Investment Intelligence Lab

Naina Garg

Markets and AI, plus deep-tech stocks and deals.

Independent, dated, and auditable. Search a ticker or browse below.

01Sister site Updated daily

Markets & AI

The daily read on markets and AI, published on the sister site.

Read the latest
02On this site 39 tech dashboards

Deep Tech Equity Research Library

Stocks

Interactive research dashboards with live prices, editable valuation models, and scenario tools.

Open the library
03On this site 21 deals

Deals

Landmark technology and private-equity deals, examined across three regions.

Explore the deals
Live dashboards

39 Tech Dashboards

Ticker Company Category Rating Price 12-mo PT Δ vs PT As of Next Release
AAPLApple Inc.Edge & On-Device AIHold$317$330+4.1%Aug 2026
Oct 29, 2026
Q4 FY26
AMZNAmazon.com, Inc.HyperscalersBuy$272$310+14.0%Aug 2026
Oct 29, 2026
Q3 CY26
GOOGLAlphabet Inc.HyperscalersBuy$362$420+16.0%Aug 2026
Oct 28, 2026
Q3 CY26
MSFTMicrosoft CorporationHyperscalersBuy$500$560+12.0%Aug 2026
Oct 28, 2026
Q1 FY27
CRDOCredo Technology Group Holding LtdNeoclouds & ConnectivityBuy$251$280+11.6%Jun 2026
Sep 2, 2026
Q1 FY27
CRWVCoreWeave, Inc.Neoclouds & ConnectivityBuy$94$125+33.0%Jun 2026
Aug 11, 2026
Q2 2026
GEVGE Vernova Inc.Power, Grid & AI ServersHold$988$1,080+9.3%Aug 2026
Oct 21, 2026
Q3 2026
SMCISuper Micro Computer, Inc.Power, Grid & AI ServersHold$26$34+30.8%Jul 2026
Aug 11, 2026
Q4 FY26
AVGOBroadcom Inc.Silicon & FoundriesBuy$365$461+26.3%Jun 2026
Sep 2, 2026
Q3 FY26
NVDANVIDIA CorporationSilicon & FoundriesBuy$219$250+14.2%Aug 2026
Aug 26, 2026
Q2 FY27
TSMTaiwan Semiconductor (TSMC)Silicon & FoundriesBuy$423$490+15.8%Aug 2026
Oct 15, 2026
Q3 2026
AMDAdvanced Micro Devices, Inc.Silicon & FoundriesHold$483$560+15.9%Aug 2026
Nov 3, 2026
Q3 2026
ASMLASML Holding N.V.Silicon & FoundriesHold$1,741$1,740-0.1%Aug 2026
Oct 14, 2026
Q3 2026
BESIBE Semiconductor Industries N.V.Silicon & FoundriesHold€221€250Aug 2026
Oct 22, 2026
Q3 2026
CBRSCerebras Systems, Inc.Silicon & FoundriesHold$188$185-1.6%Jul 2026
Aug 12, 2026
Q2 2026
INTCIntel CorporationSilicon & FoundriesHold$102$110+7.8%Aug 2026
Oct 22, 2026
Q3 2026
MRVLMarvell Technology, Inc.Silicon & FoundriesHold$267$255-4.5%Jun 2026
Aug 27, 2026
Q2 FY27
MTSIMACOM Technology Solutions Holdings, Inc.Silicon & FoundriesHold$311$360+15.8%Aug 2026
Nov 5, 2026
Q4 FY26
SMTCSemtech CorporationSilicon & FoundriesHold$138$150+8.7%Jul 2026
Sep 2, 2026
Q2 FY27
CRWDCrowdStrike Holdings, Inc.Endpoint SecurityHold$195$205+5.1%Jul 2026
Aug 26, 2026
Q2 FY27
NETCloudflare, Inc.Network SecurityHold$300$300+0.0%Aug 2026
Nov 5, 2026
Q3 2026
PANWPalo Alto Networks, Inc.Network SecurityHold$384$365-4.9%Aug 2026
Sep 1, 2026
Q4 FY26
ZSZscaler, Inc.Network SecurityHold$147$160+8.8%Jul 2026
Sep 3, 2026
Q4 FY26
000660SK hynix Inc.Memory & StorageBuy₩1.42M₩2.1MAug 2026
Oct 27, 2026
Q3 2026
MUMicron TechnologyMemory & StorageBuy$967$1,400+44.8%Aug 2026
Sep 29, 2026
Q4 FY26
005930Samsung Electronics Co., Ltd.Memory & StorageHold₩231,000₩283,000Aug 2026
Oct 29, 2026
Q3 2026
CMXTChangXin Memory TechnologiesMemory & StorageHold¥49¥45Jul 2026
Aug 31, 2026
H1 2026 report
SNDKSandisk CorporationMemory & StorageHold$1,212$1,350+11.4%Aug 2026
Nov 5, 2026
Q1 FY27
PLTRPalantir TechnologiesAI PlatformsHold$172$175+1.7%Aug 2026
Nov 2, 2026
Q3 2026
AIC3.ai, Inc.Enterprise AI AppsHold$9$9+0.0%Jul 2026
Sep 9, 2026
Q1 FY27
NTNXNutanix, Inc.Infrastructure SoftwareHold$54$58+7.4%Jul 2026
Aug 26, 2026
Q4 FY26
NOWServiceNow, Inc.Systems-of-RecordBuy$125$135+8.0%Aug 2026
Oct 28, 2026
Q3 2026
ORCLOracle CorporationSystems-of-RecordBuy$143$190+32.9%Aug 2026
Sep 8, 2026
Q1 FY27
IONQIonQ, Inc.Quantum ComputingBalanced$43$40-7.0%Aug 2026
Nov 4, 2026
Q3 2026
QBTSD-Wave Quantum Inc.Quantum ComputingBalanced$21$15-28.6%Aug 2026
Nov 5, 2026
Q3 2026
RGTIRigetti Computing, Inc.Quantum ComputingBalanced$17$13-23.5%Aug 2026
Nov 10, 2026
Q3 2026
MBLYMobileye Global Inc.Robotics & Physical AIHold$9$10+11.1%Aug 2026
Oct 22, 2026
Q3 2026
TSLATesla, Inc.Robotics & Physical AIHold$320$370+15.6%Aug 2026
Oct 21, 2026
Q3 2026
SPCXSpace Exploration Technologies Corp.Space & LaunchHold$133$135+1.5%Aug 2026
Nov 10, 2026
Q3 2026
Cross-coverage map

