CHAPTER 9: THE 11-STOCK "DUAL-SCREEN" ELITE AI PORTFOLIO AND 5-YEAR CAPEX WAVE ROTATION ENGINE
— Guy M. Girardi, The Investing Machine (Chapter 5: Reading the Blueprint, p. 42)
Global full-stack AI infrastructure capital expenditure is projected to total $8.9 trillion between 2026 and 2030 (encompassing Gartner’s $7.0 trillion core IT/data-center hardware subset plus $1.9 trillion in civil, grid, and nuclear power buildouts, and aligning with Columbia Business School’s 8-year $10.3 trillion 2025–2032 U.S. macroeconomic projection), driving U.S. data-center power demand from ~30 GW to 90–118 GW. To harvest this buildout without overpaying for speculative hype or exposing the portfolio to 100% single-sector chip cyclicality, this chapter enforces two strict quantitative guardrails:
- The Dual-Screen Valuation Calipers (Forward P/E ≤ 29.0x AND PEG < 1.00): Every holding must trade at or below a 29.0x Forward P/E ceiling and strictly below a 1.00 PEG ratio. The resulting 11-stock basket achieves a 19.46x weighted Forward P/E (cheaper than the S&P 500 SPY at 22.5x) and a 0.39 weighted PEG ratio (vs. QQQ at 1.55)—meaning earnings are compounding more than 2.5x faster than valuation multiples.
- 62.0% Non-Silicon Physical Infrastructure Weighting: Because roughly half of every AI CapEx dollar is spent outside semiconductor wafers, 62.0% of the portfolio is allocated to the "Permanent 40%" physical bottlenecks: baseload nuclear electricity (Talen Energy, 12%), direct-to-chip liquid cooling (Vertiv, 11%), industrial HVAC/piping (Comfort Systems, 10%), high-voltage electrical MEP (EMCOR, 10%), site rock blasting/excavation (Sterling Infrastructure, 7%), and 800G/1.6T active electrical rack cabling (Credo Technology, 12%), alongside 38.0% in foundational silicon and HBM4 memory (NVIDIA 11%, TSMC 10%, Broadcom 9%, SK hynix 4%, Micron 4%).
- The 5-Year Capex Wave Rotation Engine & 50% Automatic Trim Rule: Capital rotates sequentially across 20 quarters from Wave 1 (Site Excavation & Initial Silicon/HBM) to Wave 2 (Interior MEP Fit-Out, Liquid Cooling, Custom Inference ASICs & Nuclear Power) and Wave 3 (4-Year GPU Replacement Cycle), while automatically trimming any holding by 50% if its Forward P/E breaches 29.0x or its PEG reaches 1.00.
9.1 Executive Thesis: Why These 11 Monopolies Beat Standard AI & Semiconductor Funds
In Chapter 5 (Reading the Blueprint) and Chapter 10 (Raw Material & Market Mechanics) of Volume I (The Investing Machine), we established that an operator never buys a flashing ticker symbol based on media narrative; you pull out the calipers to verify an impenetrable Economic Moat and demand a strict Margin of Safety on valuation.
When retail investors attempt to invest in the $8.9 trillion global full-stack AI buildout (2026–2030), they typically make two fatal mechanical errors: (1) They buy standard semiconductor ETFs that expose 100% of their capital to single-sector silicon cyclicality and the 3-to-6-year perishable compute decay we proved in Chapter 3; or (2) They chase unprofitable software wrapper startups trading at 50x to 100x forward earnings with PEG ratios above 2.50.
Roughly half of every AI capital expenditure dollar is spent on physical, non-silicon bottlenecks that must be permitted, excavated, wired, piped, cooled, and powered before a single GPU can be turned on: site rock blasting and grading, high-voltage interior electrical switchgear, direct-to-chip liquid cooling, 24/7 baseload nuclear electricity, and active electrical rack cabling. By enforcing two unbreakable quantitative calipers across the entire four-tier supply chain—a Forward P/E ceiling of 29.0x and a PEG ratio (Price/Earnings-to-Growth) strictly below 1.00—this 11-stock basket eliminates overpriced hype and isolates profitable infrastructure monopolies whose earnings are compounding more than 2.5x faster than their valuation multiples.
