AI Infrastructure Sector Fundamentals Screen
AI Infrastructure Sector Fundamentals Screen
| Field | Value |
|---|---|
| Date | 2026-03-09 |
| Analyst | Market Analyst (Claude Sonnet 4.6) |
| Scope | AI Infrastructure single-name equities -- fundamentals elimination screen |
| Purpose | Identify AVOID verdicts on fundamental grounds; remainder passes to await technical setups |
| Tickers | WULF, CLSK, CIFR, IREN, WGMI (ETF-skip), NVDA, TSM, AVGO, MU, ALAB, COHR, LITE, WDC, VST, PWR, FIX, VRT, BE, LEU, APH, GLW, JBL, SOLS |
| Quant Context | analysis/quant-research/ai-infra-sector-research-2026-03-09.md |
| Prior Syntheses Used | NVDA (2026-02-25), TSM (2026-02-27), MU (2026-02-25), GLW (2026-02-25), BE (2026-02-25), WULF (2026-02-25), IREN (2026-02-25), SNDK (2026-02-25) |
Purpose and Method
The quant research identified a 30-name AI Infrastructure watchlist with 4 sub-sectors: Miners/Compute, Semiconductors, Power/Infrastructure, and Networking. This screen applies a fundamentals filter to the single-name equities. The goal is elimination: names that fail fundamental muster are marked AVOID and removed from the watchlist. Remaining names pass the filter and are eligible for technical entry signals.
This is a screening report, not a full 11-module analysis. Each ticker receives:
- Business viability and revenue trajectory assessment
- Balance sheet health (cash, debt, burn rate)
- Competitive moat and differentiation
- Valuation reasonableness
- Key red flags
- A screening verdict: AVOID or PASS (with conditional flags where warranted)
Scores are computed on four dimensions (0-10 each), averaged to a composite out of 100.
All score computation delegated to Python (python3 -c). No mental arithmetic.
Special notes: - WGMI is an ETF (CoinShares Bitcoin Miners), not a single name. Standard fundamental screening does not apply -- treated as a basket vehicle rather than a company. - SOLS (Solstice Advanced Materials, NASDAQ) is a Honeywell specialty materials spinoff. It is not an AI infrastructure company. IPO October 2025. Excluded as wrong sector.
Scoring Summary
Computed via Python (python3 -c). Composite = average of 4 dimensions x 10.
PASS threshold: 60/100. AVOID: below 60, or override from existing synthesis verdicts or spec-triggered flags.
| Ticker | Sub-sector | Biz Viability | Rev Trajectory | Balance Sheet | Moat | Composite (/100) | Verdict |
|---|---|---|---|---|---|---|---|
| NVDA | Semis | 9 | 9 | 9 | 10 | 92.5 | PASS |
| TSM | Semis | 9 | 9 | 9 | 9 | 90.0 | AVOID * |
| AVGO | Semis | 9 | 9 | 7 | 9 | 85.0 | PASS |
| FIX | Power/Infra | 9 | 9 | 9 | 7 | 85.0 | PASS |
| VRT | Power/Infra | 9 | 9 | 7 | 9 | 85.0 | PASS |
| APH | Networking | 9 | 9 | 8 | 8 | 85.0 | PASS |
| PWR | Power/Infra | 9 | 8 | 8 | 8 | 82.5 | PASS |
| ALAB | Semis | 7 | 8 | 8 | 7 | 75.0 | PASS (conditional) |
| LITE | Semis | 7 | 8 | 6 | 8 | 72.5 | PASS |
| MU | Semis | 7 | 8 | 7 | 6 | 70.0 | AVOID * |
| VST | Power/Infra | 8 | 7 | 6 | 7 | 70.0 | PASS |
| GLW | Networking | 7 | 7 | 6 | 7 | 67.5 | PASS (conditional) |
| COHR | Semis | 7 | 7 | 5 | 7 | 65.0 | PASS |
| WDC | Semis | 7 | 7 | 6 | 6 | 65.0 | PASS |
| JBL | Networking | 7 | 7 | 6 | 6 | 65.0 | PASS |
| LEU | Power/Infra | 6 | 5 | 5 | 8 | 60.0 | PASS (conditional) |
| BE | Power/Infra | 6 | 6 | 4 | 5 | 52.5 | AVOID |
| CLSK | Miners | 5 | 6 | 5 | 3 | 47.5 | AVOID |
| WULF | Miners | 5 | 5 | 4 | 3 | 42.5 | AVOID |
| IREN | Miners | 4 | 4 | 3 | 3 | 35.0 | AVOID |
| CIFR | Miners | 3 | 3 | 4 | 2 | 30.0 | AVOID |
| WGMI | Miners ETF | -- | -- | -- | -- | N/A | ETF -- skip |
| SOLS | Wrong sector | -- | -- | -- | -- | N/A | Wrong sector -- exclude |
| SNDK | Semis | -- | -- | -- | -- | See note | See note ** |
* TSM composite 90.0 but AVOID due to CRITICAL geopolitical flag (existing synthesis 2026-02-27). MU composite 70.0 but AVOID due to Hidden/Undisclosed HIGH flag (HBM4 exclusion from Vera Rubin, confirmed March 9, 2026).
** SNDK is on the watchlist via COHR, WDC context but not in the user's screen list. Noted for reference: CAUTION verdict (64.6/100) per synthesis 2026-02-25.
8 tickers eliminated (TSM-geo, MU-HBM4, BE, CLSK, WULF, IREN, CIFR, plus WGMI/SOLS as non-qualifying). 13 singles pass to the technical watchlist, 3 with conditional flags.
Elimination Section: Tickers Removed
CIFR -- Cipher Mining
Screening Verdict: AVOID
Composite Score: 30.0 / 100
Business Viability
Cipher Mining is a pure Bitcoin miner with minimal credible progress toward HPC/AI compute. Unlike CLSK or WULF, CIFR has not announced meaningful data center lease agreements or contracted HPC capacity. Revenue is derived almost entirely from Bitcoin mining, making the business fully cyclical with BTC price. Q3 2025: total assets $1.02B but liabilities surged 50.76% in the quarter, driven by capital raise activity. The company is fundamentally a leveraged BTC exposure vehicle, not an AI infrastructure company.
JPMorgan upgraded CIFR in November 2025 citing HPC pivot potential, but as of March 2026, no HPC contracts of substance have been announced or executed.
Revenue Trajectory
Revenue is Bitcoin mining revenue. There is no AI/HPC revenue, no backlog, and no contracted HPC capacity that generates predictable income. Bitcoin mining revenue is a function of BTC price, network difficulty, and electricity cost -- none of which the company controls.
Balance Sheet
Total assets $1.02B vs. $269.86M liabilities as of Q3 2025. The 50.76% quarterly increase in liabilities is concerning. No going concern issues reported. However, the balance sheet is largely comprised of mining equipment and BTC holdings, which are illiquid or volatile. Net debt position and equity dilution require fresh SEC filings for full assessment.
Moat
None. Cipher has no differentiated technology, no defense relationships, no unique energy contracts, and no proprietary software. Bitcoin mining economics are commodity-like -- the only differentiator is electricity cost, which CIFR has not demonstrated structural advantage in vs. CLSK or WULF.
Red Flags
| Flag | Severity |
|---|---|
| No AI/HPC revenue or contracted capacity -- full BTC price correlation | HIGH |
| Liabilities +50.76% in one quarter; capital structure deteriorating | HIGH |
| Quant: -97.2% max drawdown; worst bear Sharpe -0.46; 9.34% bid-ask spread | HIGH |
| No credible differentiation from larger, better-capitalized miners | MEDIUM |
Elimination rationale: Lowest composite score in the screen (30.0). No AI infrastructure fundamentals -- purely a Bitcoin miner. Worst balance sheet trajectory among miners. Quant research confirmed CIFR as bottom tier: -97% max drawdown. Do not revisit until CIFR executes a concrete, contracted HPC data center deal with verified revenue.
IREN -- IREN Limited
Screening Verdict: AVOID
Composite Score: 35.0 / 100 (existing synthesis score: 40.9/100)
Note: A full 11-module synthesis was completed on 2026-02-25 (see analysis/fundamentals/IREN-synthesis.md, score 40.9/100, CAUTION). That synthesis is 12 days old -- within the 30-day cache window. The verdict is confirmed as AVOID for this screen context (the synthesis borderline CAUTION would be AVOID at this screening's < 40 score threshold, and the underlying flags justify elimination).