Thesis Graph: How It Connects

The sector deep-dive · follow the ~$725B

The Library: Five Expressions of One Trade

The hyperscalers’ ~$725B AI build-out, sampled across the risk spectrum. Pick a sector below, or read the cross-sector Synthesis.
Sector I

AI Infrastructure & Compute

Hyperscaler-4 capex 2026E
~$725B
+77% YoY
Global DC power demand 2026E
~132 GW
+27% (Gartner)
Cloud infra spend
+35% YoY
>$500B run-rate
TSM share of leading-edge
~90%
~99% frontier-AI silicon

Hyperscaler capex: 2025 → 2026E (US$ bn)

The read

The single largest capital-formation event in technology: ~$725B of 2026 hyperscaler capex (+77% YoY) pulling an entire stack to its physical limits: the silicon that computes (NVDA, AVGO, AMD, MRVL, TSM, ASML, INTC), the neoclouds and connectivity that rent it (CRWV, CRDO), the hyperscalers that buy it (MSFT, GOOGL, AMZN), and the power and servers that run it (GEV, SMCI). The bottleneck has migrated from chips to memory to electricity. The trade is to own the toll-roads and watch where the build over-runs first.
What matters now
The toll-roads are owned, the challengers are priced. The build must route through TSM (~90% of leading-edge, ~99% of frontier-AI silicon) and ASML’s EUV monopoly; NVDA stays the capstone Buy ($250) with NTM P/E in the 11th percentile while FY28 consensus migrated $290B→$485B. The re-rated challengers (AMD Hold $560 after +264%, MRVL Hold $255 after +231% YTD) already price the MI450/custom-ASIC ramp; AVGO’s XPU franchise (>$100B FY27 AI) is the hyperscalers’ hedge against paying NVDA’s margin.
The capex is settled; the return on it is not. Cloud is reaccelerating (+35% YoY to a >$500B run-rate) but FCF is collapsing under the build. AMZN’s TTM FCF has turned slightly negative and our base underwrites a negative 2026 print. The leveraged expression is the rental layer: CoreWeave’s ~$99B contracted RPO (98% take-or-pay) is the bull case, ~$24.9B debt at ~640% D/E the bear, and Credo’s AECs/optical DSPs inflect with every GPU generation.
Power is the binding constraint. DC power demand +27% to ~132 GW while gas-turbine and transformer slots sell out 4–5 years deep. The gate is generation and interconnect, not wafers. GE Vernova’s $176B backlog (~100 GW of turbines, large-frame sold into 2028) reprices on scarcity (Hold $1,080, ~43x fwd EBITDA already prices the supercycle), and Super Micro is the purest AI-server beta on the same demand.
19 dashboards

Silicon & Foundries

AVGONASDAQ
Broadcom Inc.
Semiconductors / Infra Software

AI custom-silicon franchise (Google, Meta, Anthropic, OpenAI XPUs) + VMware annuity. FY27 AI guided >$100B. SOTP $461 PT.

datacentersinfrastructure softwareAI networkingVMware
Price
$365
Rating
Buy
12-mo PT
$461
PT as of Jun 2026 Excel model
MRVLNASDAQ
Marvell Technology, Inc.
Custom Silicon / Compute Silicon & Foundries

The premier AI custom-silicon + optical franchise. Forward EV/Sales scenarios. Hold $255. A great franchise, but fairly-to-fully valued after a +231% YTD run; head-to-head analysis vs AVGO.

custom siliconAWS TrainiumMicrosoft MaiaNVLink Fusion
Price
$267
Rating
Hold
12-mo PT
$255
PT as of Jun 2026 Excel model
MTSINASDAQ
MACOM Technology Solutions Holdings, Inc.
Analog / RF & Optical Semiconductors

A differentiated analog/RF + optical franchise with a real AI-optical tailwind (Data-Center growth raised to ~74%). Forward EV/Sales scenarios. Hold $360 at ~40x fwd EPS, still below the Street’s ~$396.

optical semiconductorsInP photonicslaser drivers / TIAGaN
Price
$311
Rating
Hold
12-mo PT
$360
PT as of Aug 2026
SMTCNASDAQ
Semtech Corporation
Analog / Connectivity & IoT Semiconductors

Best-in-class power-efficient AI-connectivity ramp (CopperEdge active-copper cables, 1.6T toward 3.2T, a hyperscaler design win, data center +58% in FY26 and guided >50% in FY27), but the multiple already prices near-flawless execution. Hold, base PT $150: balanced risk/reward at ~$138 after a +157% 52-week run.

analogCopperEdge active copperAI interconnectLoRa / LoRaWAN
Price
$138
Rating
Hold
12-mo PT
$150
PT as of Jul 2026 Excel model
NVDANASDAQ
NVIDIA Corporation
Semiconductors / Compute Silicon & Foundries

The capstone of AI compute. Operating story accelerating, multiple decelerating in opposite directions. EV/Sales scenarios. Buy $250. NTM P/E at 11th %ile of 10-yr range while FY28 consensus migrated $290B → $485B.