9.2 Master Portfolio Allocation & Valuation Matrix (October 2026 Baseline)
| Ticker | Company Name | Tier & Bottleneck Role | Current P/E | Forward P/E | PEG Ratio | Target Wt | Per $100,000 | Per $500,000 |
|---|---|---|---|---|---|---|---|---|
| NVDA | NVIDIA Corp. | Tier 1: GPU Compute | 26.83x | 24.88x | 0.22 | 11.0% | $11,000 | $55,000 |
| TSM | Taiwan Semiconductor | Tier 1: Wafer Foundry | 23.52x | 21.41x | 0.41 | 10.0% | $10,000 | $50,000 |
| AVGO | Broadcom Inc. | Tier 1: Custom ASICs | 41.28x | 19.42x | 0.45 | 9.0% | $9,000 | $45,000 |
| SKHY | SK hynix Inc. (ADR) | Tier 1: HBM4 Memory | 7.85x | 3.80x | 0.66 | 4.0% | $4,000 | $20,000 |
| MU | Micron Technology | Tier 1: US DRAM/HBM | 11.54x | 4.93x | 0.01 | 4.0% | $4,000 | $20,000 |
| TLN | Talen Energy Corp. | Tier 2: Nuclear Power | 379.4x* | 10.28x | 0.35 | 12.0% | $12,000 | $60,000 |
| VRT | Vertiv Holdings Co. | Tier 3: Liquid Cooling | 44.13x | 21.38x | 0.41 | 11.0% | $11,000 | $55,000 |
| FIX | Comfort Systems USA | Tier 3: HVAC / Piping | 34.46x | 24.32x | 0.33 | 10.0% | $10,000 | $50,000 |
| EME | EMCOR Group Inc. | Tier 3: Electrical MEP | 20.42x | 17.63x | 0.67 | 10.0% | $10,000 | $50,000 |
| STRL | Sterling Infrastructure | Tier 3: Site Excavation | 29.75x | 17.65x | 0.71 | 7.0% | $7,000 | $35,000 |
| CRDO | Credo Technology | Tier 4: AEC Cabling | 57.93x | 28.93x | 0.18 | 12.0% | $12,000 | $60,000 |
| WEIGHTED PORTFOLIO AVERAGES / TOTALS | 30.56x* | 19.46x | 0.39 | 100.0% | $100,000 | $500,000 | ||
9.3 Company-by-Company Investment Theses & Specific Rotation Rules (All 11 Monopolies Unpacked)
Tier 1: Foundational Compute Silicon, Wafer Foundry & HBM4 Memory (38.0% Total Weight)
Trading 49% below its 10-year median P/E (53.02x), NVIDIA commands full-stack AI rack pricing power as its 20-petaflop Vera Rubin NVL72 racks and containerized DSX Pods ramp into full production. Protected by its 5-layer CUDA/NVLink moat and the U.S. National Security Floor ("The Pentagon Put" detailed in Chapter 8), a mandatory 3-to-4-year GPU thermal replacement cycle creates a permanent upgrade annuity starting in 2029–2030.
The undisputed N3/N2/A16 advanced wafer foundry and CoWoS-L packaging monopoly (58% revenue from HPC; >60% gross margin). TSMC functions as the ultimate semiconductor toll bridge: it wins regardless of whether merchant GPUs (NVDA), challenger GPUs (AMD), or custom hyperscaler ASICs (AVGO) capture marginal compute share, and it also fabricates the custom logic base dies for 12-high and 16-high HBM4 memory stacks.
