Ai Value Chain Sector IC Memo

Executive Summary & Action Plan

Verdict

BUY (selective — FINAL, all 18 workups complete) — The sweep's headline finding inverts the tape's instinct: the layers the market loves (hyperscale compute, data platforms, security) all price at or far above our fair values, while the layer the market left for dead — front-office applications — holds the only two names with real margin of safety. Top Pick: Salesforce (CRM), BUY, FV $220 (+34.7%), with Adobe (ADBE, BUY, FV $302, +35.0%) as the co-pick. Final tally: 2 BUY · 7 HOLD · 9 SELL.

Loading chart

Price $163.32 | Market cap $133.8B | Target $220 | Upside +34.7% IWANNAVY Fair Value (full pipeline, 2026-07-12) · price as of 2026-07-11 close · Street consensus $245.60 (reference) — the reverse-DCF at $163 implies ~1.8% perpetual growth (below inflation) for a franchise guiding +11% with Agentforce compounding at +205%

Loading chart

Abstract

We mapped the AI industry's software value chain across four investable layers — cloud/hyperscale compute, data platforms, security, and front-office applications — and completed full five-specialist workups on all 18 public names in a single day, every fair value built from SBC-expensed DCFs, forward cash-flow multiples, and price-verified inputs. The finding inverts the tape: hyperscaler AI revenue is re-accelerating (Azure +40%, Google Cloud +63%) behind a $635–665B capex cycle, yet every compute name prices at or above our fair value (MSFT +3.6%, AMZN +0.3%, META −11.5%, GOOGL −18.2%, ORCL −19.7%, CRWV −19.0%, SPCX −14.7%); the data cohort's owner earnings vanish into stock compensation (SNOW −23.1%, DDOG −22.0%, PLTR −17.2%; only MDB, with the best SBC hygiene, holds at −10.8%); and security trades past its own Street consensus (CRWD −45.5%, PANW −42.0%) except post-reset ZS (+7.0%). The only margin of safety in the map sits in the SaaSpocalypse wreckage: CRM (BUY, +34.7%) and ADBE (BUY, +35.0%), both GAAP-profitable cash machines with proven AI attach, priced for no growth. Final stances: Cloud Neutral (was OW — quality yes, price no), Data UW, Security UW (ZS excepted), Applications OW (was UW). Principal risks: frontier-lab release shocks, lockup supply (SPCX Aug–Dec, FIG Aug), backlog-credit repricing, and the seat-compression tripwires flagged in each BUY report.

Action Plan

Own the de-rated app-layer cash machines: CRM at market and $150, ADBE at market and $200. Hold-accumulate the reset names on weakness (ZS $125/$115, NOW $95/$85 post-print, FIG only after the August lockup at $18/$16, MDB $305/$270). Reduce everything trading above its bull case (GOOGL, SNOW, PLTR, DDOG, CRWD, PANW) and the leveraged backlog stories (ORCL, CRWV, SPCX) into strength.

Loading chart

Risk/Reward (CRM) ~4.3:1 | Prob-weighted 12M return (CRM) +37.8% | Confidence: Medium-High. ADBE co-pick: R/R ~4.2:1, PW +25.0%, governance-catalyst-gated. Five names sit above even our bull-case values (GOOGL, SNOW, PLTR, DDOG, CRWD ranked by severity) — the sweep's recurring pattern.

Value Chain Map

Four layers, ranked by where the marginal AI dollar lands today. Bottleneck = what the layer controls that others must pay for.