Vera RubinBlackwellHBM4Spectrum-X
Price
$219
Rating
Buy
12-mo PT
$250
PT as of Aug 2026 Excel model
CBRSNASDAQ
Cerebras Systems, Inc.
Semiconductors / AI Accelerators

The wafer-scale bet on inference speed. Fastest silicon for memory-bound inference (>2,200 tok/s on Llama 70B), a $24.6B OpenAI-anchored backlog, and FY25 revenue of $510M (+76%). The catch: ~86% of 2025 revenue from two UAE-linked related parties, gross margin guided down to 36-38%, a ~$55B cap at ~104x sales, and NVIDIA’s CUDA moat. EV/Sales scenarios. Hold, $185 fair value (Bull $300 / Bear $110), below the ~$292 Street average.

wafer-scaleWSE-3AI inferenceOpenAI
Price
$188
Rating
Hold
12-mo FV
$185
PT as of Jul 2026
TSMNYSE (ADR)
Taiwan Semiconductor (TSMC)
Semiconductors / Compute Silicon & Foundries

A toll on the AI build-out: ~90% of leading-edge and ~99% of frontier-AI silicon, at record 68% gross / 56% net margins. Buy $490 (DCF + reverse DCF). The honest catch: the DCF says ~fairly valued (~$425), so this is a quality-and-cycle Buy, not deep value; the Taiwan tail + a pending ITC ruling + ~44%-of-revenue capex are the bear.

foundryleading-edgeN2CoWoS
Price
$423
Rating
Buy
12-mo PT
$490
PT as of Aug 2026 Excel model
AMDNASDAQ
Advanced Micro Devices, Inc.
Semiconductors / Compute Silicon & Foundries

The credible #2 in AI accelerators, now fully re-rated. SOTP by segment. Hold $560 after a ~126% YTD run. The OpenAI + Meta 6GW Instinct deals and record ~46% EPYC server share are real, but ~65x FY26 earnings and ~40x forward Data-Center EBITDA price the MI450/Helios ramp in; ~20% warrant dilution + NVDA dominance are the checks.

Instinct MI350MI400 HeliosEPYCROCm
Price
$483
Rating
Hold
12-mo PT
$560
ASMLNASDAQ
ASML Holding N.V.
Semiconductors / Lithography & Capital Equipment

The EUV lithography monopoly. SOTP by value pool. Hold $1,740. The sole EUV/High-NA supplier, with a recurring service annuity and a EUR 44-60B 2030 runway, but at ~44x forward earnings we see it fairly valued at $1,740, with the Street (~$2,100) more bullish; a China DUV-servicing ban the swing risk and a conservative DCF nearer $1,000.

EUVHigh-NADUVlithography monopoly
Price
$1,741
Rating
Hold
12-mo PT
$1,740
PT as of Aug 2026 Excel model
BESIEuronext Amsterdam
BE Semiconductor Industries N.V.
Semiconductors / Advanced Packaging Equipment

Hold €250: the precision leader in hybrid-bonding equipment, the only viable path to sub-10µm 3D interconnect, and a genuine AI advanced-packaging chokepoint (Applied Materials co-develops the Kinex bonder and holds ~9%). After a pullback to ~€221 the multiple has compressed to ~40x forward earnings; we trim to €250 (from €280) on a more conservative multiple as the HBM catalyst slips to 2027+ and the near-term ramp stays logic-led (TSMC SoIC). Our €250 sits ~10% below the Street median.

hybrid bondingadvanced packagingdie attachTSMC SoIC
Price
€221
Rating
Hold
12-mo PT
€250
PT as of Aug 2026 Excel model
INTCNASDAQ
Intel Corporation
Semiconductors / Compute Silicon & Foundries

Turnaround, still unproven. After a ~2.8x 2026 run to ~$102 on foundry catalysts (Google >3M-TPU order, Apple/Nvidia 18A/14A, 18A-P on track) plus a big Q2 revenue beat, Hold $110 (trimmed from $135). The $110 SOTP base implies only ~8% upside and sits in line with the Street (~$115); we value Foundry at a TSM-like ~10.3x EV/Sales, not a scarcity premium, while its ~$2.1B/qtr losses and unproven external volume keep us neutral. Foundry ~55% of EV; a fairly-valued, two-sided Hold, not conviction.

x86Intel Foundry18APanther Lake
Price
$102
Rating
Hold
12-mo PT
$110
PT as of Aug 2026 Excel model

Neoclouds & Connectivity

CRWVNASDAQ
CoreWeave, Inc.
AI Infrastructure / Neocloud GPU-as-a-Service

NVIDIA-preferred neocloud. $99.4B backlog (8x FY26 rev, 98% take-or-pay), 11.5% NVIDIA stake + $6.3B capacity backstop, first-to-GA every NVIDIA generation. EV/Sales scenarios. Buy $125. Cheapest neocloud despite largest backlog, but $24.9B debt / 640% D/E warrants smaller sizing.

neocloudGPU cloudNVIDIA preferredVera Rubin
Price
$94
Rating
Buy
12-mo PT
$125
PT as of Jun 2026 Excel model
CRDONASDAQ
Credo Technology Group Holding Ltd
AI Connectivity Silicon / AI Infrastructure

AI back-end interconnect pure-play. EV/Sales scenarios. Buy $280. FY26 revenue +206% YoY to $1.335B, FY27 mgmt commentary >80% growth on optical DSP + AEC ramp.

