kWh Analytics 2024 Solar Risk Assessment pegged the median P90-to-P50 production ratio at roughly 90%, meaning lender base cases sit about 10% above the 90th-percentile downside. That one number reframes solar P50 P90 energy yield underwriting from a mechanical spreadsheet step into the largest single lever on debt sizing. Credit committees that treat P90 as an afterthought tend to oversize senior tranches on residential solar and utility-scale portfolios alike. This guide walks the assumptions bankable independent engineers use, the databases behind them, and where credit analysts should push back.
P50, P90, and P99 in solar P50 P90 energy yield underwriting
P50 is the median expected annual production: half of years fall above it, half below. P90 is the exceedance level annual output meets or beats in 90% of years. For solar P50 P90 energy yield underwriting, senior lenders size debt to P90 or lower because coupon payments must clear a bad-resource year, not the average.
Inter-annual variability drives the spread. P99 sits at the far downside tail, reserved for stress runs and cash sweep triggers rather than primary sizing. Utility-scale sites in the desert Southwest show tighter P90/P50 ratios than variable-climate sites in the Northeast or Pacific Northwest. As pv magazine USA reporting on kWh Analytics 2024 Solar Risk Assessment summarized, the median P90/P50 ratio for U.S. utility-scale projects sits near 90%, though outliers reach 85% on the low end or 95% on the tight end depending on geography and dataset length. CAISO and ERCOT desert Southwest sites typically post P90/P50 ratios of 92 to 95%, while Northeast and Pacific Northwest projects with higher inter-annual cloud variability often land at 85 to 88%. Equity investors watch P50 because it drives base-case IRR. Debt investors watch P90 because it drives coverage tests.
Irradiance databases and IE firms behind solar P50 P90 energy yield underwriting
Almost every bankable U.S. solar energy model starts with irradiance data from NREL National Solar Radiation Database (NSRDB), which provides over 30 years of hourly resource data at 4 km resolution. In solar P50 P90 energy yield underwriting, lenders require at least one satellite-derived source paired against local ground-truth pyranometer data.
Commercial supplements include SolarAnywhere, Meteonorm, and Solargis. On the IE side, the short list of bank-accepted firms in 2026 for U.S. utility-scale projects includes DNV, Black and Veatch, Leidos, Enertis, ArcVera, Fichtner, and E3. For residential solar portfolios, third-party model validators such as kWh Analytics and PVEL back ABS transactions rated by Fitch, KBRA, and Moody's. Any lender syndicate should cross-check the IE against the DOE Solar Energy Technologies Office reference materials on solar resource assessment. Beyond the roster, ask which time-series length the P90 was computed against. A 5-year dataset gives a tighter P90 than a 30-year dataset because rare downside years get sampled out. NREL recommends at least 15 years, and the NSRDB provides more than 30.
DSCR calibration for solar P50 P90 energy yield underwriting scenarios
Most 2026 term sheets for contracted utility-scale solar require a P50 base-case minimum DSCR of 1.30 to 1.35x, alongside a P90 one-year minimum near 1.10x. Merchant-exposed deals push higher: 1.40x P50, 1.20x P90, plus a P99 no-default check. In solar P50 P90 energy yield underwriting, the P90 test is often the binding constraint on senior tranche size.
The mechanics are straightforward. Start with the IE P50 case, apply degradation, availability, and loss stacks, then rerun cash flow at the P90 exceedance level. The lower of (max debt at 1.30x P50) and (max debt at 1.10x P90) sets the sized senior tranche. In the 2026 term loan syndications we have reviewed at SunRaise Capital, the P90 case has been the binding sizing limit in the majority of contracted utility-scale deals we have underwritten. Rating agencies including S&P and Fitch, per SEC-filed prospectuses for recent utility-scale term loan B and ABS transactions, expect a P95 or P99 stress that still clears interest-only coverage. Contracted revenue under a PPA reduces the P90 shock because the offtaker pays for delivered kWh at a fixed price, but curtailment and volume risk still travel through the P90 case. Merchant deals warrant deeper stress, as our merchant solar project finance underwriting 2026 guide details.
| Deal type | P50 DSCR floor | P90 DSCR floor |
|---|---|---|
| Contracted utility-scale | 1.30 to 1.35x | 1.10x |
| Merchant utility-scale | 1.40x | 1.20x |
| Residential solar ABS pool | Portfolio P50 metric | Portfolio P90 metric |
For fuller mechanics, our utility-scale solar project finance debt sizing and DSCR 2026 breakdown walks a worked term-loan example end to end.
Degradation, soiling, and availability assumptions
In solar P50 P90 energy yield underwriting, bankable 2026 loss models converge on a narrow set of well-tested assumptions. Monocrystalline silicon modules degrade at 0.5% per year after a first-year drop of 2 to 2.5% for light-induced degradation, matching IEC 61215 accelerated test outputs and manufacturer linear performance warranties.
Bifacial and TOPCon technologies show similar or better degradation curves in NREL PV module reliability studies. Soiling losses run 1 to 3% in humid Northeast and Southeast climates and can reach 5 to 8% in the desert Southwest without regular washing. Snow losses add another 1 to 4% depending on tilt and latitude. Availability, meaning uptime of inverters, transformers, and grid interconnection, is modeled at 98 to 99.5% for standard O and M contracts, a point covered in our solar O&M underwriting equipment warranty coverage guide. Shading and mismatch losses are site-specific but typically stack another 2 to 4%. The gross-to-net performance ratio for a well-designed U.S. utility-scale system lands between 82 and 86% in most bankable models. Lenders scrutinizing solar P50 P90 energy yield underwriting outputs should ask the IE to break out each loss line rather than accept a single lumped derate.