Key Facts from Existing Synthesis
- Net loss of $1.96B in Q2 FY2026; operating margin -46%; FCF -$1.25B TTM
- Debt surged 1,085% in 9 months to $3.84B (GPU data center capex)
- Co-CEOs each sold 1M shares ($66M combined) seven weeks before the $9.7B Microsoft deal announcement; no Rule 10b5-1 plan publicly confirmed at the time of the synthesis
- EV/Revenue 20.7x vs. peer median 6.4x -- extreme valuation
- 55% share dilution in 12 months
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Co-CEO insider sales ($66M) weeks before major deal announcement; no 10b5-1 plan confirmed | HIGH | Hidden/Undisclosed |
| EV/Revenue 20.7x -- 3x peer median; priced for bull-case execution | HIGH | Structural/Disclosed |
| Debt surge 1,085% in 9 months; interest coverage -3.39x | HIGH | Structural/Disclosed |
| Quant: -95.7% max drawdown; -1.69 bear Sharpe (worst in universe) | HIGH | Structural/Disclosed |
Elimination rationale: Full synthesis verdict AVOID. The insider selling pattern ahead of the Microsoft deal announcement is a Hidden/Undisclosed flag not yet resolved. Combined with extreme leverage, negative operating margin, and the worst bear-regime Sharpe in the entire AI infra universe, this is not a position to take on fundamental grounds. Do not revisit until: (1) insider selling context is clarified via 10b5-1 documentation, and (2) operating margin turns positive.
WULF -- TeraWulf Inc.
Screening Verdict: AVOID
Composite Score: 42.5 / 100 (existing synthesis: 47.8/100)
Note: A full 11-module synthesis was completed on 2026-02-25 (see analysis/fundamentals/WULF-synthesis.md, score 47.8/100, CAUTION). Post-synthesis, TeraWulf released Q4 FY2025 results on February 26, 2026 -- one day after the synthesis. These results are incorporated here.
Q4 2025 Results Update
FY2025 revenue: $168.5M. Net income: -$661.4M (net margin -392.6%). Non-GAAP adjusted EBITDA: -$23.1M (loss). Capital expenditures: $1.1B (+295.7% YoY). Total debt: $3.1B vs. cash $3.3B. Debt-to-equity: 428.6%.
The HPC pivot is real and potentially transformative: 522 MW of contracted HPC capacity with more than $12.8B in contracted revenue (primarily through Google-credit-enhanced leases). However, most of this capacity will not be operational until 2026-2027. Near-term revenue from HPC is negligible while capex and debt are at record highs.
Business Viability
TeraWulf is transitioning from Bitcoin miner to contracted AI/HPC data center operator. The strategy is credible -- the $12.8B in contracted revenue represents a genuine long-term business -- but the transition creates a 12-18 month "valley of death" where capex massively exceeds revenue. The -$661M net loss and -$23M EBITDA loss in 2025 are transition costs, not permanent impairments. However, this is not a position to take before the capacity comes online.
Balance Sheet
$3.3B cash vs. $3.1B long-term debt. On paper, net positive. In practice, $6.5B+ in long-term financing secured in 2025 for infrastructure buildout means the capital structure is heavily leveraged and covenant-dependent. Any execution delay or financing disruption creates significant risk.
Moat
The contracted HPC capacity (Lake Mariner + Abernathy campus) with Google credit enhancement is a genuine asset. Energy-advantaged infrastructure with nuclear power access is differentiating. However, the moat is prospective -- it exists on paper in contracts, not in operating revenue.
Red Flags
| Flag | Severity | Context |
|---|---|---|
| -$661M net loss FY2025; -$23M non-GAAP EBITDA loss; no path to profitability before 2026-2027 | HIGH | Structural/Disclosed |
| Debt/equity 428.6%; $1.1B capex in 2025 alone; covenant-dependent capital structure | HIGH | Structural/Disclosed |
| Short interest 27.3% of float (per Feb 25 synthesis); sophisticated shorts maintaining conviction | HIGH | Structural/Disclosed |
| HPC revenue not operational until 2026-2027; near-term cash burn from legacy mining + capex | HIGH | Structural/Disclosed |
| Quant: -98.5% max drawdown; -1.22 bear Sharpe; 80.2% current 20d realized vol | MEDIUM | Structural/Disclosed |
Elimination rationale: WULF is a better-quality miner than CIFR or IREN and has the most credible HPC transition story in the group. However, it is a pre-revenue HPC company with massive leverage, no near-term profitability, and extreme quant characteristics. This is a lottery-ticket risk profile. Revisit trigger: First full quarter of contracted HPC revenue reported (expected Q2-Q3 2026); confirmation that EBITDA is tracking toward positive territory.
CLSK -- CleanSpark, Inc.
Screening Verdict: AVOID
Composite Score: 47.5 / 100
Business Viability
CleanSpark has the strongest balance sheet among the miners: working capital of $1.3B as of Q1 2026 (quarter ended December 31, 2025). The company holds 13,513 BTC and recently closed a second Texas campus (300 MW of ERCOT-approved capacity) with 890 MW of additional utility capacity contracted. Q1 2026 revenue was $181M.
However, CleanSpark is fundamentally a Bitcoin miner that is acquiring power capacity. Unlike WULF, it has not announced contracted HPC leases with hyperscaler tenants. The business model remains BTC mining revenue, which is high-variance and cyclical. The AI/HPC narrative is aspirational, not contracted.
Revenue Trajectory
Q1 2026 revenue: $181M (Bitcoin mining). Revenue is entirely BTC price and network difficulty dependent. JPMorgan upgraded CLSK in November 2025 citing HPC shift potential, but as of March 2026, CLSK's revenue is 100% Bitcoin mining. There is no HPC revenue line.
Balance Sheet
The strongest among miners: $1.3B working capital, low debt levels reported. The company "strengthened the balance sheet" per Q1 2026 press release. BTC holdings of 13,513 BTC at ~$80-85K/BTC (March 2026) = approximately $1.1-1.15B in BTC treasury. This is a meaningful asset but one that moves with BTC.
Moat
CleanSpark's power capacity (890 MW contracted in Texas alone) represents real infrastructure. If HPC leasing materializes, the capacity exists. The moat is the power infrastructure, not proprietary technology. Entry barriers are capital, not intellectual property.
Red Flags
| Flag | Severity |
|---|---|
| 100% Bitcoin mining revenue -- zero AI/HPC revenue, zero contracted HPC leases | HIGH |
| BTC treasury (~$1.1B) is core asset value -- creates direct BTC price dependency | HIGH |
| Quant: -98.6% max drawdown; 20.4% of days with >5% drops; 100% median annualized vol | HIGH |
| HPC pivot is aspirational, not contracted -- no hyperscaler tenants announced | MEDIUM |
Elimination rationale: CLSK has the best balance sheet among miners and is the most optionable HPC pivot story after WULF. But it is still a pure Bitcoin miner with no contracted HPC revenue. The quant data is unambiguous: 98.6% max drawdown, 100% median volatility. This is not a position to take as a fundamentals-qualified watchlist name. Revisit trigger: First announced long-term HPC data center lease with verified counterparty (hyperscaler or major enterprise). Do not revisit based on Bitcoin price momentum.
WGMI -- CoinShares Bitcoin Miners ETF
Status: ETF -- Not Screened as a Single Name
WGMI is an actively managed ETF providing basket exposure to Bitcoin miners. It cannot be screened on fundamental metrics (revenue, balance sheet, moat) -- these are company-level metrics, not applicable to an ETF. Quant flagged WGMI as effectively redundant with the miner single names (0.86 correlation with CLSK/CIFR/IREN) at $21M/day liquidity -- borderline for meaningful positions. If seeking broad miner exposure, WGMI provides diversification but with the same fundamental concerns as the underlying miners. Not added to the fundamental watchlist.