AI interconnectActive Electrical Cablesoptical DSPSerDes
Price
$251
Rating
Buy
12-mo PT
$280
PT as of Jun 2026 Excel model

Hyperscalers

GOOGLNASDAQ
Alphabet Inc.
Hyperscalers / Public Cloud

Search + YouTube + Google Cloud + AI; the Gemini/TPU AI hyperscaler with DOJ Search remedies as the binary risk

SearchYouTubeGoogle CloudGemini
Price
$362
Rating
Buy
12-mo PT
$420
PT as of Aug 2026 Excel model
AMZNNASDAQ
Amazon.com, Inc.
Hyperscalers / Public Cloud

A sum-of-the-parts where AWS is ~58% of value on ~21% of revenue. Buy $310. AWS reaccelerating (+37%) + a high-margin advertising annuity + a market-multiple retail base, less net debt; the swing risk is the free-cash-flow drought under ~$220B AI capex.

AWSadvertisinge-commercesum-of-the-parts
Price
$272
Rating
Buy
12-mo PT
$310
PT as of Aug 2026 Excel model
MSFTNASDAQ
Microsoft Corporation
Hyperscalers / Public Cloud

The Azure AI compounder; OpenAI exclusivity + Copilot enterprise attach driving cloud + software growth

AzureOpenAICopilotAI infrastructure
Price
$500
Rating
Buy
12-mo PT
$560
PT as of Aug 2026 Excel model

Power, Grid & AI Servers

GEVNYSE
GE Vernova Inc.
Energy & Electrification / Power Generation

The arms-dealer of the AI power supercycle: gas turbines (100GW backlog, large-frame slots sold out into 2028) + grid electrification + nuclear/SMR. Record $176B backlog, FCF inflecting to $11.5–12.5B. Forward EV/EBITDA scenarios. Hold $1,080. Best-in-class franchise, but ~43x forward EBITDA already prices much of the supercycle; accumulate on pullbacks.

AI powergas turbinesgridnuclear SMR
Price
$988
Rating
Hold
12-mo PT
$1,080
PT as of Aug 2026 Excel model
SMCINASDAQ
Super Micro Computer, Inc.
AI Servers / Networking & Server Hardware

Public-market purest AI-server bet, repriced. EV/Sales scenarios. Hold $34. Revenue scaled to ~$40B but margins swung to a 6.3% trough, ~$7.5B net debt, and a Taiwan chip-smuggling probe cap the multiple.

AI serversGB200 NVL72direct liquid coolinghyperscaler
Price
$26
Rating
Hold
12-mo PT
$34
PT as of Jul 2026 Excel model

Edge & On-Device AI

AAPLNASDAQ
Apple Inc.
Consumer Hardware + Services / Edge AI

The best consumer franchise in tech, fully valued. Record Q3 FY26 ($109.4B rev +16%, EPS $2.02) and a re-rate on the ~$1B/yr Google Gemini–Siri deal, but guided Q4 to a soft +9–11% at ~34x FY27E. Hold $330 (~consensus): a ~2.5B-device installed base + $109B Services annuity (~75% GM) + ~$137B FCF + $100B buyback, against a two-sided setup, the ~$20B Google-TAC appeal and China/Huawei share. Own to compound; no margin of safety at the multiple.

iPhoneServicesApple Intelligenceon-device AI
Price
$317
Rating
Hold
12-mo PT
$330
PT as of Aug 2026

On the radar: AMAT, NBIS, IREN, APLD, WULF, CIFR, HUT, BITF, META, VRT, ETN, PWR, CEG, VST, SMR, CSCO, ANET, DELL, HPE, QCOM.

Silicon & Foundries
Neoclouds & Connectivity
Hyperscalers
Power, Grid & AI Servers
Edge & On-Device AI
See these names marked to market
Sector II

Cybersecurity

Global infosec spend 2026E
~$244B
+13.3% (Gartner)
Cloud security growth
+28.8%
fastest subsegment
PANW NGS ARR
$8.1B
+60% rep / +28% org
CRWD vs PANW
~21x / 14x
NTM EV/Sales

Security platform scale: ARR (US$ bn)