Dominates custom hyperscaler AI accelerators (XPUs for Google TPU and Meta MTIA, with quarterly semiconductor revenue surging +127% YoY to $20.8B) alongside high-speed Tomahawk 5/6 Ethernet switch silicon and VMware enterprise private-cloud software. Positioned as the primary semiconductor beneficiary of the 2027–2029 structural shift from frontier pre-training toward cost-efficient, high-volume AI inference.
Trades at the single lowest forward multiple in global AI hardware (3.80x Forward P/E) following South Korea’s 40-day KOSPI liquidation foreshock (Chapter 6). Protected by its proprietary MR-MUF (Mass Reflow Molded Underfill) liquid-epoxy thermal packaging process and its direct TSMC logic base-die partnership for 12-high and 16-high HBM4 memory stacks used on NVIDIA’s Rubin platform.
The sole U.S.-domiciled manufacturer of HBM3E/HBM4 and high-density low-power enterprise LPDDR5X DRAM (FY2026 revenue tripled to $133.2B at an 81% gross margin). Triple-digit earnings expansion compresses Micron’s forward multiple under 5.0x (0.01 PEG ratio) while CHIPS Act grants subsidize its domestic megafabs in Idaho and New York, providing a vital geographic hedge against East Asian seismic or geopolitical risk.
Tier 2: Baseload Nuclear Power Anchor (12.0% Total Weight)
Owner of the 2.5-gigawatt Susquehanna nuclear generating facility directly co-located with Amazon Web Services (AWS) data-center campuses inside the capacity-constrained PJM Interconnection grid. Because permitting new high-voltage transmission lines takes 5 to 7 years, existing nuclear baseload plants command extraordinary pricing power. Talen captures multi-decade nuclear Power Purchase Agreement (PPA) repricings and rising PJM capacity auction payments at a fraction of peer utility valuation multiples (10.28x Forward P/E).
Tier 3: Liquid Cooling, Interior Mechanical/Electrical MEP & Site Excavation (38.0% Total Weight)
The pure-play global leader in data-center thermal and critical power management (>75% revenue from data centers; order backlog up +109% YoY to $15B). Vertiv co-designs liquid-cooling reference architectures directly with NVIDIA, supplying the Coolant Distribution Units (CDUs), cold-plate manifolds, rear-door heat exchangers, and chillers mandatory for 120 kW to 300+ kW Blackwell and Vera Rubin racks.
The premier national specialty contractor for complex industrial chillers, heavy mechanical piping loops, and off-site modular data-center cooling skids. A severe national shortage of certified industrial HVAC technicians, pipefitters, and welders grants Comfort Systems extraordinary backlog pricing power and margin expansion across hyperscale campuses.
The leading mission-critical high-voltage electrical and mechanical contractor in the United States—wiring interior substations, switchgear, Uninterruptible Power Supply (UPS) halls, and backup generator yards across gigawatt hyperscale campuses at the lowest forward multiple in the specialty construction tier (17.63x Forward P/E) with a fortress net-cash balance sheet.
The literal "first check written" on any multi-hundred-acre AI data-center, semiconductor megafab, or reshoring industrial groundbreaking. Sterling’s E-Infrastructure Solutions division performs heavy rock blasting, precision site grading, water retention, and underground electrical duct-bank trenching 12 to 18 months before vertical building construction or server installation begins.
Tier 4: High-Speed Active Electrical Rack Interconnects (12.0% Total Weight)
The pioneer and market-share leader in Digital Signal Processor (DSP)-enabled Active Electrical Cables (AECs) for 800G and 1.6T rack-to-rack and switch-to-server AI interconnects. Credo’s purple copper AEC cables cut power consumption by roughly 50%, improve reliability, and cost significantly less than short-reach optical transceivers inside dense liquid-cooled clusters—driving +315.8% YoY EPS growth and compressing its PEG ratio to 0.18. Operational Note: At 28.93x Forward P/E, CRDO sits just 0.07x below our 29.0x valuation ceiling and must be monitored closely during quarterly audits.
