Top Sales: AMAT, KLAC, TJX, MS, WMT
Anomaly:
Top Buys: NVDA, JBS, MA, RACE, Q
Top Sales: GIL, MTZ, V, GOOGL, WMT
APG:
Top Buys: LQD, DLR, WELL, VMRK, VTR
Top Sales: COLD, HYG, AVB, MAA, BXP
Appaloosa:
Top Buys: SPCX, MU, ORCL, NVDA, AMAT
Top Sales: GOOG, AAPL p, NRG, AMD, EWY
AQR:
Top Buys: MSFT, NVDA, META, AAPL, ORCL
Top Sales: WDC, EIX, PCG, TER, CRS
Avala:
Top Buys: MSFT, SNDK, INTC, CLS, META
Top Sales: LITE, WDC, CORZ, AMD, AGX
Balyasny:
Top Buys: HYG p, MSFT, IWM c, NVDA, SMH c
Top Sales: SPCX, AMZN, IVV, MMM, KVUE
Barrow Hanley:
Top Buys: AMZN, NVDA, VRT, VMRK, TEL
Top Sales: MRK, URTH, AXTX, BNS, VICI
Baupost:
Top Buys: HONA, SITE, GPC, CME, WCC
Top Sales: TFX, NCLH, GOOG, LBTYK, UNP
Berkshire:
Top Buys: LEN, BSP, LEN.B, M, DAL
Top Sales: BAC, DVA, VRSN, COF, KR
Blackrock:
Top Buys: META, CRWD, CVX, TMO, AMD
Top Sales: LRCX, AVGO, CAT, GOOGL, GLW
Blackstone:
Top Buys: JMKE, QQQ c, INTC, ET, PAA
Top Sales: DLR, SPY, SPY p, TXNM, SPCX
Bridgewater:
Top Buys: QCOM, APP, MU, QRVO, SNDK
Top Sales: SPY, IVV, GOOGL, VOO, AMZN
BVF:
Top Buys: TYRA, ENTX, STOK, ALKS, PYXS
Top Sales: XOMA, MIRM, GPCR, IRON, GULE
CalPERS:
Top Buys: MSFT, NVDA, AAPL, META, GOOGL
Top Sales: VOO, KLAC, TSLA, UNH, GLW
CalSTRS:
Top Buys: IVV, TEVA, SUNB, VMRK, LITE
Top Sales: TSLA, AMAT, KLAC, AVGO, LRCX
Capital International Investors:
Top Buys: MSFT, AAPL, JNJ, DHR, XOM
Top Sales: AZN, KLAC, CAT, CRM, ASML
Capital Research Global Investors:
Top Buys: META, NVDA, ORCL, INTC, ANET
Top Sales: AMAT, RCL, LLY, AVGO, TDG
Capital World Investors:
Top Buys: MSFT, AAPL, NET, CNQ, JNJ
Top Sales: MU, KLAC, TSLA, IBM, UNH
Caxton:
Top Buys: WBD, GM, CRWD, PLTR, AMZN
Top Sales: AAPL, MSFT, NVDA, XLE, GOOGL
Cinctive:
Top Buys: QQQ p, WBD, MTZ, XOM, AMZN
Top Sales: SPY p, VST, IWM, APGE, NUVL
Citadel:
Top Buys: META c, SPY p, AMD c, AAPL c, SPY
Top Sales: SMH p, MU p, QQQ c, TSLA c, SPY c
Coatue:
Top Buys: MSFT, CRM, SKHY, META, SHOP
Top Sales: FPS, CBRS, GEV, ETN, GOOGL
Cohen & Steers:
Top Buys: EQIX, PLD, SBAC, OHI, CUBE
Top Sales: CCI, UDR, INVH, KRG, WY
Corvex:
Top Buys: MCHP, SPY p, VST, MLM, FLR
Top Sales: ILMN, PLNT, META c, MSFT, QSR
D1:
Top Buys: LSCC, MAIR, Q, MLM, FLEX
Top Sales: SPCX, META, FERG, SHW, GOOGL
Davidson Kempner:
Top Buys: SPY c, ROKU, WBD, IWM c, SPY p
Top Sales: EA, IWM p, QQQ p, SPCX, APGE
D.E. Shaw:
Top Buys: QQQ p, INTC, PLTR, MSFT, CAT
Top Sales: QQQ c, SPY, TSLA, CRWD, AMD
Delta Global:
Top Buys: MSFT, AER, FIVE, NTNX, ICE
Top Sales: PEP, WMT, GOOGL, AMZN, HLT
Deerfield:
Top Buys: UTHR, CELC, TEM debt, CHE, BMRN
Top Sales: NUVL, LNTH, PFE, SUPN, CDNA
Discovery:
Top Buys: EWZ, IFS, AMZN, BAP, TTE
Top Sales: COMP, CVX, JPM, GGAL, COF
Driehaus:
Top Buys: ZETA, FSLY, AMLX, WPM, PBR
Top Sales: CRNX, APGE, PL, VICR, CW
Duquesne Family Office:
Top Buys: STX, U, FOXA, STX c, AMZN
Top Sales: TSM RSP c, EWZ c, STM, GOOGL
Elliott:
Top Buys: QQQ p, IWM p, XLV p, GDX c, XAR p
Top Sales: IGV p, XLY p, IHI p, SNPS, HPE
Engineers Gate:
Top Buys: SYY, GM, MSFT, TT, CAT
Top Sales: SHEL, SGI, DECK, MUSA, TGT
Exodus Point:
Top Buys: QQQ c, SNDK p, SNDK, GLD c, META
Top Sales: AMZN, AMZN p, AAPL, AAPL p, QQQ p
FMR:
Top Buys: MSFT, AAPL, AMD, META, MU
Top Sales: SPCX, AVGO, MRVL, ALAB, AMZN
Fred Alger:
Top Buys: MSFT, NVDA, MU, META, TMO
Top Sales: SPCX, AMZN, APP, TSLA, NVO
Freestone Grove:
Top Buys: SYY, PH, WAT, BURL, HLT
Top Sales: IVV, USB, GE, AIG, ABT
Glenview:
Top Buys: TEVA, XLV p, META, HUM, ONTO
Top Sales: CVS, SPY p, TEVA, VTRS, GPN
HG Vora:
Top Buys:
Top Sales: PENN, UGI debt, FOSL, OPTU
Holocene:
Top Buys: AMZN, PG, GOOGL, AAPL, AMD
Top Sales: SOXX, MSFT, NBIS, KLAC, CSX
Hudson Bay:
Top Buys: SMH c, BRK.B p, AMD c, BRK.B, LITE p
Top Sales: IWM p, SNDK, SNDK p, TSM p, TSM
Icahn:
Top Buys: SD, IEP
Top Sales: IFF, JBLU, ECHO, AEP, BLCO
Interval Partners:
Top Buys: IWM p, JBHT, TSCO, DG, FERG
Top Sales: SPY p, CMG, HD, LOW, VVV
Jain Global:
Top Buys: SPY c, GLD c, GLD p, MRVL p, AMD p
Top Sales: XLF p, SPY p, TLT c, SMH p, SPY
Jana:
Top Buys: CWST, MOD, FISV, COO, FUN
Top Sales: MKL, MRCY, P
Jericho:
Top Buys: SNDK, SKHY, DOCN, META, INTC
Top Sales: ON, WDC, GOOGL, LION, CIEN
Kerrisdale:
Top Buys: FPS, ONT, CORZ, LGN, RTO
Top Sales: RRX, SWKS, AZTA, DV, BDC
Land & Buildings:
Top Buys: DLR, AMT, SBAC, SKT, INVH
Top Sales: CURB, SPG, RHP, DHI, BXDC
Lighthouse:
Top Buys: VLO p, RCL c, RCL, HIG c, DUK
Top Sales: SPY p, SPCX, WMT, CBRS, WBD p
LMR:
Top Buys: DRAM c, MRNA p, SPY c, CRWV p, NBIS p
Top Sales: IVV, QQQ p, GLD, p, SPY, NVDA p
Lone Pine:
Top Buys: AMZN, DHR, META, V, TSM
Top Sales: STX, HD, AMAT, LIN, NBIS
Longaeva:
Top Buys: MDT c, RSP, DHR c, CCI c, MDT
Top Sales: SPY p, SPCX, JPM, XLV c, BAC
Long Pond:
Top Buys: O c, H, ADC, LADR, FBRT
Top Sales: SBAC, SBAC c, JAN, POOL, FERG
Mane Global:
Top Buys: MSFT, FOXA, AMZN, MA, ULTA
Top Sales: GOOGL, CMG, TXRH, DELL, YUM
Man Group:
Top Buys: GILD, CRM, CMI, CSCO, TXN
Top Sales: SPYM, SPY p, APH, TEL, PEP
Marshall Wace:
Top Buys: SPY, VOO, NVDA, GOOG, META
Top Sales: IVV, MU, BAC, UNH, AMZN
