POWL · Powell Industries, Inc. · Stock research

Current Thesis

The gap between Powell's order book and its income statement — the whole reason this name was ever interesting — did not close on 2026-08-03. It widened to the largest it has ever been. Fiscal Q3 (quarter ended 2026-06-30) booked $934M of new orders, +158% YoY versus $362M, for a 3.0x book-to-bill and a record $2.4B backlog, +69% YoY and +35% sequentially off the $1.8B struck at 3/31/26. Three orders above $50M landed inside the quarter: the previously flagged behind-the-meter data-center award above $400M, a $75M petrochemical order and a $60M LNG order. The reported P&L did what it has done for four quarters — lagged. Revenue of $311.7M (+9% YoY vs $286.3M, +5% sequentially vs $296.6M) came in under the $315.168M consensus, and diluted EPS of $1.42 (vs $1.32 a year earlier) missed the $1.47 estimate. Shares fell 13.4% in the 2026-08-04 pre-market session, the second consecutive quarter where the headline line items were the thing that got traded. The 2026-08-14 close of $213.48 sits 33.7% under the $321.94 52-week high, and the $220 weekly level named in the prior note on this name gave way. The fundamental condition paired with it did the opposite of breaking: backlog is up 35% in a quarter and book-to-bill is 3.0x, not below 1x. That divergence is the read, and there is no dated event to resolve it until the fiscal-year print in the December quarter.

Bullish and bearish views on Powell Industries, Inc.

The model's bull view on Powell Industries, Inc. (POWL), in brief: New orders $934M in fiscal Q3 (reported 2026-08-03), +158% YoY versus $362M, book-to-bill 3.0x — the largest single order quarter in the company's 79-year history. The bear view: Two consecutive quarters of missing the printed estimate: Q2 EPS $1.25 versus ~$1.34 (2026-05-04), Q3 EPS $1.42 versus $1.47 with revenue $311.7M versus $315.168M (2026-08-03). Both cases follow in full.

Bull Case

• New orders $934M in fiscal Q3 (reported 2026-08-03), +158% YoY versus $362M, book-to-bill 3.0x — the largest single order quarter in the company's 79-year history.
• Backlog $2.4B at 6/30/26, +69% YoY and +35% sequentially from $1.8B at 3/31/26; first time above $2B.
• Three mega orders inside the quarter: data center above $400M (behind-the-meter on-site generation), petrochemical $75M, LNG $60M — order concentration is spread across three end markets rather than one.
• Gross margin 30.6% on gross profit of $95.3M (+8% YoY); net income $52.2M versus $48.2M; the miss was on the top line and estimate level, not on unit economics collapsing.
• Cash and short-term investments $633.6M with working capital of $606.5M at 6/30/26 — the capacity build is self-funded off the balance sheet.
• Cantor Fitzgerald, cutting its target on 2026-08-14, stated that executing the $2.4B backlog "does not depend on the next major capacity addition": existing leased space including a 50,000 sq ft Houston facility and added Ohio capacity covers it, and the 275,000–300,000 sq ft facility under development is for growth beyond the current book. That removes the capacity-constraint objection that dominated the May discussion.
• Jacintoport fabrication-yard expansion is expected complete by the close of fiscal 2026 (year-end 2026-09-30), with production ramp thereafter.

Bear Case

• Two consecutive quarters of missing the printed estimate: Q2 EPS $1.25 versus ~$1.34 (2026-05-04), Q3 EPS $1.42 versus $1.47 with revenue $311.7M versus $315.168M (2026-08-03). Record bookings have not stopped the sell response — the 2026-08-04 pre-market decline was 13.4% on the best order quarter the company has ever reported.
• Revenue converted at +9% YoY while orders grew 158%. Backlog that large is only worth the multiple if it turns; every quarter of high-single-digit revenue growth against a 3.0x book-to-bill lengthens the conversion runway the market is being asked to underwrite.
• Petrochemical revenue fell 49% YoY in the quarter even as commercial and other industrial rose 54% and electric utility rose 18% — the mix shift toward data-center and utility work is not additive, it is partly substitution.
• Cantor Fitzgerald cut its target from $320 to $235 on 2026-08-14 while holding Neutral, and cited the stock at $211.20 on a 40.13 P/E. The pre-print target cluster carried in the prior note — JPMorgan $360, Roth $333, Street high near $370, an outlier $650 from Texas Capital, all as of late July — has not been re-marked in public coverage since the print; treat those as stale until refreshed.
• Structure: lower highs into lower lows since the May peak ($264.86 on 7/1 → $242.67 on 7/5 → $236.58 on 7/9 → ~$232.09 on 7/25 → $213.48 on 8/14). The ~$255 breakout shelf and the ~$226 late-July shelf are both gone, with no reclaim attempt.

