Cross-Name Scenario Governance

Research discipline

Consistent scenario definitions and underwriting discipline, with company-specific probabilities based on evidence and risk. The evidence hierarchy and downside discipline are standardized; identical probabilities and identical economics are not required.

ARR/MW, ACV/MW, NOI/MW, utilization, capex/MW, financing and customer economics are company-specific and evidence-driven. Process is standardized; economic assumptions are not.

Primary filings and contractual terms take priority over company presentations, management commentary and third-party context. DR estimates remain separately labeled. New initiations use a Base-case target and the published ±15% rating convention, subject to an explicit analyst review of downside. Historical report exceptions remain identified. Live prices and interactive diagnostics do not automatically change a published rating. See the rating and target methodology.

Current scenario priors

Failure represents capital loss or a broken investment case; Bear represents material disappointment; Base is the central research case; Bull represents a strong but plausible outcome. Scenario weights are analyst priors, not empirically calibrated frequencies. Scenario weights are separate from simulated outcome probabilities. Review the company Coverage page for scenario mechanisms and weighted-value contributions.

Published rating and target methodology

Only a published report assigns a formal recommendation. The standard convention for new initiation reports is a Base-case target over 12–18 months, with probability-weighted value and the simulation median shown separately. Above +15% target upside is BUY, below −15% is SELL, otherwise HOLD, measured against the dated report reference. Upside uses the published rounded target, not the unrounded Base value.

Targets round to the nearest $5 at $50 or above and the nearest $1 below $50. A departure from this convention requires an explicit explanation in the report. Existing reports retain their original conventions: IREN uses probability-weighted value; IREN and NBIS retain their report-specific recommendation judgments. These exceptions are historical, not changes to the current model.

A threshold is a starting convention, not a substitute for downside review. Every new recommendation must discuss scenario losses, financing, dilution, evidence quality and upside-tail dependence. If PW or the simulation median is below the dated reference, a BUY requires a written investment justification. Model updates do not automatically change a published rating; a formal recommendation change requires a dated research update.

PW = sum of scenario weight × scenario value. Weights and Monte Carlo distributions are analyst judgments, not empirically calibrated probabilities. Neither PW nor the simulation median should be described as a verified expected market return.

Modeling rules

• IREN retains its frozen new-cohort pricing curve. NBIS retains its separately underwritten cohorts. CRWV applies the common future-sales curve to an estimated 50% of 2027 additions; the rest retains prior pricing. Its primary valuation is replacement-aware DCF with an ARR cross-check. Announced deal economics are unchanged. Commercial MW, active MW and critical IT MW are not interchangeable denominators.
• Power readiness affects delivery timing, financing capacity and residual infrastructure value. It does not create an automatic ARR/MW premium.
• Internal/customer cash and modeled debt capacity precede residual common equity. Future financing is an estimate, not a commitment.
• Project NAV incorporates the stated completion capital and project debt. New funding sensitivities must include proceeds as well as shares.
• Existing security claims and future ungranted SBC reserves are separate. Convert principal and full conversion shares are never charged together.
• Operating cloud simulations and aggregate landlord NAV simulations have different granularity. Their distributions and limits are disclosed in each Risk & MC tab.
• MU separates HBM from other DRAM and NAND, models shipment volume and pricing independently, and normalizes cash returns after manufacturing reinvestment. Customer deposits are not incremental equity value.
• COHR models high-speed optics within Datacenter & Communications. Delivered products, manufacturing costs, reinvestment and funding determine cash returns. Terminal EV/EBIT is scenario-specific; perpetuity FCFF remains a separate cross-check. Partnerships receive no automatic valuation premium.
• NVDA separates Data Center customer groups from Edge Computing. Procurement commitments affect inventory and cash; investment gains are excluded from recurring EBIT. Partner guarantees receive explicit scenario loss estimates, not automatic full-debt treatment.
• AMD values delivered Instinct and EPYC products after SBC and reinvestment. Conditional customer warrants enter dilution once; redacted milestone schedules use disclosed estimates. Long-term targets are not guaranteed orders.
• AEHR separates equipment, contactors and service sales. Bookings require delivery and acceptance; customer forecasts are not orders. The fiscal transition and elapsed-quarter cash bridge precede valuation. Existing awards, future compensation and completed equity funding are counted once.

Company-specific valuation

Company source registers distinguish reported information, management commentary and DR estimates. Recovered inputs and unfinished primary-source verification are disclosed; no universal readiness percentages are imposed on the reconstructed models.