The read

Security spend is re-accelerating (~$244B in 2026, +13.3%) and consolidating. CISOs are collapsing dozens of point tools into a handful of suites, and that consolidation dividend accrues to the two platform franchises here: CRWD from endpoint outward (Falcon + Charlotte AI) and PANW from network and identity inward (platformization + CyberArk + Cortex). The shared question, extending to the SSE pure-play ZS (zero-trust leader, ~56% off its high) and the edge premium NET (Cloudflare, priciest in the set at ~24x sales), is whether any can defend a premium multiple against a Microsoft that bundles “good-enough” security free inside E5.
What matters now
Platformization is the whole bull case, and the multiple debate. PANW’s ~1,550 platformized customers and $8.1B NGS ARR (+60% reported / +28% organic) validated the consolidation play, but after a +117% YTD run PANW now trades at ~22x fwd EV/Sales; CRWD prints a record Q1 net-new ARR ($256M) and $5.5B+ ARR at ~35x NTM EV/Sales. We rate PANW Hold ($365) after the re-rate, CRWD Hold ($205), the SSE pure-play ZS Hold ($160), and the edge premium NET Hold ($250) on those valuation and growth-durability gaps, not on franchise quality.
Microsoft is the structural bear for the whole group. Defender bundled into ~75% of F500 E5 licenses caps mid-market wallet expansion and pressures pure-play pricing: the single risk that travels across both names. The defense is depth and switching cost (Falcon’s single agent; PANW’s Strata/Prisma/Cortex breadth); if Defender closes the endpoint-AI gap, mid-market churn toward ‘free-with-E5’ structurally caps multiples.
AI is reshaping the moat both ways. The next leg is the agentic SOC. Attacks scale (deepfakes, autonomous intrusion) while defense moves to agents: PANW’s Cortex AgentiX and CRWD’s Charlotte AI are optionality the Street isn’t fully paying for. With enterprises still spending ~17x more on AI tooling than on securing AI, ‘securing AI’ is a coming budget line that favors scaled platforms; M&A (Google-Wiz $32B, PANW-CyberArk $25B) keeps concentrating the category.
4 dashboards
CRWDNASDAQ
CrowdStrike Holdings, Inc.
Cybersecurity / Endpoint & XDR Platform

Franchise firing, multiple math-constrained. EV/Sales scenarios. Hold $205 (post 4:1 split). Q1 FY27 ARR $5.51B +24%, record FCF, but ~39x sales and ~150x forward earnings at all-time highs cap the upside.

EDRXDRFalcon platformChannel File 291 outage
Price
$195
Rating
Hold
12-mo PT
$205
PT as of Jul 2026 Excel model
PANWNASDAQ
Palo Alto Networks, Inc.
Cybersecurity / Network & Platform Security

Platformization tipping point + CyberArk identity flank + Cortex AgentiX agentic SOC. EV/Sales scenarios. Hold $365 after a +117% YTD run through the old $310 target. ~1,550 platformized customers, NGS ARR $8.1B (+60% reported / +28% organic); ~22x FY27E EV/Sales now full, upside is the Cortex AgentiX optionality.

NGFWSASEPrismaCortex XSIAM
Price
$384
Rating
Hold
12-mo PT
$365
PT as of Aug 2026 Excel model
ZSNASDAQ
Zscaler, Inc.
Cybersecurity / Zero Trust & SSE

The SSE / zero-trust leader, ~56% off its high after a 16-17% FY27 preview and a sales-leadership exodus. Hold $160. Still compounds ARR at 25% ($3.53B) with ~27% FCF margins and ~$1.7B net cash, but organic net-new ARR is only +14%, Red Canary props the headline, and Palo Alto/Microsoft bundling caps the pure-play premium at ~6.9x EV/Sales.

SSESASEZero Trust ExchangeZIA
Price
$147
Rating
Hold
12-mo PT
$160
PT as of Jul 2026
NETNYSE
Cloudflare, Inc.
Cybersecurity / Edge, SASE & Developer Platform

The programmable edge (CDN + security + Cloudflare One SASE + Workers/Workers AI), priced for perfection. Hold $300. Elite quality (Q2 revenue +36%, DBNRR 120%, RPO +35%, FCF inflecting) but ~28x forward sales with the Street (~$250) now below spot. The 20% agentic-AI-first restructuring is a margin tailwind and an execution risk. A great company at a full price.

edge networkCDNCloudflare OneSASE
Price
$300
Rating
Hold
12-mo PT
$300
PT as of Aug 2026

On the radar: S.

See these names marked to market
Sector III

Memory Super-Cycle

HBM market 2026E
~$54.6B
+58% (BofA)
DRAM / NAND price rise
+125% / 234%
2026E (Gartner)
MU gross margin
~85%
Q3 FY26
AI share of DRAM output
~20%
HBM crowds out the rest

HBM market share (2025)

The read

The memory super-cycle: HBM and DRAM/NAND “memflation” as AI demand crowds out commodity supply. MU (HBM/DRAM) and SNDK (NAND) are the cleanest public expressions, MU a Buy and SNDK a Hold, owned through the cycle peak: deeply cyclical names where HBM’s AI pull tightens the whole complex, and where the same operating leverage that prints record margins on the way up reverses violently if AI capex rolls.
What matters now
Memflation is the 2026 story. Gartner sees DRAM/NAND prices up ~125%/234% with relief not before late-2027; MU is a Buy ($1,400) printing ~85% gross margin on ~$100B of contracted RPO, SNDK a Hold ($1,350) on a structurally tighter NAND cycle after the WDC spin, but near a cyclical peak after +2,300% off the spin.
HBM crowds out commodity supply. AI is now ~20% of DRAM output and the HBM4 transition is the next node-race: own the HBM/RPO-levered makers, not the commodity bits, because the AI mix is what keeps the complex tight even as the headline cycle ages.
The risk is the cycle itself. This is a name owned through the peak: the bull case (record margins, contracted backlog) and the bear case (a violent reversal if hyperscaler capex flattens) are the same fact viewed up vs down. Position size, not franchise quality, is the live debate.
5 dashboards
MUNASDAQ
Micron Technology
Semiconductors