Maverick:
Top Buys: DHR, MSFT, MLM, META, MDLN
Top Sales: ALAB, DKS, V, BSX, COF
Meridiem:
Top Buys: MSFT, CRWV, NVDA, SNOW c, EBAY
Top Sales: DHR, UNHM CRL, GOOGL, MTD
Millennium:
Top Buys: QQQ p. NVDA, p, SPY p, NVDA c, MSFT p
Top Sales: IVV, SPCX, SPY c, SPY, WDC debt
Miller Value:
Top Buys: CNXC, CRGY, COTY, GTN, SPUU
Top Sales: BLMN, BFH, LNC, CALM, QUAD
Moore:
Top Buys: META, IWM p, MSFT, SPY p, CTVA
Top Sales: TLT c, QQQ p, IWM c, UBS, AMAT
Neuberger Berman:
Top Buys: MSFT, SPCX, IWM c, META, AMZN
Top Sales: GOOGL, MU, HD, AZN, GOOG
NFJ:
Top Buys: AMZN, MSFT, AAPL, BLK, CVS
Top Sales: GOOGL, PNFP, NTR, DINO, UMBF
Nuveen:
Top Buys: MSFT, NVDA, META, MU, AMD
Top Sales: TSLA, AMAT, CAT, GOOGL, WDC
Palestra:
Top Buys: EQIX, FOXA, WWD, INTC, META
Top Sales: STX, CPNG, AMZN, TTMI, AMD
Paulson:
Top Buys: PVH, WBD, THRY, MDGL
Top Sales: GOOGL
Pentwater:
Top Buys: ORCL, ORCL p, MU p, WBD, QQQ p
Top Sales: EA, TECK, BA p, UNH c, DBRG
Pershing Square:
Top Buys: ICE, ALC, NFLX
Top Sales: BN, MSFT, QSG, UBER, AMZN
Pointstate:
Top Buys: SPY p, SPY c, ON, GH, AMZN c
Top Sales: META c, SMH p, DASH, PCG, AMZN
Point72:
Top Buys: MSFT, CLS, INTC, T, BA
Top Sales: CRDO, AMD, ASML, MKSI, TSM
Renaissance Technologies:
Top Buys: MU, AMD, UNH, SPCX, SNDK
Top Sales: NVDA, META, AMZN, PLTR, AAPL
Rokos:
Top Buys: NVDA, MU, XBI c, TSLA c, TSLA p
Top Sales: QQQ p, QQQ, META p, AAPL c, GE p
Ruane Cunniff & Goldfarb:
Top Buys: META, BRK.B, CHTR, GMED, NFLX
Top Sales: GOOGL, GOOG, ALGN, ZTS, LBRDK
Samlyn;
Top Buys: WAT, BLK, CBRE, INTC, META
Top Sales: META c, ITUB, MKSI, AMAT, TTMI
Schonfeld:
Top Buys: EEM p, HYG p, SKHY p, EWZ c, ARKB
Top Sales: MU p, IVV, NVDA p, MU, SMH p
Situational Awareness:
Top Buys: SHAZ, BE c, NBIS c, SNDK, DRAM c
Top Sales: SNDK, MU, BE, TSM, NBIS
Soros:
Top Buys: XLE p, XOP p, INTC, TLT c, XLU c
Top Sales: EA, GPN debt, LITE debt, SMCI pfd, DBX debt
Squarepoint:
Top Buys: QQQ c, QQQ p, MSFT c, SNDK p, NVDA
Top Sales: IVV, SMH p, XLC, SPY, SMH c
Starboard Value:
Top Buys: IJH, FLS, KNF, SHAK, IWM p
Top Sales: LW, GPGI, MTCH, NWSA, HR
Steadfast:
Top Buys: BURL, MHK, INTC, CBOE, AMZN
Top Sales: COST, DKS, MSCI, VRSK, RRX
Summittx Capital:
Top Buys: IVV, IWB, IWM, RSP, TSM
Top Sales: SPY, QQQ, ADI, NXPI, SPY p
Surgocap:
Top Buys: MSFT, MCK, HLT, MELI, TSM
Top Sales: GOOG, GLW, AGX, ISRG, MDLN
Thiel Macro:
Top Buys: SOXX, VIST, MSFT
Top Sales: AEP, DTE, FE, CMS, XE
Third Point:
Top Buys: META, UNM, NVDA, MSFT, FWONK
Top Sales: WBD, AMZN, GOOGL, MSC, CRS
