AI Market Cascade

1. Executive Summary

The single clearest finding across all 31 industries and 205 cross-industry linkage vectors is that AI value creation has decoupled from software and anchored directly into physical capital, clinical biology, and heavy infrastructure. The average 5-year Relative-Value Score (RVS) across our model is 1.91x. Twenty-two industries land in net-positive territory, but the composition of the leaders defies the initial narrative of the generative AI boom. Pure-play software is no longer where the structural economic rent accumulates; that rent has migrated up the compute stack to physical fabricators and out into industries that deploy compute against tangible physical constraints.

The clear structural winners are clustered in three core verticals: life sciences, silicon architecture, and heavy utility/capital infrastructure. Biotech & AI Drug Discovery (#1, 2.49x), Big Pharma (#4, 2.41x), and Chip Design & Fabless (#2, 2.38x) lead the model. The underlying mechanism is simple: these sectors deploy AI models against massive, high-dimensional chemical, physical, and mathematical bottlenecks where successful inference creates immediate intellectual property or hardware monopolies. In contrast, Capital Goods (#3, 2.35x) acts as the essential picks-and-shovels provider to the entire compute ecosystem, monetizing multi-gigawatt backlogs in gas turbines, transformers, and autonomous operational machinery.

At the bottom of the stack, the picture is severe value erosion driven by margin compression and business-model obsolescence. Commercial & Professional Services (#31, 0.62x) and Enterprise SaaS (Legacy) (#30, 0.67x) represent the sharpest risk profiles. The fundamental threat is not that these businesses cannot use AI, but that AI directly undermines their monetization architecture. Billable-hour business models cannot easily replace billable human time with software speed without top-line contraction, while per-seat SaaS providers face an existential reckoning: if autonomous AI agents compress enterprise department headcounts by 30% to 40%, the seat-based recurring revenue engine implodes.

2. The Bull Case: Where AI Creates the Most Value

The high RVS scores in our top cohort are driven by specific model components: large, positive cross-industry ripple effects (crossEffects = +0.9000), aggressive direct top-line contribution (revenue), and capital efficiency gains (capEff). The value is accumulating where high barriers to entry protect the productivity dividend.

Accelerating Hardware and the Industrial Supply Chain

Chip Design & Fabless (#2, 2.38x) and Foundry & Equipment (#6, 2.28x) are operating under multi-year forward revenue visibility. Broadcom's trajectory toward approximately $115 billion in AI revenue for fiscal 2027 and $230 billion in fiscal 2028, paired with NVIDIA guiding quarterly revenue above $108 billion, confirms this is an ongoing architectural buildout rather than a short-term spike. On the manufacturing side, SEMI projects wafer fab equipment sales climbing from $165.9 billion in 2026 to $229.5 billion by 2028. TSMC lifting capital expenditure to $60-64 billion confirms the durable requirement for physical silicon fabrication capacity.

The Biopharma Operational Shift

In Biotech & AI Drug Discovery (#1, 2.49x) and Big Pharma (#4, 2.41x), the economic driver is the mitigation of the $300 billion patent cliff due by 2030. Clinical translation is moving from theory to late-stage execution: Insilico Medicine's Phase III trial of rentosertib for idiopathic pulmonary fibrosis (320 patients across 47 sites) represents the first true late-stage clinical test for an AI-generated molecule. In parallel, operational gains are immediate: Novo Nordisk reports cutting select clinical documentation pipelines from 40-50 personnel over 15 weeks down to a 3-person review team operating in minutes via Claude on AWS. These internal efficiency gains translate directly into protected operating margins.

The Non-Consensus Outperformer: Capital Goods

The primary surprise in our top tier is Capital Goods (#3, 2.35x). Traditional software investors routinely underappreciate how quickly compute bottlenecks translate into electrical and mechanical bottlenecks. GE Vernova's gas turbine backlog and slot reservations expanding to 116 GW (with orders up 88% organically in Q2 2026 to $24.2 billion) shows that capital goods providers are the primary toll collectors on data center energization. Caterpillar's Power Generation division expanding 29% to $3.098 billion in a single quarter reflects the identical dynamic: when compute farms outrun grid interconnection queues, heavy industrial capital equipment absorbs the spend.

3. The Bear Case: Where AI Destroys or Redistributes Value

Value destruction in our model is rarely about operational incompetence; it is about economic displacement and margin transfer. The bottom cohort suffers from severe negative cross-industry drag (crossEffects down to -0.6750) and competitive margin compression (compDyn).