Layer (stance)	Bottleneck controlled	Tier 1	Tier 2 public	Private disruptors
Cloud & Hyperscale (Neutral)	Compute capacity, power, custom silicon	Amazon (AMZN), Microsoft (MSFT), Alphabet (GOOGL), Meta (META), Oracle (ORCL), CoreWeave (CRWV), SpaceX (SPCX)	Nebius (NBIS), IREN (IREN), Applied Digital (APLD), Digital Realty (DLR), AST SpaceMobile (ASTS)	OpenAI, Anthropic, xAI, Crusoe, Lambda
Data Platform (UW)	Governed context agents must query	Snowflake (SNOW), MongoDB (MDB), Palantir (PLTR), Datadog (DDOG)	Confluent (CFLT), Elastic (ESTC), Teradata (TDC)	Databricks ($6.9B RR +80%), Supabase, ClickHouse, Pinecone
Security (UW; ZS excepted)	Identity & inspection chokepoints for humans and agents	CrowdStrike (CRWD), Palo Alto (PANW), Zscaler (ZS)	Fortinet (FTNT), SentinelOne (S), Cloudflare (NET), Okta (OKTA), Netskope (NTSK)	Wiz (now Google), Cyera, Abnormal, Torq, Island
Front-office & Apps (OW)	System-of-record + workflow distribution	Salesforce (CRM), Figma (FIG), ServiceNow (NOW), Adobe (ADBE)	HubSpot (HUBS), Workday (WDAY), Atlassian (TEAM), Monday (MNDY)	Anthropic (Cowork/Design), OpenAI, Canva, Sierra, Vercel v0

Adjacent layers (semis/NVDA, power, networking) are covered in our companion sector reports — the 2026-06-18 AI value chain (compute) memo and the 2026-07-10 AI/Quantum/Energy/Space fair-value screen.

What Happened / State of the Sector

2026 has been the year the market started pricing AI's second-order effects. January's Claude Cowork launch triggered the "SaaSpocalypse" — a $285B one-day software wipeout and ~$1T over six sessions (IGV −23%, CRM −30%+), followed by repeated 5–30% single-day air pockets on each frontier-lab release (Claude Code Security in February, Claude Design in April, Google AI Threat Defense in May). A late-June rotation back into beaten-down software (NOW/CRM/ADBE +10%+) showed the worst case failing to materialize; agent monetization is now measurable (Agentforce ARR $1.2B +205%, Now Assist tracking to $1.5B ACV, Adobe AI-first ARR >$500M).

Below the apps, the backlog-credit question dominates. Oracle's June 10 print (OCI +93%, RPO $638B) still crushed the stock ~35% in June because FY26 FCF was −$23.7B and FY27 capex guides to $70B — the market now pays ~$0.64 per $1 of ORCL RPO (Motley Fool, 2026-07-06). Meta went the other way: the "Meta Compute" cloud push (CNBC, 2026-07-01) plus its Broadcom-designed Iris accelerator entering production in September (Reuters, 2026-07-09) re-rated META from capex spender to infrastructure owner (+6.0% on 7/10).

GAP verification (2026-07-10 session, five ±5% moves): No macro shock — S&P 500 +0.4%, Nasdaq +0.3% (Yahoo Finance). The common thread was rotation out of high-multiple AI software into AI hardware, catalyzed by SK Hynix's record $26.5B US listing debut (+12.8%) and NVDA +4%: MDB −5.7% and the security group (CRWD −5.7%, ZS −5.3%, PANW −3.7%) fell on no company news. Idiosyncratic layers: DDOG −4.3% extended Bernstein's 7/6 downgrade; FIG −5.2% was a failed rally into its Aug lockup; SPCX −4.5% was lockup front-running (~44% of shares unlockable by early September); META +6.0% was the Iris chip. Verdict: mixed — group rotation amplified by name-specific supply and positioning.

What to Watch Now

July 22–30 hyperscaler gauntlet: GOOGL (7/22), MSFT + META (7/29), AMZN (7/30) — whether $635–665B combined capex guides get raised or hedged is the single biggest signal for the whole map.
August–September software prints: DDOG 8/6 → PLTR 8/10 → FIG 8/13 → CRWV 8/18 → PANW 8/24 (E) → SNOW 8/26 (E) → MDB 8/27 (E) → CRM late-Aug (E) → CRWD 9/2 (E) → ZS 9/8 (E) → ADBE 9/10 → ORCL 9/14 — the test of the "AI-lab spend plateau" thesis.
Lockup supply: SPCX first tranche (~20%) unlocks after Q2 results (8/6 E), rolling to ~34% by 9/10 (~44% only if the $175.50 early-release trigger hits); FIG's ~$6–7B extended lockup frees Aug 17–31. Both are binary float events.
Frontier-lab release calendar: every 2026 lab launch has repriced a segment 5–30% in a day; Black Hat (8/1–6) and fall model releases are the next scheduled shock windows.

