CSIQ · Canadian Solar Inc. · Stock research

Current Thesis

The policy-relief leg that carried this name into May is still dead, and the month since the last note tested that read in the cleanest way available: on 2026-07-24 CS PowerTech formally opened Phase I of the Jeffersonville, Indiana cell plant — the first American PV cell facility built for heterojunction bifacial n-type cells, nearly $1B of local investment and >1,200 jobs at full build-out — and the stock closed 2026-08-07 at $15.91, still 52.6% below the $33.58 adjusted 52-week high. That is the reshoring narrative's first hard, dated proof point arriving to no bid. What sits between here and a re-rate is one event: the Q2 2026 print on 2026-08-27 (scheduled 2026-07-30, call 8:00 a.m. ET), guided on the 2026-05-14 Q1 call to 13–15% gross margin against 25.1% in Q1. Price has stopped falling — the 2026-07-08 low of $14.75 has not been retested and RSI(14) sat at 55.0 on 2026-08-07 — but a four-week range under a broken structure is a stabilization attempt, not a completed base.

Bullish and bearish views on Canadian Solar Inc.

The model's bull view on Canadian Solar Inc. (CSIQ), in brief: The pivot stopped being a slide and became a building. The bear view: The guided margin air-pocket has not yet printed. Both cases follow in full.

Bull Case

• The pivot stopped being a slide and became a building. Phase I of the Jeffersonville PV cell facility opened 2026-07-24 at River Ridge Commerce Center; the site is designed for more than 6 GWp of HJT bifacial n-type cells annually at full build-out, with phase-two work expected to begin before the end of 2026 (company release, 2026-07-24). The prior note carried this as "commercial operation targeted July 2026, unconfirmed" — it is now delivered and dated.
• Vertical integration is the 45X mechanism. Paired with the Mesquite, Texas module plant, US-made cells feeding US-made modules is the structure management framed on the 2026-05-14 Q1 call as the route back to normalized margin — the domestic-credit position First Solar has had and Canadian Solar has not.
• e-STORAGE is the non-commodity franchise. $3.5B contracted backlog and 34 GWh under long-term service agreements as of 2026-05-08; 20+ GWh cumulative shipped as of 2026-03-31. Storage economics do not track module ASPs.
• Volume guidance implies a sequential step-up. Q2 guide (2026-05-14): 3.1–3.3 GW modules and 2.8–3.2 GWh storage on $1.0–1.2B revenue, against 2.5 GW and 2.1 GWh recognized in Q1.
• Sell-side floor has not been cut into the weakness. Mizuho raised its target from $15 to $18 (Neutral) on 2026-06-15; aggregator-published consensus in early August 2026 remains Hold with average targets clustered in a wide $15–19 band and a full range spanning roughly $9 to $30 — dispersion that reflects an unresolved 2027 margin question rather than a directional call.

Bear Case

• The guided margin air-pocket has not yet printed. 13–15% Q2 gross margin versus 25.1% in Q1, and Q1 itself carried roughly 860bps of tariff-refund accrual. The 2026-08-27 release is the first hard mark on how deep the compression actually goes.
• Losses are current; the credit is 2027. Q1 2026: net loss $32M, -$0.71 per diluted share, a $29M FX hit, negative operating cash flow. Phase II trial production at Jeffersonville is guided to early 2027 (+4.2 GWp); the 45X margin payoff lands with it.
• The July 24 opening drew trade-press coverage and no price response. Syndicated coverage ran through the week of 2026-07-24 (Solar Power World, PR Newswire wire pickup); the 2026-08-07 close of $15.91 remains inside the post-July range. A milestone that does not move price is evidence about the size of the marginal bid.
• Capital need against a roughly $1B build. A $230M convertible bond funds the US expansion; the Jeffersonville full build-out is described as nearly $1B of local investment. Funding the gap is an open question the Q2 balance sheet will speak to.
• Two C-suite changes in eight weeks into the worst margin quarter. Colin Parkin became CEO effective 2026-05-14 (founder Shawn Qu to Executive Chairman & CTO); COO Dylan Marx stepped down 2026-07-06 to run subsidiary Recurrent Energy.
• Trade-case exposure is idiosyncratic. The Section 337 TOPCon case (337-TA-1494) remains active, and a China-weighted supply chain into US tariff policy is the specific risk First Solar's domestic footprint does not carry.
• Demand hangover is structural. The July 4, 2026 ITC safe-harbor construction-start deadline elapsed; safe-harbored projects must be placed in service by 2027-12-31 and the ITC fully expires 2030-12-31 under the OBBB. The pull-forward is behind the industry.