9.4 The 5-Year Capex Wave Rotation Engine & 20-Quarter Trigger Schedule
In precision manufacturing, a rotary transfer machine does not engage all 11 cutting spindles at the exact same second; the raw casting rotates sequentially from Station 1 (rough milling) to Station 2 (drilling/tapping) and Station 3 (precision finishing). Our 11-stock AI portfolio operates on the exact same physical sequence across a five-year (20-quarter) buildout timeline:
- Wave 1 — Site Prep & Foundational Silicon Overweight (Years 0.0 to 1.5 / Q4 2026 – Q1 2028): Overweight the immediate ground-breaking excavation and initial rack-shipment leaders—STRL, NVDA, TSM, SKHY, MU, and CRDO—by +25% (1.25x weight multiplier) relative to base weights, while holding interior MEP contractors (VRT, FIX, EME), custom inference ASICs (AVGO), and nuclear power (TLN) at 1.00x base weight while shells are being graded.
- Wave 2 — Interior MEP Fit-Out, Liquid Cooling & Inference Flip (Years 1.5 to 3.0 / Q2 2028 – Q3 2029): As 2026–2027 groundbreakings become framed concrete buildings and hyperscalers shift budgets toward custom inference and interior fit-out, rotate profits out of early site excavation (STRL) and overweight VRT, FIX, EME, AVGO, and TLN (+25% / 1.25x multiplier).
- Wave 3A — Cycle Maturation, Memory/Builder Trim & Nuclear Scarcity (Years 3.0 to 3.75 / Q4 2029 – Q2 2030): Trim cyclical memory (SKHY, MU), early silicon (NVDA), and site excavation (STRL) by -35% (0.65x weight multiplier) as initial campus construction matures, while maintaining maximum overweight (+25%) in baseload nuclear power (TLN) as multi-gigawatt PPAs reprice higher.
- Wave 3B — The 4-Year GPU/HBM Hardware Replacement Cycle (Years 3.75 to 5.0 / Q3 2030 – Q4 2031): As the $1.5T+ of GPUs and HBM stacks installed in 2026–2027 hit their 4-year thermal and economic obsolescence wall, RE-LOAD (+25% / 1.25x multiplier) NVDA, TSM, SKHY, and MU alongside core nuclear anchor TLN, while trimming late-stage building contractors (VRT, FIX, EME, STRL, CRDO, AVGO).
- The Unbreakable 29.0x Forward P/E & 1.00 PEG Guardrail: Audit Forward P/E and PEG ratios quarterly during Weekend Offline Diagnostics. Any stock whose Forward P/E rises above 29.0x or whose PEG reaches 1.00 is automatically flagged for a 50% Valuation Trim (0.50x weight multiplier), with proceeds reallocated pro-rata into the remaining sub-25x holdings.