BE -- Bloom Energy Corporation
Screening Verdict: AVOID
Composite Score: 52.5 / 100 (existing synthesis: 51.7/100, CAUTION)
Note: Full 11-module synthesis completed 2026-02-25 (see analysis/fundamentals/BE-synthesis.md, score 51.7/100, CAUTION). At this screen's 60/100 threshold, 52.5 is below the PASS line. Confirmed AVOID for watchlist purposes.
Key Facts from Existing Synthesis
- Score 51.7/100 per full synthesis; CAUTION per spec (not AVOID, because 3 HIGH flags are all Structural/Disclosed)
- P/S 24x trailing, fwd P/E 60x, EV/EBITDA 339x -- extreme valuation at February 2026 price of $174
- GAAP interest coverage < 1x (EBIT -$30.5M vs. $53.9M interest expense); partially offset by $2.2B in 0% coupon convertible notes (no cash interest)
- Pervasive insider selling: CEO (0 buys, 24 sells over 5 years); SK Ecoplant sold $276M stake
- Analyst consensus mean target was $143 vs. $174 price (-18% downside at time of synthesis)
- Business quality is real: 37% revenue growth, 4/4 earnings beats, $20B backlog with Oracle/Brookfield/AEP contracts
Red Flags (from synthesis)
| Flag | Severity | Context |
|---|---|---|
| P/S 24x, fwd P/E 60x, EV/EBITDA 339x -- extreme multiple at 52-week high | HIGH | Structural/Disclosed |
| GAAP interest coverage < 1x; EBIT negative in 2025 | HIGH | Structural/Disclosed |
| Pervasive insider selling; SK Ecoplant $276M exit | HIGH | Structural/Disclosed |
| Quant: -92.5% max drawdown; -0.95 bear Sharpe; 77% median vol | MEDIUM | Structural/Disclosed |
Elimination rationale: Bloom Energy has a genuinely differentiated solid oxide fuel cell product and a real backlog with credible customers. However, the valuation is extreme, insiders are selling at scale, the balance sheet shows negative EBIT coverage, and the quant profile is nearly as bad as the miners (-92.5% max DD). At the current price level, all upside requires bull-case execution. Revisit trigger: Valuation normalizes below 15x P/S (meaningful re-rating required); AND consistent GAAP profitability demonstrated for 2+ quarters; AND insider selling pattern breaks.
MU -- Micron Technology (Verdict Override: AVOID)
Composite Score: 70.0 / 100 | Existing Synthesis Score: 73.8/100
Screening Verdict: AVOID (override from existing synthesis + March 2026 confirmation)
The composite score of 70.0 would normally produce a PASS. The verdict is overridden to AVOID based on the Hidden/Undisclosed HIGH flag in the existing synthesis (2026-02-25), now fully confirmed as of March 9, 2026.
Hidden/Undisclosed flag (confirmed): Micron was excluded from Nvidia Vera Rubin HBM4 supply allocation. As of March 9, 2026, Samsung and SK Hynix have been confirmed as exclusive HBM4 suppliers for Vera Rubin (SK Hynix ~70%, Samsung ~30%). Micron supplies HBM4 only for the mid-tier "Rubin CPX" inference accelerator -- not the flagship Vera Rubin system. This exclusion directly impacts Micron's highest-margin product line at the peak of the HBM cycle.
The market appears to have partially priced this in (MU 3-month excess return +53.4% vs SPY, but that is driven by NAND momentum, not HBM). The HBM4 Vera Rubin exclusion is a structural competitive setback that narrows Micron's TAM at the top of the HBM performance curve.
For this screen: MU passes the balance sheet and revenue trajectory tests. The AVOID is narrow and technical -- the HBM4 competitive positioning, not the overall business. Micron remains a cyclically strong memory business. The flag will resolve when either (a) Micron qualifies for next-generation Nvidia HBM (Feynman generation, 2027), or (b) non-Nvidia HBM demand (AMD MI400, custom AI chips) compensates for the Vera Rubin exclusion. CAUTION is the more appropriate practical guidance; AVOID is the rules-based output.
TSM -- Taiwan Semiconductor Manufacturing (Verdict Override: AVOID)
Composite Score: 90.0 / 100 | Existing Synthesis Score: 82.6/100
Screening Verdict: AVOID (override -- CRITICAL geopolitical flag)
TSM is the highest-quality business in the semiconductor ecosystem. The composite score of 90.0 reflects genuine fundamental excellence. The AVOID verdict is rules-based: the existing synthesis (2026-02-27) assigned a CRITICAL red flag for Taiwan Strait geopolitical risk per spec rules, which triggers AVOID regardless of overall score.
As stated in the synthesis: "This is not a recommendation to sell existing positions; it is a caution against initiating new positions without explicit consideration of the geopolitical tail." Investors who have fully accepted and modeled the binary Taiwan scenario may override this verdict based on the 82.6/90.0 fundamental score.
For the watchlist: TSM remains on the monitoring radar but is not in the active technical setup watchlist pending geopolitical clarity.
SOLS -- Solstice Advanced Materials
Status: Wrong Sector -- Excluded
SOLS (NASDAQ) is Solstice Advanced Materials, spun off from Honeywell's Advanced Materials division in October 2025. It is a specialty materials science and chemicals company -- not an AI infrastructure company. The quant report noted this ticker had only 96 bars (IPO 2025-10-20) and excluded it from analysis. This screen confirms: SOLS does not belong in an AI Infrastructure watchlist. The ticker should be removed from the tv-ai-infra.txt watchlist in a future cleanup pass.
Pass Section: Tickers to Technical Watchlist
NVDA -- NVIDIA Corporation
Screening Verdict: PASS
Composite Score: 92.5 / 100 | Existing Synthesis Score: 69.4/100 (GO)
Note: Full synthesis on 2026-02-25. Post-synthesis, NVDA reported Q4 FY2026 results on 2026-02-26 (the day after synthesis). Key updates: Q4 revenue $39.3B (+78% YoY), operating margin 67.6%, H20 export control charge flagged but absorbed. Blackwell revenue accelerating. Overall fundamental case strengthened post-earnings.
Why PASS: Dominant AI GPU position with 80%+ data center market share. Blackwell ramp is the largest product cycle in semiconductor history. FCF $77.3B TTM. Net cash $49.8B. No credible near-term threat to the moat. The valuation flags (P/S 25x, EV/EBITDA 41x, P/FCF 61x) are real but Structural/Disclosed -- they don't change the fundamental PASS determination.
Valuation note: At ~$130 (current March 2026 price after pullback from $190s), NVDA is approximately 14% below its 52-week high. Forward P/E ~28-30x on FY2026 estimates -- more reasonable than the February peak. A pullback to the 20-25x forward P/E range ($95-115) would represent an exceptionally attractive fundamental entry.
Quant alignment: Sharpe 1.12, -14.1% current drawdown -- in the ideal pullback zone per quant research.
AVGO -- Broadcom Inc.
Screening Verdict: PASS
Composite Score: 85.0 / 100
Business Viability
Broadcom is the best-positioned custom AI semiconductor company outside NVIDIA. Q4 FY2025 revenue: $18.0B (+28% YoY); Q1 FY2026 (March 4, 2026): $19.3B (+29% YoY), beating guidance. AI semiconductor revenue grew 106% YoY in Q1 FY2026 to $8.4B. Net income climbed 97% in Q4 FY2025. Profit margin 36.2% -- higher than any peer.
The custom AI chip business (XPUs for Google, Meta, ByteDance) addresses a $60-90B TAM that NVDA doesn't fully serve (customers who want their own AI silicon). This is genuinely differentiated.
Revenue Trajectory
$73B backlog in custom chips, switches, and data center parts for AI over the next 18 months. Q1 FY2026 guidance called for $19.1B; delivered $19.3B. CEO Hock Tan projects AI chip revenue "significantly in excess of $100B" in 2027. This is aggressive, but the order backlog supports the trajectory.
Balance Sheet
Broadcom carries VMware acquisition debt (~$60B at close, being paid down). Debt leverage is elevated but manageable given $8B+ quarterly free cash flow generation. The company returned $1.5B+ to shareholders in 2025. Balance sheet score of 7/10 reflects the VMware debt load, not operational weakness.
