[Guest Klothilde]

Gotta love your JASO IR! This is what I got only a few hours after emailing them:

Hi XXX (aka Klothilde),

Thanks for getting in touch. In answer to your question, module tolling accounted for 15% of total module shipments and 11% of total module revenue in Q4.

Thanks,

Nick

From this we get:

Module revenue: $167.3M

Module shipments: 273.7MW

Module ASP: $0.611/W

Module tolling revenue: $20.7M

Module tolling shipments: 48.3MW

Module tolling ASP: $0.428/W

Cell & Cell tolling revenue: $80.2M

Cell & Cell tolling shipments: 178.0MW

Cell & Cell tolling ASP: $0.450/W

Total revenue: $268.1M

Total shipments: 500.0 MW

Total ASP: $0.536/W

61 cts. module ASP is imho consistent with selling about half of their modules in China. 43 cts. module tolling ASP points to wafer - module tolling.

[explo 648]

Thanks. Yes, 61 cents makes sense. Now only the cogs is in the dark. Great module tolling ASP. Must be like you said, including both cell and module tolling.

[Guest Klothilde]

...Now only the cogs is in the dark...

Actually the mud is slowly clearing. My cost model aligns pretty good with their overall COGS. I did a costing based on spot poly in Q3 and their stated non-si cost of 53 cents. Assumptions for breakdown: poly cost: 0.103 (10% above multi poly cost assuming mostly mono cells - thus $17.4/kg * 5,4 g/W *1,1) wafer conversion: 0.16 (mostly mono) cell conversion: 0.17 module conversion: 0.20 Module cost per watt: 0.103+0.16+0.17+0.20 = $0.633 Module shipments: 273.7MW Module COGS: $173.3M Module tolling cost per watt: 0.17+0.20 = $0.37 Module tolling shipments: 48.3MW Module tolling COGS: $17.9M Cell & cell tolling cost per watt: 0.103+0.16+0.17 = 0.433 (assume only cells - no cell tolling) Cell & cell tolling shipments: 178.0MW Cell & cell tolling COGS: $77.1M Total COGS: $268.3M I assumed little to none cell tolling since historically the tolling numbers were pretty low and their high cell&cell tolling ASP (45 cents) also suggest little tolling volume. $268.3M is not too far away from their overall COGS of $280.5M. Higher cost poly and wafer inventory can help explain some of the difference.

[explo 648]

Nice. They are buying wafers too, so you need to model that. Did they confirm low poly cost?

[Guest Klothilde]

No, they didn't say anything about their poly. Though looking at the aggregate COGS of my model space for expensive poly is limited, approx to 13 cts. Yes they also buy wafers, this was a first shot to model the thing in-house. I think working in external wafers won't change results much, though, since wafers were trading at next to zero margin in q4.

[explo 648]

Thanks. The cell processing cost looks high. Probably why they don't want to discuss it. 16 cent mono wafer processing sounds low (I usually model 10 cents multi 20 cents mono), but it could be a strength they have.

[Guest Klothilde]

Sorry - maybe I wasn't clear. All they said was that non-si cost was 53 cts. per watt. The breakdown shown is based on my own assumptions, i.e. 16/17/20. 20 cents for mono wafer conversion like you suggest strikes me as too high. Do you have any source / link for mono wafer conversion costs relative to multi? I'm still trying to nail down cost per watt for mono. What I've gathered so far points to 10% higher poly cost, 20% higher wafer conversion cost, 5% higher cell conversion cost, and 10% lower module conversion cost, all relative to multi. 100% higher wafer conversion relative to multi is completely new to me.

[explo 648]

Mono and multi ingot growth are very different processes. Diff has been 10 cent and I think that remained duing the major cost improvement of multi ingot growth. Would be good to get some real numbers on this though.

[Guest Klothilde]

That I know, CZ vs. casting. However look at YGE, who mentioned that the total cost gap at module level between Panda and multi was 15 cents and that they wanted to reduce it to 10 cents. And this is n-type mono, which on top of the CZ wafering includes much higher cell conversion costs than multi. Have a look at page 19. This is outdated but useful for a "feelin": http://www.semi.org/ch/sites/semi.org/files/docs/1205_SolarSector_DAIWA.pdf

[explo 648]

the total cost gap at module level between Panda and multi was 15 cents

that translates to around 18 cents on wafer level.