| Ticker (Base Wt) | Wave 1: Yrs 0.0 – 1.5 (Q4 '26 – Q1 '28) |
Wave 2: Yrs 1.5 – 2.75 (Q2 '28 – Q2 '29) |
Wave 3A: Yrs 2.75 – 3.75 (Q3 '29 – Q2 '30) |
Wave 3B: Yrs 3.75 – 5.0 (Q3 '30 – Q4 '31) |
Valuation Cap Guardrail |
|---|---|---|---|---|---|
| NVDA (11%) | PHASE IN (1.25x) Vera Rubin ramp |
CORE HOLD (1.0x) Peak cash flow |
TRIM -35% (0.65x) Cloud digestion |
RE-LOAD (1.25x) 4-Yr GPU refresh |
50% Trim if Fwd P/E > 29.0x |
| TSM (10%) | PHASE IN (1.25x) Yr 0–1.75 CoWoS/N2 |
CORE HOLD (1.0x) Yr 1.75–3.5 Foundry |
CORE HOLD (1.0x) Through Yr 3.5 |
RE-LOAD (1.25x) Yr 3.5+ 1.4nm refresh |
50% Trim if Fwd P/E > 29.0x |
| AVGO (9%) | CORE HOLD (1.0x) Yr 0–1.0 Base; IN Yr 1+ |
PHASE IN (1.25x) Yr 1.0–3.25 XPU flip |
CORE HOLD (1.0x) Yr 3.25–4.25 Hold |
TRIM -35% (0.65x) Yr 4.25+ Maturation |
50% Trim if Fwd P/E > 29.0x |
| SKHY (4%) | PHASE IN (1.25x) Yr 0–1.5 HBM4 lock |
CORE HOLD (1.0x) Yr 1.5–2.5 Hold |
TRIM -35% (0.65x) Yr 2.5–3.75 Supply cap |
RE-LOAD (1.25x) Yr 3.75+ HBM5 refresh |
50% Trim if PEG ≥ 1.00 |
| MU (4%) | PHASE IN (1.25x) Yr 0–1.5 HBM sold out |
CORE HOLD (1.0x) Yr 1.5–2.5 Hold |
TRIM -35% (0.65x) Yr 2.5–3.75 Cycle peak |
RE-LOAD (1.25x) Yr 3.75+ HBM5 refresh |
50% Trim if PEG ≥ 1.00 |
| TLN (12%) | CORE HOLD (1.0x) Yr 0–1.25 Base Wt |
PHASE IN (1.25x) Yr 1.25–3.0 AWS PPA |
PHASE IN (1.25x) Peak power scarcity |
PHASE IN (1.25x) Multi-decade nuclear |
50% Trim if Fwd P/E > 29.0x |
| VRT (11%) | HOLD Yr 0–0.75 IN Yr 0.75+ |
PHASE IN (1.25x) Yr 0.75–3.0 CDU boom |
CORE HOLD (1.0x) Yr 3.0–4.25 Installed |
TRIM -35% (0.65x) Yr 4.25+ Shells done |
50% Trim if Fwd P/E > 29.0x |
| FIX (10%) | HOLD Yr 0–0.75 IN Yr 0.75+ |
PHASE IN (1.25x) Yr 0.75–2.75 Chiller fit |
CORE HOLD (1.0x) Yr 2.75–3.75 Backlog |
TRIM -35% (0.65x) Yr 3.75+ Buildout peak |
50% Trim if Fwd P/E > 29.0x |
| EME (10%) | HOLD Yr 0–0.75 IN Yr 0.75+ |
PHASE IN (1.25x) Yr 0.75–3.0 HV wiring |
CORE HOLD (1.0x) Yr 3.0–4.0 Switchgear |
TRIM -35% (0.65x) Yr 4.0+ Late cycle |
50% Trim if Fwd P/E > 29.0x |
| STRL (7%) | PHASE IN (1.25x) Yr 0–1.75 First dirt check |
CORE HOLD (1.0x) Yr 1.75–2.75 Grading |
FIRST TRIM (0.65x) Yr 2.75+ Rotate to MEP |
TRIM -35% (0.65x) Site prep completed |
50% Trim if Fwd P/E > 29.0x |
| CRDO (12%) | PHASE IN (1.25x) Yr 0–2.25 800G/1.6T AEC |
CORE HOLD (1.0x) Yr 2.25–3.5 Rack scale |
TRIM -35% (0.65x) Yr 3.5+ Optical shift |
TRIM -35% (0.65x) Late-cycle harvest |
28.93x Now — Trim at >29.0x! |
| Ticker & Name | Tier & Role | Base Wt | Stage Wt | Allocation ($) | Sim Fwd P/E | Sim PEG | Active Phase Signal |
|---|
You now hold the complete blueprint for our 11 pre-cleared Elite Inventory monopolies across the $8.9 trillion full-stack AI buildout. Before advancing into Module 4 (Chapters 10–12) to run live derivative options spindles across these 11 tickers, verify that these three tolerances from The Investing Machine (Volume I) are locked into your Digital Clipboard:
- 1. Sleeve Weight vs. Total Account Net Liq (The 5% Master Rule, Volume I, Chapter 10, p. 76): The target weights in Table 9.1 (from 4.0% MU/SKHY to 12.0% TLN/CRDO) represent 100% of your AI Infrastructure Sleeve. When you size that sleeve at 15% to 40% of your total portfolio Net Liquidating Value (keeping 60%–80% in Dual-Yield T-Bill Collateral and the SPY/SPX Sandbox), even a 12% sleeve weight equals just 1.8% to 4.8% of your total portfolio equity—strictly obeying the 5% Master Rule!