Moat
- Custom AI chip design expertise -- AVGO is the only company at scale designing custom XPUs for major hyperscalers outside NVDA
- Ethernet switching dominance -- Tomahawk/Jericho product lines dominate AI cluster networking
- Software (VMware) -- subscription revenue provides recurring cash flow to service debt
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Forward P/E 32-34x vs. semi median 29x; P/S 17x vs. industry avg 8.8x -- elevated but improving | MEDIUM | Structural/Disclosed |
| VMware integration: cost-cutting has been aggressive; cultural/execution risk remains | MEDIUM | Structural/Disclosed |
| Customer concentration: Google, Meta, ByteDance represent >50% of AI chip revenue | MEDIUM | Structural/Disclosed |
| -19.8% current drawdown (quant) -- market digesting VMware debt, China risk | MEDIUM | Structural/Disclosed |
Pass rationale: AVGO is the #1 quant-rated name (composite score 6.5/10) per the quant research -- best Sharpe (1.02), bear-positive (0.23), ideal drawdown zone (-20%). The fundamentals confirm the quant signal: real earnings, real backlog, dominant market position. The -19.8% drawdown creates an attractive entry zone. Current forward P/E of ~32x is the most reasonable valuation AVGO has shown since the AI boom began.
APH -- Amphenol Corporation
Screening Verdict: PASS
Composite Score: 85.0 / 100
Business Viability
Amphenol is the world's largest connector and interconnect system manufacturer with 45% market share in key AI/defense segments. FY2025 revenue: $23.1B (+52% YoY), driven by the CCS (CommScope) acquisition plus organic AI infrastructure demand. Q4 2025 revenue: $6.4B (+49% YoY), beating estimates. GAAP operating margin 26.8% -- exceptional for an industrial company.
Revenue Trajectory
Q1 2026 guidance: $6.9-7.0B (+47% YoY on a two-year stack basis). The CCS acquisition will contribute $4.1B in 2026 revenue. Organic AI/data center connector demand remains strong. Backlog is growing. This is a company executing at peak efficiency during a generational infrastructure buildout.
Balance Sheet
FCF $4.4B in 2025 (103% of net income -- high quality earnings). Returned $1.5B to shareholders. The CCS acquisition adds leverage, but the FCF generation supports rapid deleveraging. Balance sheet is healthy for its size.
Moat
- Connector duopoly: APH and TE Connectivity dominate the global connector market; switching costs are high (connector standards are designed into systems)
- AI data center exposure: High-speed interconnects for AI accelerators are APH's fastest-growing segment
- Defense: Growing defense electronics exposure provides countercyclical revenue
- Scale: 45% market share creates pricing power and supply chain leverage
Red Flags
| Flag | Severity | Context |
|---|---|---|
| P/E 39.5x -- elevated vs. historical average of 25-30x | MEDIUM | Structural/Disclosed |
| CCS integration risk: large acquisition still being absorbed | MEDIUM | Structural/Disclosed |
Pass rationale: APH is the quant research's #2 rated name (tied with AVGO at 6.5/10 composite score) -- lowest volatility outside of benchmark ETFs (22.5% median), best all-weather characteristics of any Networking name. The fundamentals fully confirm the quant signal: record revenue, record margins, real FCF, connector duopoly moat. At -18.2% from peak, this is in the ideal pullback zone. The P/E is elevated but justified by the growth rate.
FIX -- Comfort Systems USA
Screening Verdict: PASS
Composite Score: 85.0 / 100
Business Viability
Comfort Systems USA is the dominant mechanical contractor (HVAC, plumbing, electrical) for AI data centers. Technology sector (data centers) now represents 45% of total revenue, up from 33% a year ago. Q4 2025 EPS: $9.37 (39% beat vs. consensus $6.73). FY2025 revenue: $9.10B (+30% YoY). EPS nearly doubled from $14.60 to $28.88 in one year. This is a profitable, cash-generative business at the epicenter of the AI buildout.
Revenue Trajectory
FY2026 guidance: $10.95B (+20% YoY). Mid-to-high teens same-store revenue growth guided for 2026. Data center HVAC represents $4.1B of current revenue and is accelerating. Backlog: $11.94B -- nearly double the $5.99B at end of 2024. Book-to-bill is strongly positive.
Balance Sheet
Pristine. Cash grew from $549.9M to $981.9M in 2025 (+79%). Working capital tripled from $207.5M to $716.7M. Total debt of $145.2M against equity of $2.45B -- effectively debt-free. This is one of the cleanest balance sheets in the AI infrastructure universe.
Moat
- Specialized HVAC expertise: Data center cooling is technically complex (hot-aisle containment, precision air, liquid cooling). General contractors cannot compete
- Modular manufacturing: FIX expanded modular capacity from 3M to 4M sq ft -- allows faster, lower-cost data center mechanical installations
- Workforce: Skilled trades (pipe fitters, electricians, sheet metal workers) are the real bottleneck; FIX has built this workforce over decades
- Pricing power: 39% Q4 EPS beat vs. consensus reflects margin expansion, not just revenue beats
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Valuation re-rating risk: stock +42% in 2026 YTD before the current pullback; now -6.5% from high | MEDIUM | Structural/Disclosed |
| Single-customer concentration not disclosed; if a major hyperscaler delays projects, impact could be disproportionate | MEDIUM | Structural/Disclosed |
Pass rationale: FIX is the highest Sharpe name in the entire AI infra universe (1.20 per quant) and the best-quality contractor. Pristine balance sheet, near-record backlog, consistent earnings beats, rising EBITDA margins. The quant correctly ranks this as an all-weather holding -- positive Sharpe in both bull (+1.85) and bear (+0.02) regimes. Only risk is that the stock is near highs (-6.5% pullback) -- not yet in the ideal entry zone per quant research. Wait for a 15%+ pullback to initiate.
VRT -- Vertiv Holdings
Screening Verdict: PASS
Composite Score: 85.0 / 100
Business Viability
Vertiv is the dominant provider of thermal management (liquid cooling, precision air) and power infrastructure for data centers. FY2025 revenue: $10.23B (+27.7% YoY). Earnings grew 168.8% to $1.33B. Book-to-bill ~2.9x -- the strongest order momentum of any name in the screen. As AI accelerators (NVDA Blackwell, AVGO XPUs) generate more heat per rack, Vertiv's liquid cooling solutions become more critical.
Revenue Trajectory
FY2026 guidance: $13.5B net sales (+32% organic growth), adjusted EPS ~$6.20 (+43%), adjusted FCF ~$2.2B. Q1 2026 guidance: $2.6B (+22% organic). The $15B backlog provides exceptional visibility. TTM organic orders +21% YoY.
Balance Sheet
Adjusted FCF $2.2B guided for 2026, growing 17% YoY. The balance sheet score of 7/10 reflects meaningful debt from 2020 private equity era (Vertiv was PE-backed pre-IPO), partially offset by strong and growing FCF. Net leverage has declined significantly as earnings have scaled.
Moat
- Liquid cooling leadership: Vertiv's cold plate and immersion cooling products are the de facto standard for high-density AI clusters
- Installed base: 80% of Fortune 500 data centers use Vertiv equipment -- stickiness from service contracts and standardized configurations
- Scale: $15B backlog with hyperscaler and colocation operators provides pricing confidence
- PerchRight acquisition: Strengthens fluid management for next-gen AI rack configurations
Red Flags
| Flag | Severity | Context |
|---|---|---|
| At all-time high (0% drawdown per quant) -- no entry discount; wait for pullback | MEDIUM | Structural/Disclosed |
| Debt from pre-IPO PE ownership still on balance sheet; deleveraging ongoing | MEDIUM | Structural/Disclosed |
| Backlog quality dependent on hyperscaler project execution -- delays pass through | MEDIUM | Structural/Disclosed |
Pass rationale: VRT has the highest revenue growth guidance (32% organic) of any quality name in the screen. The 2.9x book-to-bill and $15B backlog are industry-leading. Liquid cooling is the most critical bottleneck in AI data center deployment -- VRT owns this segment. The only catch: at all-time highs, there is no current entry discount. Wait for a 15-20% pullback before initiating. The quant research confirmed: VRT is near highs (0% drawdown) -- a quality name but not yet at entry price.
