DK · Delek US Holdings, Inc. · Stock research

Current Thesis

The leg on offer is a ~302,000 bpd Gulf Coast/Permian refiner where the margin cycle and a company-specific cost program are delivering in the same quarter, while the share price sits below most of the targets the sell-side set in July. Q2 2026, reported 2026-08-05, was the confirmation: adjusted EPS $5.48 versus a $2.74 consensus (Benzinga), revenue $4.087B versus $3.611B, adjusted EBITDA approximately $639M, and logistics-segment adjusted EBITDA of $144M — the highest quarterly result in company history. Management raised the Enterprise Optimization Plan (EOP) run-rate target to at least $220M, against the roughly $100M figure the plan carried when it was introduced. Price did not confirm. The last completed daily close, 2026-08-07, was $58.46 — 13.5% under the $67.62 52-week high, RSI(14) at 37.6, with the three-month return still +27.4%. The June advance is intact and the marginal buyer did not appear for the best print of the cycle. That is the signature of a MATURING narrative: well known, still working on fundamentals, moderating flow.

Bullish and bearish views on Delek US Holdings, Inc.

The model's bull view on Delek US Holdings, Inc. (DK), in brief: Q2 2026 beat was not marginal. Adjusted EPS $5.48 vs $2.74 consensus, revenue $4.087B vs $3.611B estimate, reported 2026-08-05. Adjusted EBITDA approximately $639M; net income approximately $170M ($2.71/sh GAAP). Self-help leg re-based higher. EOP contributed roughly $60M to P&L… The bear view: The beat was sold. Between the 2026-08-05 print and the 2026-08-07 close at $58.46, the stock sat 13.5% below its $67.62 high with RSI(14) 37.6. A blowout quarter that fails to produce a new high says the incremental bid has thinned. The adjustment gap is large. Headline… Both cases follow in full.

Bull Case

• Q2 2026 beat was not marginal. Adjusted EPS $5.48 vs $2.74 consensus, revenue $4.087B vs $3.611B estimate, reported 2026-08-05. Adjusted EBITDA approximately $639M; net income approximately $170M ($2.71/sh GAAP).
• Self-help leg re-based higher. EOP contributed roughly $60M to P&L in Q2 2026 alone, and management moved the run-rate target to at least $220M from an initial ~$100M (Q2 2026 earnings call, 2026-08-05). That is a structural cash-flow input independent of crack spreads.
• Logistics is compounding. Delek Logistics (DKL) delivered record quarterly adjusted EBITDA of $144M on Permian crude, gas and water volumes, with DKL full-year 2026 EBITDA guided to $520–560M. DK's majority interest plus GP economics in a separately listed partnership anchors the sum-of-the-parts under the refining stub.
• Sell-side targets sit above spot. TD Cowen raised to $76 on 2026-07-21, after Goldman Sachs $73 (2026-07-17), JP Morgan Neutral $62 (2026-07-14) and Raymond James $70 (2026-07-13) — all above the 2026-08-07 close of $58.46.
• Capital return continued through the quarter. Approximately $20M of buybacks and $16M of dividends against $263M of operating cash flow in Q2 2026; quarterly dividend of $0.255/share declared 2026-07-23, record 2026-08-03, payable 2026-08-10.

Bear Case

• The beat was sold. Between the 2026-08-05 print and the 2026-08-07 close at $58.46, the stock sat 13.5% below its $67.62 high with RSI(14) 37.6. A blowout quarter that fails to produce a new high says the incremental bid has thinned.
• The adjustment gap is large. Headline adjusted EBITDA of approximately $639M includes a 50% RVO adjustment; RVO-adjusted EBITDA was $490M. GAAP net income was approximately $170M against $5.48 of adjusted EPS. Renewable Fuel Standard obligations were flagged on the 2026-08-05 call as a source of significant uncertainty and elevated cost.
• Sell-side band has caught price. MarketScreener, citing FactSet-polled analysts around the Q2 print, put the mean target at $60.58 versus the $58.46 close — a narrow gap after a $58-to-$76 target march between 2026-06-29 and 2026-07-21. The repricing that drove the June–July leg is largely done.
• Margin engine is mean-reverting. Q3 2026 guidance assumes throughput of 296,000–316,000 bpd with opex $220–230M, G&A $50–55M, D&A $110–120M and net interest $75–85M. Management flagged that the steep crude backwardation that helped Q2 is expected to ease in Q3.
• Working capital swings the cash line. A $138M working-capital outflow in Q2 2026 sat against $263M of operating cash flow, and buybacks of roughly $20M were modest relative to the quarter's earnings.