Tiger Global:
Top Buys: NVDA, RACE, CBRS, DHR, MELI
Top Sales: GOOGL, TSM, SPOT, LRCX, AMAT
Tudor:
Top Buys: QQQ c, QQQ p, SPY p, SPY, META p
Top Sales: IVV, GLD c, MSFT, MU p, TLT c
Trian:
Top Buys: RTO
Top Sales: JHG, IVZ
T. Rowe Associates:
Top Buys: MSFT, PLTR, TTE, AMZN, SKHY
Top Sales: LRCX, AVGO, MU, CAT, AMAT
T. Rowe Investment Management:
Top Buys: AMD, BSP, ALNY, NVDA, ARES
Top Sales: AVGO, GOOGL, IR, SBUX, RVTY
Two Sigma:
Top Buys: NVDA, MSFT, KO, CAT, QQQ
Top Sales: WDC debt, BRK.B, ISRG, MA, AMZN
Valueact:
Top Buys: A, MDLN, AJG, RYAN, MSI
Top Sales: RKT, TOST, RBLX, CBRE, KKR
Vanguard:
Top Buys: MSFT, META, PLTR, CVX, AMZN
Top Sales: SNDK, GE, GEV, UNH, EA
Verition:
Top Buys: WBD, MDT, IVV, GLD p, GOOGL c
Top Sales: QQQ p, SMH p, EA, SPY, SEM
Viking Global:
Top Buys: MA, HONA, TSLA, GOOGL, TSCO
Top Sales: SCHW, JCI, UNP, MSFT, V
Voya:
Top Buys: MSFT, META, PLTR, TMO, LITE debt
Top Sales: GOOGL, MU, CAT, STX debt, GOOG
Walleye:
Top Buys: QQQ p, NVDA c, GOOGL p, SPY c, GOOGL c
Top Sales: IVV, SPLG, JPM, IBIT, DISH debt
Wellington:
Top Buys: MSFT, META, AAPL, JPM, AMZN
Top Sales: KLAC, XOM, TSLA, GLW, DHR
Woodline:
Top Buys: IWM, MSFT, SOXX, AAPL, NVDA
1
630
A thesis can be right while the near-term timing is wrong.
A lot of people on X are confused right now, wondering how record earnings and massive CapEx commitments are resulting in an AI washout. Everyone wants to blame Wall Street FUD or short reports, but that is missing the forest for the trees. This is purely about the macro risk environment.
When the Strait of Hormuz closed during the Iran war, it sent oil surging and pushed the 10-year Treasury yield straight into the 5% range. Institutional capital operates on risk models and cost of capital. When oil spikes and yields hit 5%, funds are forced to de-risk out of high-beta tech and park cash in defensive bond proxies. That is the exact reason Costco and Walmart are trading at 43x to 49x earnings right now.
If the 10-year were sitting in the low 4s and oil were trading between $60 and $70, no amount of "peak CapEx" FUD or sell-side revisions would be able to slow down the capital flooding into high-beta AI infrastructure names.
The good news is that time is on our side. Hyperscaler CapEx is still set to surge into 2027 and climb again in 2028. Have patience in your work, make sure you own names with sold-out capacity and locked-in 2027/2028 revenue visibility, and wait for macro conditions to ease so you can reap the rewards in 2027 and 2028.