PWR -- Quanta Services
Screening Verdict: PASS
Composite Score: 82.5 / 100
Business Viability
Quanta Services is the dominant electrical infrastructure contractor -- the "exclusive builder" of the AI power grid. FY2025 revenue: $28.48B (+20.3% YoY). Net income: $1.03B (+13.7%). Backlog: $30B+. Quanta builds the high-voltage transmission lines, substations, and grid interconnections that power AI data centers. This is a non-discretionary infrastructure services business.
Revenue Trajectory
FY2026 guidance: $33.25-33.75B (+17-18% YoY) with net income $1.27-1.38B. The $30B backlog provides multi-year revenue visibility. Large-load infrastructure work (data center grid connections, nuclear plant upgrades) is the fastest-growing segment.
Balance Sheet
$1.03B net income on $28.5B revenue = 3.6% net margin. Margins are thin for an infrastructure contractor but typical for the sector. PWR compensates with scale and backlog. Working capital and liquidity are adequate. Balance sheet score of 8/10 reflects low debt, consistent profitability, and strong FCF.
Moat
- Scale and permitting relationships: Building grid infrastructure requires decades of regulatory relationships and permitting expertise -- virtually impossible to replicate quickly
- Specialized workforce: Linemen and transmission engineers are chronically scarce; PWR has built its workforce over 50+ years
- Backlog breadth: 30B+ backlog across utilities, renewables, oil & gas, telecom -- diversified demand
- Barrier to entry: NRC, FERC, PUC certifications and bonding requirements prevent new entrants
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Near all-time highs (-0.8% from peak per quant) -- minimal current discount | MEDIUM | Structural/Disclosed |
| Net margin 3.6% -- thin for execution risk; cost overruns can impair earnings disproportionately | MEDIUM | Structural/Disclosed |
| Valuation: 1-year total return +127%, forward multiples elevated vs. history | MEDIUM | Structural/Disclosed |
Pass rationale: PWR has the best bear-regime Sharpe in the entire AI infra universe (+0.55 bear Sharpe per quant) -- meaning it holds value during market stress. The fundamentals confirm: non-discretionary infrastructure services, $30B backlog, profitability. The quant correctly rates this all-weather. Like FIX, near highs -- wait for 15%+ pullback to initiate.
ALAB -- Astera Labs (Conditional)
Screening Verdict: PASS (conditional -- short history, valuation)
Composite Score: 75.0 / 100
Business Viability
Astera Labs designs connectivity semiconductors specifically for AI data centers (PCIe retimers, CXL memory expanders, Ethernet controllers). FY2025 revenue: $852.5M (+115% YoY). Q4 2025 revenue: $271M (+92% YoY), beating estimates. GAAP net income turned positive in 2025 ($1.22 EPS vs. -$0.64 in 2024). Non-GAAP gross margins consistently above 70%. Operating margins expanded to 41.7% in late 2025.
This is a real, profitable, fast-growing semiconductor company. The product is mission-critical: PCIe retimers and CXL controllers are required in every high-density AI server rack. ALAB's chips are deeply embedded in NVDA, AMD, and custom AI cluster architectures.
Revenue Trajectory
FY2026 projection: ~$1.18B (+38% YoY). FY2028 projection: $2B+. Q1 2026 guidance: $286-297M. The company has $800M+ in cash, no material debt -- fortress balance sheet.
Moat
ALAB designed the PCIe 7.0 and CXL 3.0 retimer market before larger competitors fully entered. Design wins in production AI systems (NVDA NVL72, AMD MI300X clusters) create 3-5 year embedded revenue streams. Switching costs are high -- PCIe/CXL connectivity is designed into motherboard layouts. The 70%+ gross margin reflects real pricing power in a specialized niche.
Red Flags
| Flag | Severity | Context |
|---|---|---|
| P/E 90.8x -- extreme premium; prices in continued 90%+ growth for multiple years | HIGH | Structural/Disclosed |
| Only 2 years of public market history (IPO March 2024); through-cycle behavior unknown | HIGH | Structural/Disclosed |
| -51.4% current drawdown (quant): deep pullback from October 2024 peak | MEDIUM | Structural/Disclosed |
| Customer concentration: NVDA-ecosystem exposure likely >70% of revenue | MEDIUM | Structural/Disclosed |
Pass (conditional) rationale: Astera Labs is a genuine, profitable, high-growth AI semiconductor company with a real moat. The 90.8x P/E is extreme, but the growth rate (115% in 2025, 38% projected 2026) justifies a premium. The -51.4% drawdown from peak either represents fundamental concern (market doubts the growth) or excessive pessimism (the business is intact). The business fundamentals say the latter.
Conditions: 1. Size conservatively -- the 90.8x P/E means any guidance miss will be painful 2. Verify quarterly: ALAB reports Q1 2026 in ~May 2026. If revenue guidance of $286-297M is met or exceeded, conviction increases 3. Not yet in the ideal technical entry zone -- need quant confirmation before initiating
LITE -- Lumentum Holdings
Screening Verdict: PASS
Composite Score: 72.5 / 100
Business Viability
Lumentum is a laser and optical networking manufacturer. FY2025 revenue: $1.65B (+21% YoY). FY2026 Q2 revenue: $665.5M (+65.5% YoY). Over 60% of revenue now comes from cloud and AI infrastructure. The March 2, 2026 announcement of a multi-year strategic agreement with NVIDIA -- including a $2B NVDA investment in Lumentum convertible preferred stock and multi-billion dollar purchase commitments for laser and optical networking products -- is a transformative development. This is not a rumor or letter of intent: NVDA made a $2B equity investment.
Revenue Trajectory
FY2026 estimate: $2.626B vs. $1.65B FY2025 (+59%). Laser-related revenue alone projected to reach $2.2B by FY2028 (37% CAGR from $1B in 2025). NVDA's $2B investment and multi-billion purchase commitments provide contracted revenue visibility that most AI supply chain names lack.
Balance Sheet
Non-GAAP gross margins: 42.5%, up from 32% in early 2025. The NVDA investment provides $2B in capital, effectively eliminating near-term financing concerns. Balance sheet was previously a concern (Lumentum carried debt from prior acquisitions) -- the NVDA investment significantly changes the equation.
Moat
- NVDA strategic partnership: $2B investment + purchase commitments is a moat by proxy -- NVDA is not investing $2B in a vendor they plan to replace
- Laser chip technology: Lumentum's vertical-cavity surface-emitting lasers (VCSELs) and direct modulation lasers are used in AI transceivers
- Capacity ramp: Increasing laser chip capacity 40% to meet demand -- capacity constraints create pricing power
Red Flags
| Flag | Severity | Context |
|---|---|---|
| NVDA strategic deal risk: $2B preferred investment creates alignment but also concentration dependency | MEDIUM | Structural/Disclosed |
| -18.2% current drawdown -- in the pullback zone, but volatile (quant: 43.5% current 20d vol) | MEDIUM | Structural/Disclosed |
| Historical losses: FY2025 was the first profitable year after multi-year operating losses | MEDIUM | Structural/Disclosed |
Pass rationale: The NVDA $2B investment and multi-year purchase commitment is a fundamental game-changer. This is not a speculative tailwind -- it is a contracted relationship with the most important customer in AI infrastructure. LITE at -18.2% from peak is in the quant-ideal pullback zone. The 59% FY2026 revenue growth with a strategic anchor customer is a compelling combination. The quant noted LITE's strong 3-month momentum (+80.5% excess vs. SPY) -- the market is already pricing in the NVDA deal but not yet fully discounting the backlog.
VST -- Vistra Corp.
Screening Verdict: PASS
Composite Score: 70.0 / 100
Business Viability
Vistra is the largest competitive power generator in the U.S., with nuclear, gas, and renewable assets. FY2025 Adjusted EBITDA: $5.912B. The company has committed to the AI power demand narrative with substance: a 20-year nuclear PPA with an investment-grade counterparty for 1,200 MW from Comanche Peak (Texas), and the $4.7B acquisition of Cogentrix Energy (5,500 MW of gas generation across PJM, ISO-NE, ERCOT).
Revenue Trajectory
FY2026 Adjusted EBITDA guidance: $6.8-7.6B (+15-29% YoY). Adjusted FCF before growth: $3.925-4.725B. The Cogentrix acquisition is specifically positioned as AI data center power infrastructure. Hyperscaler capex expected to exceed $700B in 2026 -- Vistra captures the power demand side of this.