Setup & Price Structure

Close of $58.46 on 2026-08-07; 52-week high $67.62; drawdown 13.5%; three-month return +27.4%; RSI(14) 37.6. Price is below every July target action (JPM $62, Raymond James $70, Goldman $73, TD Cowen $76) and roughly at the FactSet-polled mean of $60.58.

Life-cycle: MATURING, dated by three events. The analyst cluster ran 2026-06-29 through 2026-07-21 and ended at the highest target of the cycle; the 52-week high was set in that window; the 2026-08-05 print beat by roughly 2x on EPS and the stock was 13.5% off the high two sessions later. Attention arrived, participation stopped expanding, and the fundamental news flow is now better than the tape. The June breakout shelf near $48 remains the structural pivot beneath price; the current RSI reading of 37.6 with a positive three-month return describes a pullback inside an uptrend, not yet a broken structure.

Crowding and positioning observables, stated without a verdict: four target raises inside 23 days; consensus mean target within roughly two dollars of spot; buybacks of approximately $20M in a quarter that produced $263M of operating cash flow; a dividend record date (2026-08-03) that has already passed; and no company-specific binary scheduled inside the next 30 days to force a re-rate in either direction.

Catalyst Calendar (next 30 days)

• 2026-08-10 — Quarterly dividend of $0.255/share payable (declared 2026-07-23, record 2026-08-03). Confirms the capital-return cadence; resolves nothing about margins.
• 2026-08-12, 2026-08-19, 2026-08-26, 2026-09-02 (Wednesdays) — EIA Weekly Petroleum Status Report. Gasoline and distillate inventory builds are the highest-frequency read on the 3:2:1 crack spread that sets DK's refining earnings.
• No scheduled company binary. The next dated company event that tests the thesis is the Q3 2026 print, historically early November (~2026-11-04, est.) — the first check on the raised ≥$220M EOP run-rate and on the 296,000–316,000 bpd Q3 throughput guide.

What Would Change Our Mind

The load-bearing piece has shifted from the crack-spread cycle to the self-help program: after 2026-08-05 the market is being asked to pay for a ≥$220M EOP run-rate and $520–560M of DKL EBITDA, not merely for wide margins. Evidence that would break that:

• EOP progress stalling — a Q3 2026 contribution materially below the ~$60M booked in Q2, or the ≥$220M run-rate target walked back on the next call.
• Q3 throughput printing below the 296,000–316,000 bpd guide, or opex above the $230M top end, which would point at unplanned downtime rather than mix.
• A widening gap between headline adjusted EBITDA and RVO-adjusted EBITDA (Q2: approximately $639M vs $490M), or an adverse RFS/RVO policy outcome that raises the compliance cost run-rate.
• Price: a weekly close below $48 loses the June breakout shelf that the July analyst cluster was built on and ends the leg. Between here and there, failure to trade back above the $67.62 high on the next crack-spread expansion would date the theme as SATURATED rather than maturing.
• DKL unit weakness that undercuts the sum-of-the-parts, since a large share of DK's implied value moves with a security it does not control the price of.

Correlation Notes

• Refining complex. DK is a high-beta expression of the same 3:2:1 crack that drives VLO, MPC, PSX, PBF, CVI and DINO; on a margin-cycle drawdown it moves more, in both directions, given its size and the operating leverage on ~302,000 bpd.
• Crude structure, not just crude level. Management attributed part of Q2 to steep backwardation and expects it to ease in Q3 — a flattening curve compresses the inventory-cost benefit even if flat-price crude is unchanged.
• Midland–Cushing differential. The Permian sourcing advantage at Big Spring and the DKL gathering footprint both track that spread.
• DKL as a separate quote. DK's sum-of-the-parts and DKL's unit price are mechanically linked; DKL trades on its own distribution-yield logic and rate expectations, which can move DK's implied refining stub without any refining datapoint changing.
• RIN prices / RFS policy. RIN costs enter through the RVO line that the company adjusts out of headline EBITDA; policy moves here are idiosyncratic to the small-refiner cohort rather than to the broad energy tape.