- 2. The Automated 29.0x Forward P/E & 1.00 PEG Trim Guardrail: During your quarterly Weekend Offline Diagnostics, any holding that breaches 29.0x Forward P/E (watch CRDO at 28.93x closely) or 1.00 PEG is automatically trimmed by 50% and reallocated into the remaining sub-25x bottleneck leaders.
- 3. The Chapter 13 Sector Pivot Backup (Volume I, p. 99): If any individual semiconductor or construction holding ever suffers an idiosyncratic moat breach or executive scandal, you do not "baghold" or average down; you immediately execute a Sector Pivot into a broad market-cap-weighted ETF (such as SMH for semiconductors or SPY/XSP for the broad market).
Your raw material is pre-cleared and staged at the machine. Proceed to Module 4 (Chapter 10) to engage the Tier 1, Tier 2, and Tier 3 options spindles on these 11 stocks.
9.5 Troubleshooting Matrix: Operating the 11-Stock Dual-Screen Portfolio
A: Trailing GAAP multiples look backward into the rearview mirror, whereas institutional capital allocation looks through the windshield at contracted forward cash flows. Talen Energy’s trailing 12-month GAAP net income included one-time non-cash post-restructuring accounting items prior to the activation of its co-located 2.5 GW Susquehanna nuclear Power Purchase Agreement (PPA) with Amazon Web Services (AWS) and surging PJM capacity auction prices. Normalized for those one-time items, Talen’s operational baseline P/E is 37.90x, and its contracted forward earnings compress its Forward P/E to just 10.28x with a 0.35 PEG ratio—making it the cheapest nuclear baseload anchor in the United States.
A: Credo’s 28.93x Forward P/E is the exact reason we built the Market Multiple Stress Slider into the interactive visualizer above (try clicking the "+15% Rally" button in the simulator to watch CRDO automatically trigger a red >29x P/E Cap Alert!). Because Credo’s earnings are growing at over 300% YoY, its forward multiple naturally compresses each quarter as new analyst EPS revisions roll forward; however, if a rapid price rally pushes CRDO’s Forward P/E past 29.05x before the next earnings report, our mechanical rule dictates cutting the position weight in half (or locking in the top half with a Volume I Chapter 23 Zero-Cost Collar) and reallocating the freed capital into our sub-20x holdings (TLN, EME, STRL, AVGO, MU, SKHY).
A: That is the exact subject of Chapter 10 (the very next chapter). High-liquidity megacaps like NVIDIA (NVDA), Taiwan Semiconductor (TSM), Broadcom (AVGO), Micron (MU), Vertiv (VRT), and Credo (CRDO) feature deep options chains ideal for our 45-DTE Cash-Secured Puts, Bull Put Spreads, 80-Delta LEAPS PMCCs, and ZEBRAs. Conversely, specialty industrial contractors with wider options bid-ask spreads or foreign ADR structures (such as SK hynix ADR, Sterling Infrastructure, Comfort Systems, and EMCOR) must be filtered through Volume I’s $0.05–$0.10 Bid-Ask Liquidity Guardrail (Chapter 15 & Chapter 24)—combining physical share tranches with selective Mid-price limit orders.