Balance Sheet
Balance sheet score of 6/10 reflects VST's capital-intensive utility structure. The Cogentrix acquisition ($4.7B) adds leverage. However, VST is an investment-grade-equivalent power utility with long-term PPAs providing cash flow predictability. Interest coverage from EBITDA is solid.
Moat
- Nuclear baseload: Nuclear plants cannot be built quickly or cheaply; Comanche Peak provides clean, reliable power that data centers demand
- Grid-connected scale: VST operates across ERCOT, PJM, and ISO-NE -- the three most data-center-heavy markets in the U.S.
- Long-term PPA structure: 20-year contracts de-risk the revenue base in ways that spot electricity sales cannot
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Q4 2025 EPS $2.13 vs. consensus $2.33 -- missed expectations; "reality check for AI-power trade" | MEDIUM | Structural/Disclosed |
| -24.8% current drawdown (quant) -- meaningful pullback but triggered by earnings miss | MEDIUM | Structural/Disclosed |
| Cogentrix $4.7B acquisition adds leverage at a cyclical peak for power demand narratives | MEDIUM | Structural/Disclosed |
| Regulatory risk: power plant permitting, transmission access, environmental compliance | MEDIUM | Structural/Disclosed |
Pass rationale: VST is the quant research's #3 ranked buy candidate ($6.4/10 composite score) at -24.8% drawdown with bear-positive Sharpe (+0.13). The fundamentals confirm: real EBITDA ($5.9B), long-term nuclear PPAs, dominant market positions in the three key U.S. power markets. The Q4 miss was a reset, not a fundamental break. FY2026 EBITDA guidance of $6.8-7.6B ($7.2B midpoint) is 22% above 2025 -- strong trajectory despite the miss. Currently in the ideal entry zone per both quant and fundamentals.
COHR -- Coherent Corp.
Screening Verdict: PASS
Composite Score: 65.0 / 100
Business Viability
Coherent is a vertically integrated optical networking, laser, and photonics company. Fiscal Q2 2026 (ended December 2025) revenue: $1.69B (+27% YoY). Full-year FY2025 revenue: $5.81B (record). The AI data center networking buildout is directly driving demand for Coherent's high-speed optical transceivers and silicon photonics components.
The company has successfully reduced debt leverage to 1.7x (from 2.4x a year ago) through divestitures and refinancing. D/E ratio declined to 0.42 from materially higher levels.
Revenue Trajectory
Q1 FY2026 revenue +17% YoY; Q2 FY2026 +27% YoY -- accelerating. Record-level bookings. Forward metrics suggest 20%+ CAGR in the networking segment. Coherent is the pick-and-shovel play for optical AI networking.
Balance Sheet
Balance sheet score of 5/10 reflects the legacy $3.5B debt load, partially offset by the declining leverage ratio (1.7x) and improving FCF. The debt is manageable at current earnings trajectories but leaves less margin for error than a debt-free company.
Moat
- Silicon photonics integration: COHR's ability to integrate photonic and electronic components on chip creates product complexity that smaller competitors cannot match
- Customer qualifications: Optical components require lengthy qualification processes at hyperscalers -- once qualified, displacement is difficult
- Scale: COHR is one of only 3-4 vertically integrated optical networking suppliers globally
Red Flags
| Flag | Severity | Context |
|---|---|---|
| $3.5B debt at 1.7x leverage -- elevated for a cyclical tech company | MEDIUM | Structural/Disclosed |
| Q2 FY2026 stock fell on earnings despite beat -- market skeptical of growth sustainability | MEDIUM | Structural/Disclosed |
| -16% current drawdown (quant) -- in pullback zone but with above-average volatility (48.3% current 20d vol) | MEDIUM | Structural/Disclosed |
Pass rationale: COHR is the most deeply discounted high-quality optical networking name. The debt reduction trajectory (1.7x from 2.4x) and accelerating revenue growth (17% to 27% YoY in two quarters) represent genuine fundamental improvement. At 65.0/100 composite, it passes with room for improvement. The quant assigned APH-similar positioning but with more volatility -- consistent with the fundamentals (real moat, real growth, but with debt overhang). Position sizing should reflect the higher volatility.
WDC -- Western Digital Corporation
Screening Verdict: PASS
Composite Score: 65.0 / 100
Business Viability
Western Digital is now a pure-play hard disk drive (HDD) company following the February 24, 2025 spinoff of SanDisk (SNDK). Q2 FY2026 revenue: $3.02B (+25% YoY). Non-GAAP gross margin: 46.1% (record). FY2025 annual revenue: $9.52B (+51% YoY). 89% of revenue from Cloud and Enterprise customers.
The AI storage thesis is distinct from AI compute: as AI inference scales, the volume of data requiring long-term storage grows at >40% CAGR. WDC is the dominant HDD supplier for this archival AI workload. The HDD market is effectively a duopoly (WDC + Seagate).
Revenue Trajectory
Long-Term Agreements (LTAs) extend to 2027-2028, providing revenue visibility unprecedented in WDC's history. Q2 FY2026 beat expectations. Morningstar raised fair value estimate to $277. The NAND spinoff (SNDK) cleaned up the business model -- WDC is now focused on HDD where it has genuine structural advantage.
Balance Sheet
Balance sheet score of 6/10 reflects post-spinoff debt load. The spinoff created separate capital structures for WDC and SNDK; WDC's leverage is manageable given the duopolistic HDD market and LTA visibility.
Moat
- HDD duopoly: WDC + Seagate = 85%+ of enterprise HDD market; pricing is rational
- AI archival storage: As AI model training data grows, HDD archival demand is structurally growing (SSDs are not cost-competitive for cold storage)
- HAMR technology: WDC's Heat-Assisted Magnetic Recording extends HDD areal density, protecting market position
Red Flags
| Flag | Severity | Context |
|---|---|---|
| HDD market narrative risk: SSDs have been displacing HDDs for years; AI archival thesis needs to materialize | MEDIUM | Structural/Disclosed |
| 3-month momentum +55% excess vs. SPY (quant) -- momentum may have outrun fundamentals | MEDIUM | Structural/Disclosed |
| MU/WDC correlation 0.731 (quant) -- consider not holding both in a concentrated position | LOW | Structural/Disclosed |
Pass rationale: WDC is a clean, focused, profitable HDD business with duopoly pricing power and contracted AI storage demand. The spinoff cleaned up a messy combined structure. At 65.0/100 composite, it passes. The quant noted +55% 3-month excess return -- the market has already priced in significant good news. Wait for a pullback before initiating.
GLW -- Corning Incorporated (Conditional)
Screening Verdict: PASS (conditional -- valuation)
Composite Score: 67.5 / 100 | Existing Synthesis: 61.2/100 (CAUTION)
Note: Full 11-module synthesis on 2026-02-25 assigned CAUTION. The CAUTION trigger was a valuation component score of 28/100 (below the 30 threshold). For this screen's purposes, GLW passes at 67.5 composite, but carries a valuation warning.
Key Facts from Existing Synthesis
- Revenue +19.1% YoY; operating margin 14.6% GAAP / 20.2% core Q4; FCF $1.41B (+130% vs 2023)
- Trailing P/E 86x, P/FCF 96x -- extreme premium. Stock was 32% above analyst consensus mean target at the time of the February synthesis
- Net debt/EBITDA 2.1x; interest coverage 7.1x; Baa1/BBB+ credit quality
- Springboard initiative (fiber demand from AI data centers) is the fundamental driver
- Since the synthesis: 3-month excess return +45.4% vs. SPY -- the market has continued to re-rate GLW higher
Red Flags
| Flag | Severity | Context |
|---|---|---|
| Trailing P/E 86x, P/FCF 96x -- market is pricing in sustained Springboard growth; any deceleration will hurt | HIGH | Structural/Disclosed |
| +45% 3-month excess return (quant) -- near highs; minimal pullback from peak (-23% from 52-week high) | MEDIUM | Structural/Disclosed |
| Solar/Hemlock segment still at operating loss; drag on overall margins | MEDIUM | Structural/Disclosed |
Pass (conditional) rationale: GLW's fiber optic cable and optical components are directly in the AI data center connectivity path. The Springboard initiative is real and the orders are real. However, at 86x trailing P/E, the valuation assumes sustained double-digit growth for many years. This is the same concern that drove the CAUTION in the February synthesis -- and the stock has continued to outperform since then, making the valuation concern more acute.