Setup & Price Structure

• Adjusted 52-week high $33.58 (May 2026 peak), -52.6% from it. Three-month return -20.7%. RSI(14) 55.0.
• Life-cycle label: DEAD. Dating it: the round trip from the May high to $14.75 on 2026-07-08; the mid-June bounce to $21.12 that failed to reclaim the declining 50-day; the July 4 sector binary elapsing with no relief; and the 2026-07-24 Jeffersonville opening producing no breakout. The narrative that drove the move failed and the structure is broken. A successor leg — US vertical integration capturing 45X — exists and now has physical proof, but it has not produced expanding participation or a price base, and its P&L arrives in 2027.
• What has changed since 2026-07-12: the $15.20 June low that broke in early July has been reclaimed, and the $14.75 print has held for four weeks. RSI at 55.0 is mid-range — no momentum extreme in either direction.
• Crowding observables, stated as observables: price is 52.6% under its 52-week high, so there is no extension-above-a-rising-average condition here; the pressure runs the other way. An earnings date sits 19 days out from 2026-08-08. A $230M convertible remains outstanding against a build-out described at nearly $1B. Canadian Solar files as a foreign private issuer, so there are no Section 16 Form 4s — insider-flow screens return nothing for this name by construction, and their silence carries no information. No evidence of retail-sentiment clustering appears in available coverage; the July 24 facility opening ran through trade press without a measurable bid.
• Overhead structure: the mid-June failure high at $21.12, and the high-$17s zone where the declining 50-day sat in mid-July, are the two reference levels a recovery attempt has to work through.

Catalyst Calendar (next 30 days)

• 2026-08-27 — Q2 2026 results and conference call, 8:00 a.m. ET (scheduled in a 2026-07-30 company release; conference ID 13762069). Resolves four things at once: whether gross margin landed inside the 13–15% guide, the first quantified Jeffersonville ramp and 45X commentary, Q3 and FY2026 shipment guidance as the post-ITC-deadline US demand read, and the e-STORAGE backlog mark against the $3.5B / 34 GWh figures from 2026-05-08.
• 2026-08-27 — Form 6-K filing of the Q2 release. Share count, convertible treatment and cash position update against the roughly $1B Jeffersonville build.

Elapsed catalysts

• No other confirmed company-dated event inside the 30-day window. Jeffersonville phase-two construction is guided only to "before the end of 2026" (2026-07-24 release) — any date inside this window is unconfirmed. (passed 16d ago)

What Would Change Our Mind

The structure that matters is the four-week shelf built off the 2026-07-08 low of $14.75, and the event that resolves it is the 2026-08-27 print. Losing the shelf ends the stabilization read: a weekly close below $14.50 re-opens the $9–10 multi-year floor and keeps the reshoring pivot un-ownable on price regardless of how the plant ramps. The constructive flip requires three things together — Q2 gross margin printing at or above the 13–15% guide, management quantifying the 2027 45X benefit per watt with the Jeffersonville ramp schedule attached, and a reclaim of the high-$17s zone that capped the June bounce. A print that hits the low end of the margin guide, trims FY26 shipments and defers 45X quantification would confirm the dead-narrative label for another quarter, and the 2026-08-27 date passing without a reclaim of that zone is itself the answer.

Correlation Notes

• Trades as a high-beta solar-policy vehicle alongside TAN, FSLR, ENPH and RUN, and inversely to the 10-year yield. Exposure taken here at these levels is largely a macro and policy expression; company-specific setup contributes less to the return path than the sector factor does.
• The idiosyncratic divergence from First Solar runs through supply chain: FSLR's domestic footprint is shielded by 45X and by AD/CVD outcomes that penalize imports, while CSIQ's China-weighted chain is exposed to the same actions. The Jeffersonville cell plant is the direct attempt to close that gap, and each phase completed narrows it.
• e-STORAGE gives partial correlation to the grid-storage and data-center power complex (FLNC, utility-scale battery demand), a driver that does not move with module ASPs.
• Recurrent Energy project monetizations make quarterly results lumpy in a way sector comparisons do not capture — a single asset sale can swing reported revenue and margin independent of shipment volume.