2
5
703
This is the exact playbook we saw in 2022 during the start of the Russia/Ukraine war. Oil spiked to $130, inflation expectations surged, and the 10-year yield more than doubled, forcing institutional risk models to dump semis and hide in defensive staples.
The key difference is that 2022 also had a post-COVID PC and gaming inventory glut, whereas today data center and optical capacity is completely sold out into 2027.
Back then, yields didn't even need to crash for tech to bottom. Once oil rolled over and the 10-year simply stabilized, capital flooded out of low-growth defensives and kicked off the massive 2023-2024 semis run.
2
246
If it wasn't already obvious from my posts, I am quadrupling down on the AI infrastructure trade.
Q3 earnings validated the physical bottleneck thesis across the board. With 2027 capacity already spoken for, time is on our side while we wait for the macro environment to turn.
Time will tell if I’m early or just foolish.
2
534
Everyone will be glued to $NVDA's call tomorrow for CEO succession updates and headline GPU numbers. If you own the upstream optical, laser, and test bottlenecks ($LITE, $COHR, $AAOI, $SIVE, $AEHR, $AXTI), the C-suite drama and top-line beats are secondary. Three specific data points I am watching tomorrow:
1. Networking vs. Compute Growth Inside the Data Center Segment ($LITE, $COHR, $AAOI): Colette Kress breaks out Networking revenue every quarter, so watch its sequential growth relative to Compute along with any commentary on 800G and 1.6T transceiver adoption alongside the Rubin ramp. As Spectrum-X and InfiniBand scale out, optical interconnect spend is outpacing GPU spend per rack, which feeds directly into $AAOI's hyperscale transceiver qualifications and internal InP laser ramp for 2027.
2. Spectrum-X Photonics and Optical Scale-Up Commentary ($LITE, $COHR, $SIVE, $AEHR): Last month at GTC Berlin, Nvidia announced plans to expand optics capacity 15x in 2027 to meet Co-Packaged Optics (CPO) demand after putting $4B into LITE andCOHR earlier this year. Listen for any Q&A or prepared remarks confirming that optical scale-up is moving into 2027/2028 cluster architectures. Moving from pluggable copper to CPO triggers two massive downstream bottlenecks: high-power CW laser arrays acting as external light sources (validating the 6-inch InP wafer allocation SIVE is ramping), and mandatory 100% wafer-level burn-in since a single failing laser ruins a co-packaged assembly (the exact catalyst drivingAEHR's silicon photonics backlog).
3. Non-Cancellable Purchase Commitments and Long-Term Supply Prepayments ($LITE, $COHR, $AXTI, $AAOI): Ignoring the street's "peak CapEx" chatter, another sequential jump in locked-in supply commitments in the CFO commentary and balance sheet footnotes confirms that physical component capacity, starting at the bottom of the funnel with InP substrates ($AXTI) and vertically integrated laser fabs ($LITE, $COHR, $AAOI), remains the real gating factor for 2027 deliveries.
Financial media will spend the next 48 hours obsessing over the CEO transition and headline GPU margins, while the real asymmetric alpha sits in reading how Nvidia's networking and supply numbers flow down to the physical optical bottlenecks.
Everyone is panicking over the SemiAnalysis report slashing AI CapEx by 19% in 2027, 28% in 2028, 41% in 2029, and 40% in 2030. Nobody is reading what they actually cut.
First, those numbers only cover AI accelerators (GPUs and custom ASICs), not total data center spend. Even after the revision, pure chip spend still grows 60% YoY in 2027 ($392B to $627B), rises another 21% in 2028 ($758B), and reaches $796B in 2029 and $923B in 2030 (up 135% from 2026). Add networking, optics, memory, power, and shells, and their total 2027 CapEx projection still lands in the $1.3T to $1.5T range before decelerating in 2028.
Second, why did 2029 and 2030 dollar figures drop 40% while cumulative 2026 to 2030 silicon wafer area was only cut by 17%? Chip price deflation, not demand destruction.
66% of the entire wafer reduction and a 66% cut to 2030 compute revenue falls on $NVDA (including lower assumed pricing on future Feynman chips) as hyperscalers pivot to cheaper custom ASICs ($AVGO cumulative revenue was cut just 11%, while $AMZN Trainium held steady and $AMD MI400/MI450 pulled forward). Custom ASIC clusters have lower compute density per package than NVLink racks, meaning they require significantly more optical transceivers and lasers per FLOP to link chips together. Cheaper silicon in 2029 and 2030 just shifts wallet share from Nvidia's 75% gross margins into optics, lasers, and networking.