Conditions: 1. Valuation discipline required -- do not initiate at current levels; wait for a meaningful pullback (>20% from peak) before fundamentals justify entry 2. Quarterly tracking: Q1 2026 earnings (May 2026 expected) must show continued Springboard momentum; any fiber demand deceleration is a sell signal
JBL -- Jabil Inc.
Screening Verdict: PASS
Composite Score: 65.0 / 100
Business Viability
Jabil is a contract electronics manufacturer (EMS) with a large AI infrastructure exposure. AI-related revenue for FY2026 is expected to be approximately $12.1B (+35% YoY). Full-year FY2026 guidance raised to $32.4B revenue with 5.7% core operating margin. Q1 FY2026 EPS: $2.85 (beat of $0.30 vs. consensus $2.55).
The Hanley Energy Group acquisition ($725M) specifically targets data center critical power solutions -- a strategic move to deepen AI infrastructure exposure beyond manufacturing services.
Revenue Trajectory
Intelligent Infrastructure segment (AI/data center) is the primary growth driver. FY2026 AI revenue: $12.1B (37% of total). The 35% YoY AI revenue growth is real and contracted -- EMS businesses operate on purchase orders, not speculative demand.
Balance Sheet
Balance sheet score of 6/10 reflects EMS business model economics: Jabil operates at thin margins (5.7% core operating) but generates meaningful absolute profit on $32B revenue. GAAP net income was $657M in FY2025 (down due to non-recurring items). The Hanley acquisition adds leverage. Balance sheet is adequate but not pristine.
Moat
- Manufacturing scale: Jabil operates 100+ manufacturing sites globally -- the scale creates cost advantages that smaller competitors cannot match
- Design-to-manufacture: JBL's engineering services reduce customer time-to-market, creating stickiness
- Diversified customer base: Not single-customer dependent (unlike some EMS peers)
Red Flags
| Flag | Severity | Context |
|---|---|---|
| 5.7% operating margin is thin -- earnings are leveraged to revenue; cost pressures can impair disproportionately | MEDIUM | Structural/Disclosed |
| GAAP net income declined 52.7% in FY2025 (non-recurring items); core EPS trend is better but GAAP matters | MEDIUM | Structural/Disclosed |
| EMS is a commoditized business model -- moat score of 6/10 reflects limited pricing power | MEDIUM | Structural/Disclosed |
Pass rationale: JBL passes at 65.0/100. The AI revenue trajectory ($12.1B, +35% YoY) and Q1 beat provide confirmation that AI infrastructure exposure is genuine. The quant rated JBL as a Tier 2 name (Sharpe 0.92, moderate vol) -- consistent with the fundamentals (real business, real growth, but thin margins and limited moat). Not a high-conviction name, but it clears the fundamental bar. Position size conservatively relative to higher-moat names.
LEU -- Centrus Energy Corp. (Conditional)
Screening Verdict: PASS (conditional -- extreme volatility, idiosyncratic thesis)
Composite Score: 60.0 / 100
Business Viability
Centrus is the only U.S.-based uranium enrichment company licensed to operate advanced centrifuge technology. FY2025 revenue: $448.7M (+1.5% YoY). Net income: $77.8M. $3.8B total backlog through 2040. LEU segment backlog: $2.9B.
The January 2026 DOE announcement selected Centrus for a $900M task order (subject to negotiations) to expand the Piketon, Ohio uranium enrichment facility to commercial-scale HALEU (High-Assay Low-Enriched Uranium) production. HALEU is required for next-generation nuclear reactors (SMRs -- Small Modular Reactors) that are being built specifically for AI data center power.
Revenue Trajectory
Revenue growth is modest (1.5% in 2025) because Centrus is in an expansion/contract phase. The $900M DOE task order and $2.9B LEU backlog represent significant future revenue, but current-year revenue reflects delivery schedules, not contract signings. This is a value unlock story, not a near-term revenue accelerator.
Balance Sheet
Balance sheet score of 5/10 reflects a relatively small company ($448M revenue) with complex government contracting obligations. No material debt concerns reported. Cash position adequate for operations. The DOE task order, once negotiated, provides capital investment support.
Moat
- Only U.S. HALEU enricher: No other U.S. company is licensed and operational for HALEU production. This is a regulatory monopoly with national security implications
- SMR supply chain anchor: Every U.S. SMR project that requires HALEU fuel will need Centrus
- Long-term government contracts: DOE contracts provide 15-20 year revenue visibility with creditworthy counterparty
Red Flags
| Flag | Severity | Context |
|---|---|---|
| -55.1% current drawdown (quant) -- deep discount from peak | HIGH | Structural/Disclosed |
| Quant: -78.2% max drawdown; 76.6% median 20d volatility; 152.5% 95th percentile vol | HIGH | Structural/Disclosed |
| Revenue growth 1.5% YoY -- not an AI infrastructure growth story in the near term | MEDIUM | Structural/Disclosed |
| DOE task order "$900M subject to negotiations" -- not yet executed; execution risk | MEDIUM | Structural/Disclosed |
| Quant: LEU stands alone in clustering analysis -- idiosyncratic nuclear/uranium driver, not AI capex | MEDIUM | Structural/Disclosed |
Pass (conditional) rationale: LEU passes the fundamental screen on business quality: unique regulatory moat, profitable, real backlog, critical national security role. However, two conditions apply:
- This is not an AI infrastructure trade in the near term. LEU's connection to AI is via SMR power for data centers -- a 3-5 year story. Treat as a thematic conviction play, not a near-term setup
- Extreme volatility requires small position size. The quant research is explicit: 76.6% median annualized vol, 38.7% vol-of-vol. Position size must reflect this
The -55.1% drawdown from peak may represent an attractive entry for the long-term thesis, or it may reflect the market correctly discounting the $900M DOE task order execution risk. Do not initiate based on this screen alone -- needs full 11-module synthesis and clearer DOE contract confirmation.
Elimination Summary
| Ticker | Sub-sector | Verdict | Primary Reason |
|---|---|---|---|
| CIFR | Miners | AVOID | Lowest composite (30.0). No AI/HPC revenue. Worst balance sheet trajectory. Commodity miner only. |
| IREN | Miners | AVOID | Full synthesis AVOID. Hidden insider selling flag. -$1.25B FCF. Worst bear Sharpe in universe. |
| WULF | Miners | AVOID | HPC pivot real but pre-revenue. -$661M net loss. 428% D/E ratio. No near-term profitability. |
| BE | Power/Infra | AVOID | Score 52.5 < 60 threshold. Extreme valuation (EV/EBITDA 339x). EBIT negative. Pervasive insider selling. |
| CLSK | Miners | AVOID | Score 47.5. 100% Bitcoin mining revenue. No contracted HPC leases. Lottery-ticket quant profile. |
| MU | Semis | AVOID * | HBM4 excluded from Vera Rubin (confirmed March 9, 2026). Hidden flag from prior synthesis. |
| TSM | Semis | AVOID * | CRITICAL geopolitical flag (Taiwan Strait). Business excellent (90.0 composite) but rules = AVOID. |
| WGMI | Miners ETF | N/A | ETF basket -- not a single company, cannot screen. |
| SOLS | Wrong sector | EXCLUDED | Solstice Advanced Materials is a specialty chemicals company, not AI infrastructure. |
* Narrow/conditional verdict -- see ticker details above.