What to Watch Next (Technology)

Custom silicon vs merchant GPUs: Trainium, TPU, Meta Iris (TSMC production from September) — the share of accelerator capex bypassing Nvidia sets 2027 fleet economics.
Power as the binding constraint: Meta 7→14GW, CoreWeave 3.5GW contracted → who can actually deploy guided capex; favors energy-first builders.
Agent identity as the new control plane: PANW-CyberArk vs Okta vs CRWD Continuous Identity — agent governance lags adoption ~8:1 (Gartner), the next $10B+ ARR pool.
Seats → consumption/outcome pricing: Agentforce work units, Figma AI credits, Sierra per-resolution pricing — decides whether agents expand or shrink each vendor's revenue pool.
Substrate convergence on Postgres/lakehouse: Databricks Lakebase, Snowflake Postgres, Supabase — commoditizes standalone vector DBs and blurs the SNOW/MDB boundary.

Fair Value Notes

All 18 names carry same-day full-pipeline fair values (five specialists each; log-median of SBC-expensed DCF and forward cash-flow multiples; AMZN inherited from its fresh 7/1 tearsheet). One line each; full audit trails in the individual tearsheets.

Applications (OW): CRM $220 BUY — reverse-DCF at spot implies sub-inflation perpetual growth; structured bear $184 exceeds spot · ADBE $302 BUY — cheapest multiples + best GAAP economics; bear DCF (terminal decline) $167 · NOW $124 HOLD-acc — needs only ~9.5% decade CAGR, but hostile print reaction function (7/22 gate) · FIG $26 HOLD — SBC-accounting debate priced, August lockup owns the tape.

Cloud (Neutral): MSFT $399 HOLD (accumulate $349/$320) · AMZN $246 HOLD (at fair value) · META $592 HOLD (trim the July compute-pivot premium; prior $827 retired) · GOOGL $292 SELL (spot above bull $305) · ORCL $113 SELL (supersedes 7/10's $185; S&P BBB− 7/9, ~47–54% OpenAI RPO; binary 9/14) · CRWV $72 SELL (EV pays $0.82/$1 backlog vs fair $0.72; interest = 26% of revenue) · SPCX $124 SELL (xAI consolidation inside the listco; ~34% lockup by 9/10; supersedes 6/18's $145).

Data (UW): MDB $305 HOLD — best SBC hygiene in cohort vs Postgres funnel decay · SNOW $201 SELL — SBC exceeds plain FCF; spot above bull · PLTR $105 SELL — 45x-forward grants still can't reach spot; FDE model being copied at hyperscaler scale · DDOG $201 SELL — owner earnings $154M vs $88B EV; Street mean below spot.

Security (UW, ZS excepted): ZS $149 HOLD-acc — only name below our warranted multiple; 9/8 formal guide is the binary · PANW $189 SELL — +14% organic behind +31% CyberArk headline at 74x FCF · CRWD $102 SELL — corrected 130.6x EV/FCF; reverse-DCF arithmetically impossible.

Catalysts & Risks

Catalyst	Timing	Impact	·	Risk	Prob	Impact
Hyperscaler Q2 prints: capex + AI run-rates	7/22–7/30	High	·	Capex guide cut ripples down the whole chain	Low-Med	High
Meta Iris production start (Broadcom/TSMC)	Sep 2026	Med	·	Frontier-lab release shock (5–30% air pockets)	High	High
ORCL Q1 FY27: RPO→revenue conversion	9/14	High	·	Backlog-credit repricing (ORCL, CRWV)	Med	High
Agent-ARR beats (Agentforce, Now Assist)	Aug–Sep	Med	·	SPCX ~34% unlock by 9/10 (44% conditional) / FIG $6–7B Aug lockup	High	Med
Databricks round close at $165–175B	Q3 2026	Med	·	Hyperscaler bundling (Google AI Threat Defense) squeezes pure-plays	Med	Med
ZS FY27 formal guide vs 16–17% reset	9/8 (E)	Med	·	AI defense capex structurally raising security cost curves	Med	Med

AI Value Chain — Cloud, Data, Security & Applications (Sector Map) (Ai Value Chain Sector)

Executive Summary & Action Plan

Verdict

Abstract

Action Plan

Value Chain Map

What Happened / State of the Sector

What to Watch Now

What to Watch Next (Technology)

Fair Value Notes

Catalysts & Risks

References