Third, cutting 2029 and 2030 power demand by ~33% (to 72.5 GW and 80.7 GW) alongside 5.8 GW of post-midterm permitting delays and tighter financing for uncontracted startups can actually be healthy for the industry/build out. It kills speculative science projects, prevents a 2000-style dark fiber glut, stretches the deployment cycle through 2030, and makes already-permitted, energized megawatts way more valuable.
Pure chip spend is still set to more than double between 2026 and 2030 while shifting toward optical-heavy custom silicon.
$POET Q3 numbers are out. It is a mixed print on near-term timing, yet a strong validation for the 2027/2028 upstream wafer and optical bottleneck thesis.
Revenue crossed $1M for the first time at $1.03M (+80% QoQ, +244% YoY), missing the ~$1.6M consensus because the 30,000 optical engines targeted for 2026 shifted to Q2 2027. Management cited new-tech packaging yield ramps, module design cycles, and industry-wide wafer and component shortages. Also note the EPS beat was helped by a $6.4M non-cash warrant revaluation gain.
Still, the 2027/2028 buildout got much more concrete. They secured tools to expand Malaysia foundry capacity 10x from ~120 to 1,250–1,500 wafers/month in 2027, alongside a >$45M Penang equipment buildout targeted to be manufacturing-ready by mid-2027. Four customers have placed orders >$1M each, Q4 shipments will exceed Q1–Q3 combined, and 7 of 9 active customers enter production in 2027 as 1.6T ramps. They also landed a PO from a second Tier 1 laser maker post-Q3, with Blazar ELS sampling in H1 2027 for early 2028 production, while hiring a new CFO and moving to U.S. GAAP reporting with live calls starting March 1, 2027.
POET ended Q3 with $774.3M in cash ($12.7M operating burn, partially offset by ~$5M in interest income). Against an ~$888M market cap, the market values the operating business and IP at just ~$114M EV.
The biggest read-through is management blaming the delay on "supply chain tightening, wafer and component shortages." As a downstream engine integrator, $POET is feeling the squeeze of the exact wafer and laser scarcity that upstream fabs ($LITE, $COHR, $AXTI, $AAOI, $SIVE, $TSEM, etc) are monetizing today.
At a ~$114M EV with $774M cash, the downside is cushioned, though the Q2 2027 backlog shift makes it a late-2027 to 2028 story while upstream laser and substrate fabs already have the sold-out capacity and pricing power now.
$POET reports Q3 results tomorrow at 7 AM ET, and it remains one of the most polarizing stocks on X.
Stripping away the hype and the hate, the balance sheet floor is real. After raising $400M in May at $21/unit (incredible timing by management with the stock now around $5.13), POET ended Q2 with $796.3M in cash against an ~$888M market cap. The market is valuing the actual operating business and optical interposer IP at roughly $90M in true Enterprise Value, while quarterly interest income ($4.3M in Q2) offsets more than a third of their operating cash burn ($12.2M).
Still, cash in the bank doesn't drive a sustained rerating on its own. Tomorrow comes down to one thing: bridging the gap between NRE science projects and binding commercial production.
Q2 revenue was just $570k (mostly non-recurring engineering fees), with consensus around ~$1.6M for Q3. Earlier this year, POET highlighted a massive headline pipeline: a $50M initial order from Lumilens (within a $500M+ five-year framework), a $2.4M post-Q2 purchase order, an external light source (ELS) deal with a Tier 1 laser company, and module engagements with Foxconn Interconnect, Luxshare, and Mitsubishi Electric. However, in August management promised "additional announcements and updates in September" alongside H2 2026 production qualification shipments, and both September and October went by without a single commercial update.
For $POET to graduate from a speculative call option into a high-conviction 2027 bottleneck play, tomorrow's MD&A needs to show:
1. Hard evidence that the $2.4M order and H2 2026 qualification units are actually shipping, pushing revenue out of the sub-$1M NRE range.
2. Concrete delivery timelines for the $50M Lumilens order and progress on the $50M equipment buildout to scale Malaysia capacity toward 1M optical engines/month by late 2027.
