STNG — Scorpio Tankers Inc.

Deep-value: STNG — Scorpio Tankers Inc.

Analysis: 09/08/2026 · Price: … · Market cap: … · NYSE · Single currency USD (reporting and price)

REFRESH regime. Revisits the analysis from 08/17/2026 (verdict then: NEUTRAL / MONITOR). Mechanical trigger: an unclassifiable 6-K from 09/03/2026. On reading, it turned out to be anything but a routine filing — the Q3 2026 rate update and three three-year time charters, i.e. exactly the forward price that tests the old thesis’s central assumption.

What I inherit, explicitly, without re-deriving: the business description and the absence of a moat; the Scorpio group’s related-party structure and the quantification of the rent; the management team’s profile and tenure; the 2022-2025 buyback history; the structural threats (a 20.7% orderbook, EU ETS/FuelEU, the energy transition); the FCF-bridge methodology.

What I re-derived, because it moved: the fleet (93 → 75 vessels); the revenue-days base; mid-cycle owner earnings; the capital structure after the convertibles; share count; the discount rate; all five valuation models; the Quality of Earnings and red-flags chapters, on the Q1 and Q2 2026 figures now read from the filings.

What I fixed from the old file. The 08/17/2026 analysis correctly complained that it lacked the Q1 and Q2 2026 financial statements (the last local 6-K was from 12/16/2025) and that three things remained unreconciled. I’ve now downloaded them directly from EDGAR (q12026earnings-exhibit991.htm from 05/05/2026, q22026earnings-exhibit991.htm from 07/30/2026, stngexhibit991-09032026.htm), together with the Q3 and Q4 2025 releases. The three unknowns are resolved — see the balance-sheet and red-flags chapters.

Correction note to the previous report. It wrote that booked Q3 2026 rates were “LR2 ~…/day at 34% of days, MR ~…-49,500 at 46%,” and acknowledged that third-party sources didn’t reconcile. Those figures were, in fact, the realized TCE in Q2 2026, not the booked rate for Q3. The primary filing from 07/30/2026 gives LR2 …/day at 34%, MR … at 46%, Handymax … at 38%. The difference isn’t cosmetic: on the correct figures, Q3 shows an orderly normalization, not a collapse.

Executive summary

The thesis, in three sentences. Scorpio Tankers owns and operates 75 refined petroleum product tankers (25 LR2, 36 MR, 14 Handymax, average age 10.2 years — the 09/03/2026 release) and earns, literally, (TCE/day − opex/day) × revenue days − G&A − dry-dock. In the last eight months management did something remarkable on the asset side: it sold 18 vessels (93 at 12/31/2025 → 75 today) collecting over … with … of accounting gain in H1, issued … of convertibles at a 1.75% coupon and sub… yield to maturity, redeemed the 7.5% Nordic bonds, and fully repaid five secured facilities — reaching … of cash against … of gross debt, with no maturity until 2031 (the Q2 2026 release, the debt table at 07/28/2026). The thesis’s problem remains unchanged: not balance-sheet quality, but position in the cycle — the fleet’s TCE was …/day in Q2 2026, 2.03x the five-year average, and the … price is 167% above the … low of 18 months ago.

What the new filing demonstrated, and why it matters more than the price move. On 09/03/2026 the company announced three three-year time charters: two LR2 (STI Gladiator, STI Jermyn) at … and …/day, one MR (STI Pontiac) at …/day. Added to those announced in July (STI Notting Hill and STI Westminster at 25,000, STI Bronx at 23,900, STI Guide extended at 33,000), these are market forward prices, paid by commercial counterparties for a horizon covering everything reasonably called “mid-cycle.” Weighted by the current fleet mix (…% LR2, 48.0% MR, 18.7% Handymax) they give …/day. The …/day assumption from the August analysis — built then from historical averages, without this confirmation — is validated by the freight market within 2.5%. This is the most important thing the refresh brings: the thesis’s main falsifier was tested against a market price and survived.

The second new fact, in the opposite direction: the fleet shrank 19%. The revenue-days base in the old analysis — … a year, calculated on ~85 vessels — no longer exists. The 09/03/2026 filing gives 6,669 expected revenue days for Q3 2026, i.e. 26,676 annualized; I use …, conservatively. Mid-cycle owner earnings fall from 430 to … The value hasn’t disappeared, it moved: the … collected from sales is in the … of cash, hence it reappears in net debt. The model has to make exactly this move, and does.

Estimated value. Five triangulated models, all starting from mid-cycle owner earnings, not current flow: range … – 145.7/share, median …, median safety margin …. Monte Carlo over 20,000 scenarios: median …, P10 at …, P90 at …, undervaluation probability …%, probability of a margin above 30% only …%. The P10-P90 range is 89 percentage points at a net-cash company — the width doesn’t come from leverage, it comes from a single variable: the daily rate.

Verdict: NEUTRAL / MONITOR — unchanged, but for a different balance of reasons. Three findings, each with its own figure.

• On mid-cycle multiples, the stock is more expensive than at the previous analysis, not cheaper. At … EV (market cap minus the reported … of net cash) is … = 7.5x mid-cycle owner earnings and 5.6x mid-cycle adjusted EBITDA (…). In August, on the old base, it was 6.2x and 4.7x. The price rose 5.5% while the earnings base fell 13% — both push the same way.
• The balance sheet improved in a way worth paying for, and I pay for it. The average cost of debt fell from 6.5% (… interest paid in 2025 on ~… average debt) to 2.0% (… at 1.75% plus … at SOFR…) — a savings of ~… USD/year, i.e. …/share. There’s no more secured debt, no more leasing, no maturity until April 2031, and the convertibles’ conversion price is … 22% above today’s quote. I lowered the discount rate from … to …% exclusively for the financial-risk component; business risk is untouched.
• Capital allocation split into two opposite behaviors, and both must be said. On assets, management behaves like a textbook Outsider: sells 18 vessels at the secondhand market peak at ~50% above book value, issues equity-linked paper 22% above market at sub… yield, retires 7.5% debt, locks in freight for three years at good historical rates. On stock, it behaves exactly the opposite: zero buybacks in 2025, with the stock between … and … and … available in the program; then … at an average price of … in Q2 2026, near an all-time high. The pro-cyclical pattern identified in August isn’t just confirmed — it gets an extra year of evidence.

What would change the verdict to buy. A price below ~… (P/B ~0.90x on the updated book value of …/share), where EPV offers real margin and the bear scenario becomes bearable; or proof, over two-three quarters of new time charters, that the …/day LR2 level isn’t a one-off fixing peak but a plateau — which would move mid-cycle owner earnings toward …-500 mil and value above … The first is a matter of patience; the second is checkable at the Q3 reporting on 10/29/2026.