Canadian Solar Lost $77 Million in Q2 — and Revenue Rose Anyway

Canadian Solar reported a $77 million net loss for the second quarter — about 8% of its entire $940 million stock-market value, gone in twelve weeks. The headline arrived the way many of these do: revenue rose 12% from the first quarter to $1.2 billion, at the top of the company's own guidance, and the loss widened anyway, to $1.40 a share against the roughly $1.01 Wall Street had forecast for it. When revenue goes up and the loss gets bigger, something structural is usually happening, and for a stock already down more than 40% this year, the cause matters more than the ticker's reaction.

Three things explain the divergence, and each one changes what you should make of the stock.

The 'recovering margins' story was partly a refund landing.First-quarter gross margin printed a healthy 25.1%, but $93 million of it was a one-time refund of emergency tariffs the company had paid under U.S. trade orders — money being collected, not the business suddenly earning more. Strip that out and the second quarter's 13.9% gross margin is closer to the real run-rate, which is exactly the neighborhood management guides for the third quarter. Among investors watching Canadian Solar as a beaten-down turnaround, that distinction matters: the spring quarter looked like an inflection, and part of it was an accounting windfall that does not repeat.

The company is deliberately making itself smaller. Module shipments fell to 3.1 gigawatts in the quarter, down 60% from a year earlier. That is not purely collapsing demand. Under CEO Colin Parkin, who took the job in May as founder Shawn Qu moved to executive chairman, the stated strategy is "profit first" — ship fewer panels at better prices rather than chase market share in a global glut where everyone loses money on volume. The growth counterweight is storage, and it is real: 3.7 gigawatt-hours shipped in Q2, up 73% year over year and ahead of guidance, against a contracted backlog of $3.5 billion. That part of the business is executing.

The money is going into U.S. factories that earn nothing yet. The new solar-cell plant in Jeffersonville, Indiana opened its first phase in July, with plans to reach 6.3 gigawatts of U.S. cell capacity by the first half of 2027, and the Mesquite, Texas module plant is doubling to 10 gigawatts this year. These plants are built to capture the premium prices U.S.-made panels command under the tariff regime, which is the right idea if you believe the policy survives. But they cost cash now, at exactly the moment the income statement is thin.

That brings us to the cash test, which is where this story actually sits. The full consolidated loss was $85.8 million, with minority partners in the manufacturing arm absorbing part of it. Cash ended the quarter at $1.9 billion, but total debt rose to $7.1 billion from $6.8 billion three months earlier: about $4.1 billion at the Recurrent Energy project-development arm ($2.6 billion of it non-recourse project debt, serviced by the projects themselves rather than the parent), $2.5 billion of manufacturing debt, and $0.4 billion of convertible notes. Over the trailing four quarters the company spent roughly $1.3 billion on capital projects, and free cash flow came out near negative $1.5 billion.

The engine meant to relieve that pressure is the other half of Recurrent's job: selling completed or nearly completed projects to recycle cash and pay debt down. That engine barely fired in the first half. Project sales were deferred, and management now says they close in the second half — the plan being to do the year's heavy lifting in the next two quarters, into a market full of tariff and tax-credit uncertainty.

The price tells you the market believes none of the carried value yet. At about $940 million, the entire company — including the share of the manufacturing arm that minority owners hold — is valued at roughly one-fifth of the $4.3 billion of equity on its books, against $7.1 billion of debt. A money-losing manufacturer at that multiple is not the market being shy; it is the market pricing in real doubt: that the U.S. factories won't earn their cost, that Recurrent's carried project values shrink, or that the shares get diluted before value appears.

For an income investor, the discipline is the same one we apply to any payout: trace where the cash comes from, then test whether it can support what's being promised. Canadian Solar promises shareholders nothing today — no dividend, no buyback — which changes the math on "buying low." Buying a beaten-down price is only reinvestment when the income engine is intact; here the engine is mid-build and negative on cash. That doesn't make the stock wrong. It makes it a speculation on execution and policy rather than a position funded by what it pays you.

So watch the second half for two specific proofs. First, Recurrent actually closing the deferred project sales this quarter — that is the deleveraging test. Second, the U.S. factories converting protected prices into positive gross margin as shipments accelerate toward the reiterated full-year target of 6.5 to 7.0 gigawatts of U.S. modules. If both show up, the retained cash starts doing what retained cash is supposed to do: build durable per-share earning power. If the sales slip again and the plants keep swallowing cash, then a $7 billion debt load against negative operating cash flow is a credit story wearing a stock's clothes. Cheap at a fifth of book is only cheap if the cash engine is intact. That is not established yet — it is exactly what the rest of the year must prove.