How lenders validate and stress-test energy assessments
Independent engineer diligence is not a rubber stamp. In solar P50 P90 energy yield underwriting, the credit team should reprice the deal under several sensitivities: irradiance dataset swap, degradation at 0.7% rather than 0.5%, availability at 96% rather than 99%, and soiling at the high end of the regional range.
Any deal that fails DSCR under those bumps is under-margined at the sized tranche. Second, cross-check the IE stated P90/P50 ratio against industry medians. If a Southeast project reports a P90/P50 of 95% while the regional norm is 88 to 91%, ask why. Third, verify inter-annual variability was computed from the full data series, not truncated to the sunniest recent decade. Fourth, confirm inverter clipping was modeled at the DC/AC ratio actually being built. Deals often ship with a design DC/AC ratio of 1.35 and get built at 1.45, which cuts peak production. Industry pre-construction reviews flag DC/AC mismatch on roughly a third of assets before financing closes. Fifth, ask how the IE treated interconnection curtailment. Grid operators in ERCOT, MISO, and CAISO have curtailed a growing share of solar output in recent years. Any P90 that omits curtailment is not really a lender-usable P90. At SunRaise Capital, we flag any P90 that omits curtailment back to the IE before issuing a credit memo. The U.S. Energy Information Administration publishes hourly generation and curtailment data by ISO that can be cross-referenced. For residential solar portfolios in ABS pools, an aggregated fleet P90 is the standard credit metric because individual home-level noise averages out. This is where solar P50 P90 energy yield underwriting shifts from a single-project engineering exercise into a portfolio statistics exercise.
Frequently asked questions
What is the difference between P50 and P90 in a solar energy assessment?
P50 is the annual production level a solar project is expected to meet or exceed in 50% of years, so it is the median or base-case forecast. P90 is the level that will be met or exceeded in 90% of years, a downside case only 10% of years fall below. For solar P50 P90 energy yield underwriting, lenders anchor senior debt to P90 because principal and interest must clear a bad resource year. Equity investors track P50 because it drives base-case IRR. Per kWh Analytics 2024 Solar Risk Assessment coverage, the median gap for U.S. utility-scale is about 10%.
Which independent engineer firms are considered bankable for U.S. solar in 2026?
The short list accepted across most 2026 project finance syndicates includes DNV, Black and Veatch, Leidos, ArcVera, Enertis, Fichtner, and E3. For residential solar ABS transactions rated by Fitch, KBRA, and Moody's, third-party validators such as PVEL and kWh Analytics also produce accepted work product. Bank counsel confirms the acceptable list at term-sheet stage. Any IE outside this cluster typically triggers a request for a second opinion from a Tier 1 firm before financing closes, adding weeks to the timeline and cost to the sponsor. Cross-referencing with SEIA member directories can help identify firms with market track record.
What DSCR should lenders require against a P90 case?
Contracted utility-scale solar deals in 2026 typically size senior debt to a minimum 1.30 to 1.35x DSCR on P50 and 1.10x on P90 one-year. Merchant-exposed projects tighten to 1.40x P50 and 1.20x P90, sometimes with a P99 no-default overlay. Residential solar ABS deals underwrite fleet-level P90 with credit enhancement sized to absorb stress on both production and homeowner default. The lower of the two DSCR-driven sizings sets the maximum sustainable tranche. Any sponsor pushing back on 1.10x P90 as a floor should be asked to model a P95, which usually holds the line. See NREL PV cost and finance references for baseline benchmarks.
What are standard degradation and soiling assumptions in bankable models?
Bankable 2026 energy models use 0.5% annual degradation for monocrystalline silicon after a first-year 2 to 2.5% drop for light-induced degradation, matching IEC 61215 test standards and manufacturer 25-year linear warranties. Soiling ranges from 1 to 3% in humid climates to 5 to 8% in the arid Southwest before washing. Snow adds 1 to 4% in northern latitudes. Availability sits at 98 to 99.5% for standard O and M contracts. Total performance ratio for a well-designed U.S. utility-scale system typically lands between 82 and 86%. Any IE reporting a system-level derate below 14% total without explicit line-item support should be challenged during diligence.
How do lenders validate an independent engineer P90?
Cross-check the IE stated P90/P50 ratio against regional industry medians published in SEIA research reports and other IE benchmark reports. If a project claims a tighter downside than its geography supports, ask for dataset length and inter-annual variability calculation. Reprice the deal with a degradation bump to 0.7% per year and an availability haircut to 96% to check if DSCR still clears. Confirm the DC/AC ratio in the model matches construction drawings. For portfolio deals, aggregate fleet-level P90 rather than trusting single-asset numbers. Ask how curtailment risk was treated in each ISO the assets sit in.
Does solar P50 P90 energy yield underwriting apply to residential solar the same way?
The math is the same but application shifts. Individual home noise means single-address P90 modeling is not useful. Instead, lenders aggregate a pool of thousands of residential solar contracts and run pool-level P90 against portfolio cash flow. This approach powers residential solar ABS underwriting from 2025 and 2026 deals by sponsors including Sunnova, Sunrun, and Mosaic in transactions rated by Fitch and KBRA. Pool P90 tends to sit closer to P50 than single-asset P90 because uncorrelated site-level noise diversifies. Diversification does not remove correlated risk such as regulatory shifts or common component defects.