Structural Disruption vs. Internal Redistribution

A vital analytical distinction is the difference between structural industry contraction and intra-industry redistribution:

• Structural Contraction — Commercial & Professional Services (#31, 0.62x): In IT consulting and corporate staffing, the model relies on billing out human labor hours at a premium. When an enterprise client uses autonomous agents to write code or process audit trails, they do not expand their IT consulting budget. Accenture's new bookings fell 2% in its fiscal third quarter, alongside management disclosures that client budgets are not expanding; AI spend is merely cannibalizing legacy implementation dollars. When output volume expands 10x but required human input hours fall 50%, billable hours collapse faster than realized pricing can rise.
• Intra-Industry Value Redistribution — Enterprise SaaS: Notice the staggering split between AI-Native SaaS (#23, 1.38x 5yr / velocity 1.20) and Legacy Enterprise SaaS (#30, 0.67x 5yr / velocity 0.50). Legacy incumbents boast nominal AI progress—Salesforce crossing $1.5 billion in Agentforce ARR and ServiceNow hitting $1.0 billion in AI annual contract value. However, these gains remain a tiny fraction of their $40B+ total revenue bases. Meanwhile, AI-native entrants like Anthropic scaled run-rate revenue from $9 billion at late-2025 to over $47 billion by May 2026, and legal-native Harvey passed $400 million ARR across 80% of top law firms. The software pool is not shrinking, but economic surplus is transferring away from seat-based workflow wrappers to underlying models and specialized vertical agent architectures.
• Zero-Marginal-Cost Content Devaluation — Media & Entertainment (#29, 0.76x): While production pipelines gain real efficiencies—Netflix deploying GenAI across 300 titles and Disney accelerating VFX workflows—the distribution economics are failing. As Google search traffic to digital publishers dropped by one-third in the year leading into late 2025 due to direct AI answer synthesis, organic traffic funnels evaporated. Concurrently, legacy advertising agencies face severe friction: WPP's like-for-like net revenue fell 4.7% in the first half of 2026. Lowering production costs cannot save a business model when synthetic supply approaches infinity and gatekeeper distribution channels close off inbound traffic.

4. Cross-Industry Cascade Effects

Our model tracks 205 specific cross-industry dependencies. AI adoption does not happen in isolation; one sector's capital investment or efficiency gain frequently functions as another sector's operational bottleneck or revenue destruction event.

Critical Macro Transmission Channels

1. The Compute-to-Power-to-Industrial Transmission Loop:

   Hyperscaler capex plans are forcing a multi-tier capital rotation. Cloud & AI Platforms (#11) spend at annual rates approaching $200B+ per vendor (e.g., Alphabet guiding 2026 capex to $195-$205 billion). This capital flows instantly into Chip Design (#2) and Foundry (#6). But fab capacity cannot be turned on without physical energy infrastructure. This triggers a direct demand cascade into Utilities (#20) (AEP contracting 69 GW of additions through 2030) and Capital Goods (#3) (GE Vernova gas turbine backlog hitting 116 GW). The compute bottleneck is no longer code; it is physical megawatts and copper.

2. The Enterprise Labor Reduction Feedback Trap:

   In Banks (#9), Jamie Dimon notes AI has already eliminated 30% to 40% of jobs in select operational units. In Insurance (#8), Travelers reduced its claim call center headcount by one-third, consolidating four centers into two. While this preserves underwriting margins for financial institutions, it fires a direct negative feedback loop back into Enterprise SaaS (Legacy) (#30) and Commercial Services (#31). For every 10,000 back-office seats eliminated at a major bank or insurer, enterprise software providers lose thousands of recurring seat licenses, and IT staffing firms see contract extension requests cut permanently.

3. The Information Intermediation Bypass:

   The rise of high-capability reasoning agents directly bypasses standard web distribution funnels. Interactive Media & Services (#19) retains monetizable power because Meta (ad impressions up 14%, pricing up 12%) and Google (Search revenue up 17% to $63.3 billion) control the consumer endpoints. However, the publishers that feed the open web are cut out entirely. This directly reduces traffic to digital properties in Media & Entertainment (#29), causing brand advertisers to pull spending from open programmatic web networks and concentrate it into closed walled gardens.

4. Biotech Acceleration as a Tool Demand Driver:

   Increased algorithm-driven molecular generation does not eliminate physical wet labs; it radically intensifies the demand for high-throughput empirical validation. Biotech & AI Drug Discovery (#1) designs candidates in seconds, which sends a wave of validation assays into Life Sciences Tools (#5). This explains why IQVIA posted $3.15 billion in R&D Solutions bookings in Q2 2026 (a 1.22x book-to-bill). Faster virtual design directly requires higher physical testing capacity.