Record HBM / AI-memory super-cycle: Q3 FY26 printed $41.5B revenue at ~85% gross margin with ~$100B of contracted backlog (RPO) across 16 strategic-customer agreements, Q4 guided to a record ~$50B. Buy $1,400 (~10x FY27E EPS); ~23% off the June high at ~$967, the discount holds. A deeply cyclical name owned through the peak.

generative AIdatacentersHBM
Price
$967
Rating
Buy
12-mo PT
$1,400
PT as of Aug 2026 Excel model
000660KRX
SK hynix Inc.
Semiconductors

Top global HBM maker and #2 DRAM, the primary NVIDIA supplier leading HBM4 into Rubin. Record Q2 2026 (₩79.3T revenue +257% YoY, ₩60.5T operating profit, ~76% margin), yet the stock de-rated ~52% from its ₩2.99M ATH on cycle-peak fears. Buy ₩2.1M (+48%): anchor fair value on the record fundamentals, not the crashed price, a buy-the-dip still well below the ~₩3.2M Strong-Buy Street; the ₩900k bear is the cycle-top caveat.

HBMDRAMNANDNVIDIA supplier
Price
₩1.42M
Rating
Buy
12-mo PT
₩2.1M
PT as of Aug 2026
CMXTSSE STAR (688825)
ChangXin Memory Technologies
Semiconductors

Hold ¥45 (fade the pop): China’s only at-scale DRAM maker and the linchpin of its memory self-sufficiency drive. IPO’d on the STAR Market Jul 27 2026 at ¥8.66 and closed up ~466% at ~¥3.28T (~$484B), the most valuable A-share. A real supercycle (Q1 2026 net profit ~¥33bn, reversing a decade of losses) but priced at a mid-teens price-to-book scarcity premium (peers ~2-3x) on peak-cycle earnings, with HBM at trial stage and an export-control ceiling. We sit well below Nomura’s ¥116 initiation. Framework, not a formal call.

DRAMDDR5Chinamemory sovereignty
Price
¥49
Rating
Hold
12-mo FV
¥45
PT as of Jul 2026
005930KRX
Samsung Electronics Co., Ltd.
Semiconductors / Memory + Foundry

The world #1 memory maker but the HBM laggard, valued sum-of-the-parts. Record Q2 2026 (₩171T revenue +129%, ₩89.5T operating profit +1,814%, memory operating profit topped Nvidia) yet the stock sold ~25% below our SOTP base on cycle-peak fears. Hold ₩283,000 (SOTP base, now ~22% above the de-rated spot); two-sided cycle risk against a Strong-Buy Street (~₩472k).

DRAMHBMNANDfoundry
Price
₩231,000
Rating
Hold
12-mo PT
₩283,000
PT as of Aug 2026
SNDKNASDAQ
Sandisk Corporation
NAND Memory / Memory & Storage Hardware

NAND pure-play riding a record memory supercycle: Q4 FY26 revenue $8.97B (+372% YoY, +51% QoQ), 84.6% gross margin, EPS $39.25, $42B backlog. Hold $1,350 (trimmed from $1,750 for coherence after a ~48% pullback from the high); the Q1 FY27 guide is up sequentially (rev ~$10.5B) but below lofty consensus, and the name is near a cyclical peak after a ~2,300% run off the WDC spin.

NANDflash memoryWDC spinKioxia JV
Price
$1,212
Rating
Hold
12-mo PT
$1,350
PT as of Aug 2026 Excel model

On the radar: WDC, PSTG, STX.

See these names marked to market
Sector IV

Agentic Software

Global software spend 2026E
~$1.44T
+15.1% (Gartner)
AI-agent software 2026E
~$206.5B
→ $376B 2027E
Oracle RPO
~$638B
+ $300B OpenAI deal
SaaS to usage pricing
≥40%
by 2030 (Gartner)

Upside to 12-mo price target (%)

The read

The application layer above the silicon, re-underwritten for the agentic era: databases, workflow, virtualization and enterprise-AI platforms that turn the ~$725B hyperscaler build into recurring software dollars. The connective question across ORCL, NOW, NTNX, PLTR and AI is no longer “is SaaS dead?” but who converts AI from a cost line into a consumption annuity: systems-of-record with data gravity re-rate as agent platforms, while thin or single-customer models get squeezed by the same hyperscalers they ride on.
What matters now
The pricing model is the thesis. The bull case is consumption capture, not seat count. NOW books ~50% of net-new ACV from non-seat models and raised its Now Assist target to $1.5B, and Gartner sees ≥40% of enterprise SaaS spend shifting to usage/agent/outcome pricing by 2030. The bear case is the same transition: if agents deflate seats faster than vendors monetize consumption, net retention compresses before it re-accelerates.
Backlog visibility vs balance-sheet skepticism is the ORCL debate. A record ~$638B RPO and the ~$300B OpenAI/Stargate deal are the re-rate fuel (Buy $190); ~$70B FY27E capex, ~$122B debt and OpenAI customer-concentration are why the market discounts the headline. Software has become a leveraged, capex-intensive infra bet here: priced on contract quality, not deal size.
Bifurcation is now a rating spread, not a slogan. Data/system-of-record franchises win (PLTR’s ~155% Rule of 40 and +149% US-commercial growth, but a Hold at ~36x sales on valuation, not quality); the VMware-displacement re-rate (NTNX) has largely played out to Hold; and the squeezed middle is explicit: C3.ai is a Hold, downgraded from Sell after a 36% FY26 revenue collapse, on persistent losses and a hyperscaler-native squeeze, with a ~$575M cash cushion the only floor. 2026 is the ROI ‘show-me’ year.
5 dashboards
NTNXNASDAQ
Nutanix, Inc.
Infrastructure Software

Cleanest public VMware-displacement bet. EV/Sales scenarios. Hold $58 after a sharp de-rating. Q3 FY26 ARR $2.43B +15% and ~$770M FCF, but revenue growth has decelerated toward 10%.