Pass Summary: Technical Watchlist
| Ticker | Sub-sector | Composite | Watchlist Tier | Notes |
|---|---|---|---|---|
| NVDA | Semis | 92.5 | Tier 1 | Dominant AI GPU. -14% from peak. Ideal quant entry zone. |
| AVGO | Semis | 85.0 | Tier 1 | Custom AI chips + backlog. -19.8% DD. Best quant score. |
| APH | Networking | 85.0 | Tier 1 | Connector duopoly. -18.2% DD. Lowest vol of quality names. |
| FIX | Power/Infra | 85.0 | Tier 1 | Data center HVAC. Pristine BS. Near highs -- wait for pullback. |
| VRT | Power/Infra | 85.0 | Tier 1 | Liquid cooling leader. At ATH -- wait for 15-20% pullback. |
| PWR | Power/Infra | 82.5 | Tier 1 | AI grid builder. Best bear Sharpe. Near highs -- wait. |
| VST | Power/Infra | 70.0 | Tier 1 | Nuclear + AI PPA. -24.8% DD. In ideal quant entry zone now. |
| LITE | Semis | 72.5 | Tier 1 | NVDA $2B strategic investment. -18.2% DD. Contracted demand. |
| ALAB | Semis | 75.0 | Tier 2 (conditional) | Real profits, high growth. 90x P/E. 2yr history. Size small. |
| COHR | Semis | 65.0 | Tier 2 | AI optical networking. Debt declining. -16% DD. |
| WDC | Semis | 65.0 | Tier 2 | HDD duopoly for AI storage. Momentum +55% -- wait for pullback. |
| JBL | Networking | 65.0 | Tier 2 | AI EMS. Thin margins. Passes bar but lower conviction. |
| GLW | Networking | 67.5 | Tier 2 (conditional) | Fiber/AI. 86x trailing P/E. Needs pullback before entry. |
| LEU | Power/Infra | 60.0 | Tier 3 (conditional) | Nuclear HALEU monopoly. -55% DD. Idiosyncratic. Small size only. |
Alignment with Quant Research
| Quant Finding | Fundamental Confirmation |
|---|---|
| AVGO #1 pullback candidate (Sharpe 1.02, bear +0.23, -20% DD) | Confirmed: $73B backlog, 106% AI chip growth, real FCF. In ideal entry zone. |
| APH #2 candidate (Sharpe 0.95, lowest vol, -18% DD) | Confirmed: connector duopoly, 52% revenue growth, $4.4B FCF. |
| VST #3 candidate (Sharpe 0.88, bear-positive, -25% DD) | Confirmed: $5.9B EBITDA, nuclear PPA, Q4 miss was not a fundamental break. |
| FIX/PWR near highs -- quality but no entry discount | Confirmed: pristine fundamentals but wait for pullback. Same conclusion. |
| VRT at ATH -- wait for pullback | Confirmed: best revenue growth in screen but priced for perfection. |
| All 5 miners: quant AVOID (extreme vol, terrible bear Sharpe) | Confirmed: CIFR/IREN/WULF/CLSK are fundamental AVOID. WGMI is an ETF. |
| BE: -92.5% max DD, -0.95 bear Sharpe | Confirmed: AVOID. Extreme valuation, negative EBIT coverage, insider selling. |
| TSM near highs, quant PASS | Fundamental AVOID due to CRITICAL geo flag. Quant/fundamental split is expected -- geo risk is not captured in price data. |
| MU strong momentum (+53% excess) | Fundamental AVOID due to HBM4 exclusion. Momentum is NAND-driven, not HBM4. |
| LITE strong momentum (+80% excess) | Confirmed: NVDA $2B investment is the fundamental driver. Momentum is real. |
| ALAB: -51% DD, 2-year history | Confirmed: PASS (conditional). Real profitable business but extreme valuation and short history. |
| GLW: Sharpe 0.73, -23% DD | Fundamental PASS (conditional). Valuation concern from prior synthesis confirmed. |
| SNDK: excluded from user screen | Not in screen. For reference: CAUTION per synthesis. Memory cycle peak risk. |
The quant and fundamental analyses are in strong agreement. The primary divergence is TSM (quant would PASS; fundamental AVOID) and MU (quant momentum-positive; fundamental AVOID) -- both divergences are driven by non-price factors that quantitative analysis cannot capture.
Data Source Log
| Source | Query / Endpoint | Status | Notes |
|---|---|---|---|
| Existing synthesis | NVDA-synthesis.md |
OK (2026-02-25, 12 days) | Within 30-day cache; Q4 FY2026 results updated via WebSearch |
| Existing synthesis | TSM-synthesis.md |
OK (2026-02-27, 10 days) | Within 30-day cache; AVOID verdict unchanged |
| Existing synthesis | MU-synthesis.md |
OK (2026-02-25, 12 days) | Within 30-day cache; HBM4 flag confirmed via WebSearch |
| Existing synthesis | GLW-synthesis.md |
OK (2026-02-25, 12 days) | Within 30-day cache; CAUTION maintained |
| Existing synthesis | BE-synthesis.md |
OK (2026-02-25, 12 days) | Within 30-day cache; CAUTION/AVOID maintained |
| Existing synthesis | WULF-synthesis.md |
OK (2026-02-25, 12 days) | Q4 2025 results updated via WebSearch (Feb 26 earnings) |
| Existing synthesis | IREN-synthesis.md |
OK (2026-02-25, 12 days) | Within 30-day cache; AVOID maintained |
| WebSearch | "MU Micron HBM4 NVDA Rubin March 2026" | OK | Samsung + SK Hynix exclusive Vera Rubin HBM4 confirmed |
| WebSearch | "AVGO Broadcom Q1 2026 earnings revenue valuation" | OK | $19.3B Q1 revenue, 106% AI chip growth confirmed |
| WebSearch | "APH Amphenol FY2025 earnings revenue valuation" | OK | $23.1B revenue, 52% growth, Q1 2026 guidance confirmed |
| WebSearch | "VST Vistra Q4 2025 2026 guidance Cogentrix nuclear PPA" | OK | $5.9B EBITDA, $6.8-7.6B 2026 guidance confirmed |
| WebSearch | "PWR Quanta Services 2025 2026 revenue guidance backlog" | OK | $28.5B 2025, $33.5B 2026 guidance, $30B backlog confirmed |
| WebSearch | "FIX Comfort Systems 2025 earnings data center HVAC" | OK | $9.1B 2025, $10.95B 2026 est, $11.94B backlog confirmed |
| WebSearch | "VRT Vertiv 2025 2026 earnings guidance backlog" | OK | $10.23B 2025, $13.5B 2026 guide, $15B backlog confirmed |
| WebSearch | "JBL Jabil 2025 2026 AI revenue guidance" | OK | $12.1B AI revenue, $32.4B FY2026 guidance confirmed |
| WebSearch | "ALAB Astera Labs 2025 2026 revenue earnings valuation" | OK | $852.5M FY2025, +115%, $1.18B FY2026 projected confirmed |
| WebSearch | "LEU Centrus 2025 2026 DOE HALEU task order" | OK | $448.7M 2025, $900M DOE task order confirmed |
| WebSearch | "COHR Coherent 2025 2026 AI photonics revenue debt" | OK | $1.69B Q2 FY26, 1.7x leverage, debt reduction confirmed |
| WebSearch | "LITE Lumentum 2025 2026 NVDA investment AI optical" | OK | $2B NVDA investment + purchase commitment confirmed (March 2, 2026) |
| WebSearch | "WDC Western Digital 2025 2026 HDD cloud storage AI" | OK | $9.52B FY2025, $3.02B Q2 FY2026, 89% cloud revenue confirmed |
| WebSearch | "CLSK CleanSpark CIFR Cipher Mining 2025 2026 HPC balance sheet" | OK | CLSK $181M Q1 rev, $1.3B working capital; CIFR liabilities +50% confirmed |
| WebSearch | "WULF TeraWulf 2025 results HPC transition" | OK | -$661M net loss, $12.8B contracted HPC, $3.3B cash confirmed |
| WebSearch | "SOLS Solaris Energy NASDAQ stock" | OK | SOLS = Solstice Advanced Materials; not AI infra; correct sector = materials |
| python3 -c | Composite score computation (21 tickers, 4 dimensions) | OK | All scores verified via Python |
Data quality notes: - AVGO: VMware debt load is the primary balance sheet uncertainty; full leverage ratio requires latest 10-Q - JBL FY2025 GAAP net income decline (52.7%) is non-recurring; core EPS trend is more representative - LEU: $900M DOE task order is "subject to negotiations" -- not yet a binding contract; treat as unconfirmed upside - ALAB: Only 2 years of standalone public history; all statistical and fundamental conclusions carry elevated uncertainty - WULF: Q4 2025 results from February 26, 2026 -- one day after prior synthesis. Balance sheet figures in this report are current