5. The 5-Year vs. 10-Year Divergence

A static 1-year or 5-year view fails to capture how capital efficiency profiles reverse over a decade. In our model, several industries exhibit pronounced divergence between their medium-term (5-year) and structural long-term (10-year) Relative-Value Scores.

Key Divergence Dynamics

• The Automotive Long-Term J-Curve: Automobiles & Components (#16) sits at a suppressed 1-year score of 1.01x, reflecting painful near-term operating economics: Tesla's operating margin compressed to 1.4% in Q2 2026 as it funded massive AI training clusters, while deployment scale remains limited. However, its 10-year score jumps to 2.82x. Waymo reaching 500,000 paid rides per week across 14 cities with a 94% reduction in serious injury crashes proves that once autonomous networks reach commercial density, capital returns decouple entirely from legacy vehicle unit manufacturing.
• The Cloud Platform Free Cash Flow Squeeze: Cloud & AI Platforms (#11) steps up strongly from 1.18x (1yr) to 2.25x (5yr), but its expansion rate flattens by year 10 (2.73x). The immediate risk is cash burn outrunning operating cash flow: Epoch AI projections show aggregate hyperscaler free cash flow crossing zero as capex grows at ~70% against 23% operating cash flow growth. Oracle booking negative $5 billion in free cash flow and issuing $20 billion in equity to finance capacity illustrates that hyper-scale cloud platforms face intense balance-sheet friction before their 10-year cash returns normalize.
• AI-Native SaaS Plateaus After Early Land-Grab: Enterprise SaaS (AI-Native) (#23) shows an immediate jump (1.20x 1yr), but its 10-year trajectory reaches only 1.50x. Once early vertical pioneers (Anthropic, Harvey, Sierra) conquer primary white space, the commoditization of foundational model intelligence exerts downward pricing pressure on application layers that do not own specialized, closed-loop proprietary data.

6. Portfolio Implications: Diagnostic Questions

Rather than relying on generic market commentary, investors managing substantial capital should stress-test their holdings against the specific structural pressures identified in our 31-industry matrix.

Questions for Holders of Bottom-Tier Industries

If your portfolio holds material allocations in Enterprise SaaS (Legacy), Commercial Services, or Media & Entertainment, evaluate whether each company has clear answers to these operational tests:

• The Seat-Count Compression Test: If a corporate customer reduces its back-office, legal, or administrative headcount by 25% over the next 36 months using autonomous agents, will your SaaS holding's net revenue retention remain positive? Does the company have a viable, deployed usage- or outcome-based pricing model that offsets falling user seats?
• The Billable Hour Replacement Ratio: For holdings in IT consulting, staffing, or outsourced legal/advisory services, what percentage of revenue is still tied directly to billed human hours? Can the firm bill clients for work completed via automated agents at margins high enough to offset the loss of billed professional hours?
• The Proprietary Data Moat Test: Does your media or information services holding own authoritative, proprietary data that models must license to function (e.g., LexisNexis growing underlying revenue 10% with 90% of new business tied to AI tools), or does it produce generic editorial content exposed to zero-click AI search synthesis?

Concentration Risk: The Hyperscaler Capex Dependency Cluster

Investors heavily allocated to mega-cap technology must understand the single largest balance-sheet risk in the market today: circular vendor financing and concentrated capital spend. Chip Design (#2), Foundry & Equipment (#6), and Tech Hardware (#12) are all heavily dependent on the capital expenditures of just four to five cloud hyperscalers. When Broadcom's CFO openly notes the firm is helping bridge financing gaps for strategic AI lab customers, and hyperscaler debt issuance climbs from 9% of capex to 32% within two years, any pacing adjustment in hyperscaler balance sheets will transmit an immediate margin shock down through the entire semiconductor and hardware food chain.

The Stealth Winners: Real Assets and Physical Enablers

Market attention remains fixated on consumer software applications and foundational model labs. Meanwhile, the durable pricing power in this dataset sits in unheralded physical enablers:

• Energy Equipment Suppliers & Industrials: Capital Goods (#3, 2.35x) and Energy (#13, 2.23x) are locking in multi-decade utility and power contracts. SLB guiding its Data Center Solutions business over a $1 billion run-rate and Chevron signing 20-year, 2.67 GW direct-supply agreements with Microsoft at targeted mid-teen returns prove that conventional energy producers are monetizing AI infrastructure buildouts without taking software obsolescence risk.
• Regulated Grid Operators: Utilities (#20, 2.07x) such as AEP (69 GW pipeline) and Southern Company (data center power demand up 55% YoY) are executing historic capital expansion programs that are formally added to rate bases, guaranteeing long-term regulated returns driven directly by enterprise AI compute deployment.