How United Solar lost the thin-film race

With ECD's stock price nearing 5-year lows, it is time to revisit the reasons for the company's troubles. The previous analysis focused on Unisolar's competitiveness vs the polycrystalline module makers. But Unisolar's failure in the thin-film segment of the market is even more striking. In the latest reported (March 2009) quarter, First Solar, the thin-film leader, shipped 196MWs, a growth of 147% year over year. Unisolar shipped 20MWs, or 10x less, and a decline of 5% yoy. Quite a reversal from the situation in year 2005, when Unisolar shipped 11% more MWs than First Solar. The reason: the virtuous cycle of manufacturing costs and scale. A company with a manufacturing process that is inherently low cost can sell profitably in a commodity market and invest the profits (and/or raise money) to built more plants, benefiting from economies of scale, thus lowering the costs even further.

 

 

The chart above (data in this spreadsheet) shows the shipments by the two companies by quarter, as well as the respective manufacturing costs per Watt. The costs are estimated on a consistent basis for both companies as the sum of the cost of product sales (aka cost of goods sold) and the change in finished goods inventory divided by the production in megawatts. This is not a perfect estimate since it ignores certain small adjustments such as inventory write-offs, varying inventory cost layers, reclassification, etc, (and, in the case of Unisolar, some non-solar product in inventory) but as the overlaid dots representing reported costs by management (just 2 for Unisolar) show, it is a pretty good one. These costs include depreciation costs, which are non-cash, but excluding them does not change the picture much - in the latest quarter, those were about 12c out of the estimated $0.86 per Watt cost for First Solar, and 26c out of the $1.90 per Watt for Unisolar.

 

First Solar, by having a head start (they quickly dropped costs from $2.94 per Watt in 2004 to below $1.60 in 2005), has been able to underprice Unisolar severely and gain market share while lowering manufacturing costs per Watt 40-45% since 2005 (vs less than 30% drop for Unisolar's). As a result, Unisolar's March 2009 costs are still 20% above the level where First Solar's costs were in 2005. In the latest quarter, the cost differential widened, as Unisolar's cost increased while First Solar's dropped, mainly due to the drop in Unisolar production causing a hit to costs from overhead absorption. To make things worse, as illustrated on the chart below, 39% of Unisolar's March quarter production (12.9MWs out of 33.3MWs manufactured) was not shipped (vs just 11% of First Solar's), which resulted in skyrocketing inventory, production cut-backs in the June quarter, and thus, another hit from overhead absorption (the full extent of which will be revealed when the June results are reported).

 

 

And while in the past Unisolar's management could have argued that growth was somehow constrained by available capacity, the March quarter clearly illustrates that's no longer the case: Unisolar has plenty of excess capacity, however First Solar can sell at 50% gross margins and still force Unisolar to operate at a gross loss. Unisolar's so-called "roadmap to grid parity" (Slide 45 from their 2008 Investor Day presentation) is now suspect, as it relies predominantly on material cost reductions and throughput and yield improvements, rather than on conversion efficiency improvements to get from $2.00 per Watt to less than $1.10 by 2012. That cost target was nowhere near grid-parity anyway, as First Solar's 2012 target is less than 70c per Watt - see Slide 60 of their 2009 Analyst Day presentation. That same slide also shows that, in contrast, First Solar is heavily relying on conversion efficiency improvements to bring costs down, and so far they have shown they can do it (Slide 89), while at Unisolar, there has been virtually no progress on efficiency, despite earlier promises of 6%-7% increases in 2008 and 2009 (Slide 15 from the 2007 Shareholder Meeting presentation).

 

Now that First Solar is experiencing price pressures from the crystalline PV makers (the reason for those 23MWs of excess production in the March quarter), they may be lowering prices faster than the cost reductions (given their gross margins, they can afford it) and it won't be long before Unisolar's quarterly shipments of 20MWs or so start looking redundant to potential customers. The latest "Solar Spot Price Tracker” (see UBS' Global Solar Industry Update, July 8th, 2009) finds ASPs such as $1.60 per Watt for First Solar modules (10% panel-level efficiency), $2.00 for Chinese crystalline modules (about 14% panel-level efficiency), $1.50 for tandem a-Si (8% panel-level efficiency) and $1.30 for single junction a-Si (6% panel-level efficiency). Compare those to Unisolar's 6-7% panel-level efficiency and $1.90 per Watt estimated cost of manufacturing in the March 2009 quarter (likely going higher in the June, and the current, September, quarters). To make fair price comparisons, these rules of thumb may apply: 1) a 4-5 efficiency points disadvantage requires a 20-40c per Watt price advantage to compensate for the higher balance of systems costs from the lower efficiency, 2) Unisolar's glue-on product saves you a rack (10c-50c per Watt), and, 3) Unisolar's warranty is just 20 years (per their datasheets) vs 25-year industry standard, which could be a 50c-70c hit. Based on these shortcuts, Unisolar may have to lower ASPs to $1.30 per Watt or below to be competitive (resulting in a mindboggling negative 50% gross margin).

 

Unisolar's management has argued in the past that they offer a unique, differentiated product that is somewhat shielded from the pressures of a commodity market, and even deserves to sell at a premium per Watt, despite its low efficiency and substandard warranty. For example, on the March-quarter earnings call, the CEO argued that "flexibility, light weight, and ability to be integrated into roofing materials are significant advantages on rooftops," with further benefits such as "easy installation with no roof penetrations" and resistance to "wind uplift." He asserted that "Unisolar laminates are ideal for the rooftop and as a result are the leading product on building integrated PV systems." His conclusion is obviously contradicted in the real world, as both the CSI data and the Frost & Sullivan European BIPV study indicate that Unisolar is currently an insignificant, second-tier PV player on the rooftop, but let's now address the so-called unique features, which, as it turns out, appear to be nothing more than red herrings:

 

1. Flexibility: Glass panels can be easily installed on curved roofs, as the picture on the left illustrates (credit: SolarMax). And there aren't many roofs with surfaces that fluctuate wildly as to require flexible modules (and if flexibility is truly needed, that can be addressed on rack level)

 

2. Light weight: Virtually all roofs in the developed world that have been built up to code in the past 20-25 years can withstand the extra 3-5 psf dead load from regular glass solar panels on nonpenetrating racks (since those roofs are built with extra dead load factors, partially, for eventual reroofing). See the PV Inspector Guidelines [cached], page 14, and the module loads column here, keeping in mind that a lightweight rack could add less that 0.5 psf. The reality is that First Solar put about 90MWs in 2007 and about 200MWs in 2008 on rooftops (according to management, 40% of shipments still end up there), while the most Unisolar could have done is 48MWs in 2007 and 109MWs in 2008 (and that's assuming channel inventories have been cleared, which is a big if, as up to 30MWs may be sitting there).

 

3. Roofing integration and easy installation with no roof penetration: There is nothing unique about Unisolar's product as far as BIPV suitability and low balance-of-systems costs are concerned. The market-share leader in California, SunPower, offers T5 and T10 tiles for flat roofs, virtually all glass panels can be attached to a metal roof with clamps such as S-5!, and BIPV products are available from Sharp, Suntech, and Solon AG, among others. The fact is, gluing laminates to the roof may not be such a good idea, as they could heat it up in hot climates (as Unisolar's laminates have only 30% effective reflectivity), could be a potential fire hazard, and may make roof repairs (for example, for leaks) quite hard without risking damage to the laminates. And in terms of cost savings from the integration and the "easy" installation (which may involve extra insulation layers, or painting the roof first), there is no evidence that such savings, even if they exist, can overcome the lower efficiency.