VMware displacementHCIAHVenterprise software
Price
$54
Rating
Hold
12-mo PT
$58
PT as of Jul 2026 Excel model
AINYSE
C3.ai, Inc.
Enterprise AI Software / Infrastructure Software

Enterprise-AI pure-play, repriced. EV/Sales scenarios. Hold $9 (downgraded from Sell) after FY26 revenue fell 36%; a ~$575M cash cushion and a Siebel-led restructuring set a floor under a still-shrinking business.

enterprise AITom SiebelShell partnershipgenerative AI platform
Price
$9
Rating
Hold
12-mo PT
$9
PT as of Jul 2026 Excel model
PLTRNASDAQ
Palantir Technologies
Enterprise AI Software / Data Analytics

The best business in software, now fully priced after a blowout Q2. SaaS DCF + dominant reverse-DCF. Hold $175. A ~155% Rule of 40 and 63% FCF margin, but ~40x forward sales leaves no margin of safety.

AIPontologyenterprise AIgovernment / defense
Price
$172
Rating
Hold
12-mo PT
$175
PT as of Aug 2026 Excel model
NOWNYSE
ServiceNow, Inc.
Infrastructure Software / Workflow & Agentic AI

A quality SaaS compounder, de-rated then recovering. Buy $135. 24%+ subscription growth, cRPO $13.2B (+21%), 98% renewals and Now Assist AI ACV past $1B, at ~30x forward earnings. The reverse DCF now implies ~3% terminal growth; the risks are the seat-to-consumption transition and US-federal demand.

Now PlatformNow Assistagentic AIworkflow
Price
$125
Rating
Buy
12-mo PT
$135
PT as of Aug 2026 Excel model
ORCLNYSE
Oracle Corporation
Infrastructure Software / Database & Enterprise Suite

Hyperscale AI-infra pivot meets capex + leverage skepticism. EV/Sales scenarios. Buy $190. $638B RPO and $300B OpenAI deal vs $70B FY27E capex, $122B long-term debt, and customer concentration.

OCIStargateOpenAIRPO
Price
$143
Rating
Buy
12-mo PT
$190
PT as of Aug 2026 Excel model

On the radar: DDOG.

See these names marked to market
Sector V

Frontier · Quantum & Physical AI

Global quantum revenue
~$3B by 2028
~30% CAGR (QED-C)
Waymo paid robotaxi rides
~500k/wk
the scale incumbent
Combined quantum revenue
<$200M
vs tens of $B market cap
TSLA forward P/E
~188x
~75% of cap = optionality

Base fair value vs market price (US$)

The read

The frontier bets outside the AI build-out: quantum-computing pure-plays (QBTS, IONQ, RGTI) priced as call options on a quantum-advantage future, and physical AI / robotics (TSLA, MBLY) where autonomy is finally metering revenue. Both are option-value sectors: you are paying for conversion velocity and survivorship, not today’s cash flow, against well-capitalized incumbents (IBM/Google in quantum; Waymo in autonomy).
What matters now
Quantum is a fund-the-roadmap race, not a revenue race. Bookings are inflecting hard (IonQ RPO +298% YoY, D-Wave bookings +1,994%) yet combined revenue is <$200M against tens of billions of market cap. Our base fair values sit below market (IONQ $40 vs $43, QBTS $15 vs $23, RGTI $11 vs $14). Capitalization, not physics, decides survivors, and Quantinuum’s ~$1.68B June IPO just raised the bar.
Physical AI is metering, but you pay for the option, not the cash flow. TSLA trades ~188x forward with ~75% of its ~$1.5T cap in robotaxi/Optimus optionality (Hold $420), and Waymo, at ~500k paid rides/week, is the scale incumbent two orders of magnitude ahead of Tesla’s unsupervised fleet, reframing TSLA as a cost-curve bet, not a first-mover claim.
Mobileye is the cautionary structural case. A genuine ADAS inflection (EyeQ units +22%, three guide raises) trapped below its peer multiple by the Intel ~77% economic / ~96.9% voting overhang and a four-front squeeze (NVIDIA Drive Thor, Qualcomm, Horizon/BYD, Tesla AI5). Hold ($10). The autonomy optionality is real but un-ownable at scale.
6 dashboards

Quantum Computing

QBTSNYSE
D-Wave Quantum Inc.
Quantum Computing

The only dual-platform quantum pure-play: commercial annealing + in-house gate-model (Quantum Circuits, acquired Jan 2026). One EV/Sales model, three scenarios: bull $40 / base $15 / bear $7.

annealinggate-modeldeep tech
Price
$21
View
Balanced
Base FV
$15
PT as of Aug 2026 Excel model
IONQNYSE
IonQ, Inc.
Quantum Computing

Pre-profit quantum pure-play. One EV/Sales model, three scenarios: bull $70 / base $40 / bear $22.

trapped iondeep tech
Price
$43
View
Balanced
Base FV
$40
PT as of Aug 2026 Excel model
RGTINASDAQ
Rigetti Computing, Inc.
Quantum Computing

Public-market superconducting quantum pure-play; the IONQ counterweight: much smaller revenue base, higher dilution risk, IBM/Google as the existential threat

superconductingdeep tech
Price
$17
View
Balanced
Base FV
$13
PT as of Aug 2026 Excel model

Robotics & Physical AI

MBLYNASDAQ
Mobileye Global Inc.
ADAS / Robotics & Autonomy

Operational stall meets the Intel cap, repriced. Hold $10 (~15% to spot). Q2 2026 revenue was roughly flat with an adjusted-EPS beat flattered by a one-time R&D tax credit, and CEO Amnon Shashua has announced his departure; a $3.8B non-cash goodwill impairment and Intel ~96.9% voting control cap the re-rate at ~3x EV/Sales.