3. Clarity on the Tier 1 laser partner and whether their Blazar hybrid ELS is hitting milestones for 2028 CPO deployments after losing Celestial AI to the Marvell buyout.
At a ~$90M EV, the downside is cushioned by the cash balance and the warrant overhang is mostly way out of the money (like the 19M warrants at $26.25 from the May raise). Now management has to prove that conditional orders are converting into binding 2027 shipping schedules.
This is the exact setup we saw in late 1998 and the 1972 Nifty Fifty era.
Whenever macro anxiety or "peak CapEx" chatter flares up, institutional money doesn't actually go to cash. Because long-duration Treasuries have lost their hedging power with the 10-year surpassing 5%, funds hide in consumer staples and retail monopolies as equity bond proxies.
The problem is that at 43x to 49x earnings for ~10% growth, you aren't buying safety. A 49x multiple on $COST is a ~2.0% forward earnings yield when risk-free 2-year paper pays 4.74%. In 1998, Coca-Cola ($KO) and Walmart ($WMT) got bid up to the exact same 45x to 50x multiples during the macro scare. Once the market realized the telecom and internet buildout was still accelerating, institutions dumped those "safe" staples (which then went sideways for over a decade as their multiples compressed) and used them as an ATM to fund the final explosive tech infrastructure melt-up.
A lot of people look at sticky inflation, 5% yields, and these stretched defensive multiples and assume we need a 20%+ index bear market right now to reset the board. Looking at the data, a full bear market before the rotation back into AI infrastructure is unlikely:
1. Credit and growth aren't breaking. High-yield credit spreads just tightened to 320 bps even with the 10-year at 5%, and Q3 GDP is tracking at 3.4%. You don't get a systemic bear market without a credit crunch or an earnings recession, and Nasdaq-100 earnings are growing ~50% YoY.
2. Hyperscaler CapEx isn't rate-sensitive. A 5% 10-year yield exposes a consumer staple trading at a 2% earnings yield, but it doesn't stop cash-rich hyperscalers from funding $1.5T+ in 2027 infrastructure, and it doesn't cancel $320B+ in binding data center leases or upfront fab prepayments.
We already lived through a stealth bear market inside high-beta AI and optical names over the last few weeks while the headline indexes were propped up by 45x P/E defensives.
As 2027 CapEx numbers ($1.5T+, up 80%+ YoY) start hitting income statements, underweight institutional managers will be forced to sell these crowded, slow-growth "safe havens" to chase contracted 50% to 100%+ earnings growth in physical AI bottlenecks (lasers, substrates, foundries, memory, and neoclouds) trading at compressed multiples.
The real cyclical bear market is a mid-late 2028 story, not a 2026 or 2027 story. It happens after this rotation plays out and 2027 AI euphoria peaks, right as hyperscaler CapEx growth naturally decelerates from ~80% in 2027 down to ~30% in 2028 and the cumulative weight of higher rates finally catches up to the broader economy.
Right now, true safety isn't paying 49x for 10% growth, it is owning sold-out 2027 capacity at a massive execution buffer.
WE ARE IN A "SAFE-HAVEN" STOCK BUBBLE.
$COST — ~49× (Historically extreme valuation)
$WMT — ~43× (Near 30-year highs)
$AAPL — ~38× (Near 20-year highs)
$MNST — ~37× (Near all-time highs)
$RSG — ~33× (Near all-time highs)
$KO — ~29× (Near 25-year highs)
$JNJ — ~26× (Near 25-year highs)
$BRK.B — ~26× (Near 25-year highs)
$MCK — ~23× (Near 25-year highs)
$ABBV — ~19× (Near 10-year highs)
*Approximate forward P/E ratios.
Here's the interesting part:
These 10 companies are projected to grow EPS by just ~10% on average next fiscal year.
Investors are paying historically extreme valuation premiums for the perception of safety.
Meanwhile, the Nasdaq-100, packed with high-growth tech companies, is nearing COVID-era forward P/E lows...
While high-growth tech stocks trade at unjustifiably low multi-year valuations, Wall Street props up the indexes with historically expensive defensive stocks.
Why is nobody talking about this???