Setup & Price Structure

Reference close 2026-08-14: $213.48, -33.7% from the $321.94 52-week high, -27.0% over three months, RSI(14) 47.2. RSI in the mid-40s after a 34% drawdown says the decline has been orderly rather than capitulatory — there is no oversold washout to mean-revert from, and no higher-low pivot to anchor a base.

The narrative is saturated, downgraded from maturing. What dates it: the theme cooled from accelerating to maturing in early June 2026; the AI-data-center tag came out of the active theme set on 2026-07-14; and on 2026-08-04 the market met the largest order quarter in company history with a 13.4% pre-market decline. A story that no longer pays for record fundamentals has a thin marginal bid, whatever the backlog says. The narrative has not failed — $2.4B of contracted work is not a failed narrative — but the equity has stopped rewarding it.

Crowding and positioning observables, stated as observables: the multiple is still 40.13x per Cantor's 2026-08-14 note at $211.20, so the de-rating has run through price rather than earnings; the first post-print target cut arrived 11 days after the print and landed at $235, roughly 10% above the 2026-08-14 close, meaning even the lowered Street mark now sits above spot; the 3-for-1 forward split effective 2026-04-02 broadened the retail-accessible share count into the top of the move; coverage is thin at four covering firms as of 2026-07-06, so single revisions move consensus disproportionately. No insider transactions appear in the filing record reviewed for this refresh. There is no earnings date inside the next 30 days to force a repricing either way.

Catalyst Calendar (next 30 days)

• 2026-09-16 — dividend payment date.
• 2026-09-30 — fiscal 2026 year-end; the Jacintoport fabrication-yard expansion is targeted for completion "by the close of Fiscal 2026" per the Q3 release. Slippage past this date would push the capacity ramp into FY2027.
• ~2026-12-03 (est.) — Q4/FY2026 results. Date not announced as of 2026-08-16; fiscal Q4 ends 2026-09-30 and the report follows in the December quarter. This is the next event that resolves anything: whether the $2.4B backlog holds and whether revenue conversion accelerates off +9%.

The honest description of the window: it is empty. Nothing between now and roughly early December re-rates the name on fundamentals.

Elapsed catalysts

• 2026-08-19 — record date for the $0.09/share quarterly dividend declared 2026-08-03. Mechanical, not a thesis event. (passed 7d ago)

What Would Change Our Mind

The leading indicator is backlog, and the leading indicator improved — so the thesis breaks on conversion, not on demand. A Q4/FY2026 print showing backlog below the $2.4B record, or book-to-bill under 1.0x, ends the order-acceleration leg regardless of what price is doing; so would gross margin coming in below the 30.6% posted in Q3 while the new capacity ramps, since the bull case now rests on turning contracted work at unchanged unit economics. On the tape, a weekly close below $200 would end the post-print stabilization attempt and extend the lower-high sequence off the $321.94 May high — at that point the drawdown stops being a pullback inside an uptrend and the burden shifts entirely to the December print. Upward: a weekly reclaim of the ~$240 area with a higher low above the post-print zone would be the first structural evidence since May that the bid has returned; absent that, the name has no base.

Correlation Notes

• Trades with the electrical-equipment and grid complex — GEV, VRT, ETN, and small-cap STRL. The sector tape has dominated the single-name read through the entire drawdown; peer prints and peer guidance move POWL more than POWL's own bookings did on 2026-08-03.
• Data-center capex is now inside the backlog directly: the >$400M behind-the-meter order booked in fiscal Q3 ties a slug of the $2.4B to hyperscaler on-site generation plans. Any hyperscaler capex deferral reads straight through.
• Oil & gas and petrochemical exposure remains material and is currently the drag — petrochemical revenue -49% YoY in fiscal Q3 — so the name is not a clean AI-power proxy; it carries a hydrocarbon-capex leg that moves with LNG FIDs and refinery turnaround budgets.
• Small/mid-cap industrial beta: at a ~$8-9B market value with four covering firms, liquidity and revision flow are thinner than the mega-cap grid names, which amplifies both the May-to-August de-rating and any eventual re-rating.