ADASEyeQSuperVisionChauffeur
Price
$9
Rating
Hold
12-mo PT
$10
PT as of Aug 2026 Excel model
TSLANASDAQ
Tesla, Inc.
Robotics & Autonomy / Physical AI

Auto cash-cow base carrying enormous, now-converting but still richly priced robotics/AI optionality. SOTP by business. Hold $420. At ~$1.5T roughly three-quarters of the cap is robotaxi + Optimus optionality (unsupervised robotaxi now live in three metros), and at ~188x the stock de-rated ~11% in 2026 to roughly fair value even as Q1 re-accelerated; a very wide bull ($650) / bear ($210) gap.

robotaxiFSDOptimusEnergy storage
Price
$320
Rating
Hold
12-mo PT
$370
PT as of Aug 2026 Excel model

Space & Launch

SPCXNASDAQ
Space Exploration Technologies Corp.
Space Launch & Satellite Broadband

The first mega-cap space name (Jun 2026 IPO, largest ever): a Starlink profit engine plus a ~50%-by-count / >80%-by-mass launch near-monopoly, bolted to a cash-torching xAI unit at ~112x sales. After the IPO the stock round-tripped to ~$133, just below its $135 offer and roughly fair value; own the franchise, but the asymmetric bear keeps it a Hold $135.

spacesatellitedeep tech
Price
$133
View
Hold
Base FV
$135
PT as of Aug 2026

On the radar: QUBT, SYM.

Quantum Computing
Robotics & Physical AI
Space & Launch
See these names marked to market
Synthesis

Five Expressions, One Trade

I

Follow the capital

The single fact that organizes 2026 tech investing is the hyperscalers’ ~$725B build-out (+77% YoY). The five sectors are stations along the chain that capital travels: silicon (TSM, ASML, NVDA: the toll-roads we’d own), the neoclouds and the power feeding them, the software layer that turns it into recurring dollars, the two security platforms that survived consolidation, and memory, where the most violent leverage sits as AI crowds out commodity supply.

II

The barbell

The bottleneck keeps migrating (chips, then memory, then electricity), and the live question under every name is who actually earns a return on that $725B. The ratings answer it as a barbell: Buys on the chokepoints, where the toll is collected no matter who wins; Holds on the re-rated challengers, where the multiple already prices the win; the library’s lone Sell, where the story outruns the cash; and the frontier quantum and robotics names as call options on survivorship, outside the chain entirely.

III

Route through, not past

Across all five sectors the test is the same: does the build have to route through this name, or merely past it? TSM’s leading edge, ASML’s EUV, the HBM duopoly, the two scaled security franchises, the systems-of-record with data gravity all sit in the path. The substitutable middles (commodity DRAM, point-product cyber, thin seat-SaaS, C3.ai’s squeezed shell) sit beside it, and are where the cycle turns first.

IV

One trade, five masks

Spread across five sectors, this still prices a single bet: that end-demand earns a return on $725B before the depreciation clock and the circular NVIDIA-to-OpenAI loops come due. An ROI air-pocket doesn’t rotate between the five; it hits all of them at once. So sizing is the whole game: own the cyclical compute and memory beta lighter and through the peak, knowing it breaks hardest, and lean on cyber and software’s recurring dollars to cushion the fall.

See all coverage marked to market
The method · behind every dashboard

How It Gets Built

Every dashboard in this library runs the same pipeline, from what earns coverage to what gets disclosed. Six steps, in order.

Scope

Coverage Philosophy

Eight categories that mirror the most contested corners of tech: AI compute and memory, hyperscalers, quantum, security, robotics and physical AI, networking and server hardware.

Valuation

Valuation Framework

Each company gets the framework its stage and earnings quality warrant. Every name carries Bull, Base, and Bear targets with probability weights, blended into one fair value.

Inputs

Data Sources

A live quote feed sits over curated per-dashboard figures, each stamped with an asOf date and reviewed quarterly after earnings.

Method

AI Workflow

AI is a force-multiplier on the inputs, not the verdict. The rating, price target, and scenario probabilities stay author judgment.

Freshness

Update Cadence

A colored Data as of badge rides every dashboard and hub card. Live values refresh on each page load.

Standing

Disclaimers

An independent research sample, built to be inspected. Not investment advice.

Read the full methodology Behind the Library, all six parts in depth.

About this work

The dashboards and downloadable Excel models in this library are an independent research sample for demonstration. They are not investment advice, and not a recommendation to buy or sell any security. The models are illustrative teaching workbooks: their assumptions are explicit and meant to be changed. Figures were compiled from public filings and data providers and will drift. Refresh before relying on anything here. Spreadsheets download rather than preview; open them in Excel, Google Sheets, or Numbers.