Looking at the raw quarterly cadence first: revenue went $849M → $861M → $986M → $1.01B → $1.04B → $1.13B → $1.27B → $1.34B. That's not "decelerating" in any meaningful sense — sequential growth accelerated in H2 2025 (12% Q/Q in Q3, 5.5% in Q4), and YoY growth expanded from 14% in Q2'24 to 33% in Q4'25. The "decelerating quarterly trend" flag from the revenue confidence module is wrong, or at best mis-specified. Net income ex the Q4'24 $332M DTA release (which was a one-time valuation allowance reversal — that's what a 33% margin print in an otherwise ~10% margin business signals) shows margins expanding cleanly: 6.9% → 8.6% → 11% → 13%. That is a genuine operating inflection, not an accounting artifact.

The synthesis model's $12-14 floor and $18-22 platform-case is directionally reasonable but I think it under-weights the operating leverage now visible. 2025 delivered $4.77B revenue with $1.68B operating income — a 35.2% operating margin on a GAAP basis versus $233M (6.3%) in 2024. That's a 7x op income jump on 29% revenue growth. If you believe even half of that margin structure persists, 2026 at $6.1B revenue and 25% op margins gets you ~$1.5B op income, which on $24B market cap is not egregious. The reverse-DCF test: at 6.3x P/S, the market needs ~20% revenue CAGR for 5 years and terminal 20% FCF margins. With revenue CAGR already 28% and accelerating, the bar is being cleared, not stretched.

The contrarian case — and it's real — is that SoFi isn't a software platform, it's a bank that reports GAAP net income while burning $3.98B of FCF. Operating cash flow of negative $3.74B means the "earnings" are an accrual artifact of loan originations held at fair value; the cash reality is that SoFi funds loans with deposits and securitizations, and spread compression or a credit cycle would expose this immediately. ROE of 5.87% and ROA of 0.99% are bank-like, not fintech-like, and a bank earning sub-6% ROE does not deserve 2.87x book. If you normalize to a 10-11% ROE mature-state bank at 1.3-1.5x book, fair value is closer to $10-11, not $18. Galileo is the swing factor, and we don't have clean segment disclosure in this file to size it.

The insider activity tag of "net insider buying" is misleading — what I see is award grants plus M-Exempt exercises followed by F-InKind withholdings and sales (9,742 and 94,958 share sales on March 18-20, 2026). That's routine option-exercise-and-sell behavior, not conviction buying. I'd call it neutral-to-slightly-negative, not a bullish signal. Also worth noting: the file has 2026-dated transactions, which means the data extends past today or the dating convention is off — either way, treat the insider signal as noise.

Where I land: the models are collectively too cautious on the trajectory and too generous on the quality. The revenue and margin inflection in 2025 is real and the synthesis "high conviction required" verdict is defensible but timid — at $18.6 with this Q4 print, you're paying ~15x forward EBITDA for a business compounding revenue near 30% with visible operating leverage. The genuine risk is not valuation, it's that this is still a credit-sensitive balance sheet dressed as a platform, and the -$3.98B FCF means a funding-market disruption is an existential variable, not a rounding error. I'd push back on the synthesis $12-14 downside — that's too low given 2025 GAAP op income of $1.68B — and call fair value $17-22 with significant right-tail optionality if Galileo inflects. Current price sits mid-range of fair, not cheap, not expensive.