Good afternoon, everyone, and welcome to First Solar's fourth quarter and full year 2019 earnings and 2020 guidance call. This call is being webcast live on the investor section of First Solar's website at investor.firstsolar.com. At this time, all participants are in a listen-only mode. As a reminder, today's call is being recorded. I would now like to turn the call over to Mitch Ennis from First Solar Investor Relations. Mr. Ennis, you may begin.
Thank you. Good afternoon, everyone, and thank you for joining us. Today, the company issued press releases announcing its fourth quarter and full year 2019 financial results, as well as guidance for 2020. A copy of the press releases and associated presentation are available on First Solar's website at investor.firstsolar.com. With me today are Mark Widmar, Chief Executive Officer, and Alex Bradley, Chief Financial Officer. Mark will begin by providing a business and technology update, and Alex will discuss our financial results for the quarter and full year 2019. Following these remarks, Mark will provide a business and strategy update for 2020. Alex will discuss the 2020 financial outlook. Following the remarks, we open the call for questions. Most of the financial numbers reported and discussed on today's call are based on U.S. generally accepted accounting principles.
In the few cases where we report a non-GAAP measure, such as non-GAAP EPS, we have reconciled this non-GAAP measure to the corresponding GAAP measure at the back of our presentation. Please note this call will include forward-looking statements that involve risks and uncertainties that could cause actual results to differ materially from management's current expectations. We encourage you to view the safe harbor statements contained in today's press releases and presentation for a more complete description. It is now my pleasure to introduce Mark Widmar, Chief Executive Officer. Mark?
Thank you, Mitch. Good afternoon, and thank you for joining us today. I would like to start by addressing our loss per share results for 2019, which was $1.09 on a GAAP basis, with earnings per share of $1.48 on a non-GAAP basis, adjusted for litigation losses. We are disappointed with the outcome, which came in below our EPS guidance range. While Alex will provide a more comprehensive overview, I want to highlight several items that had a material impact on this result. Firstly, as initially disclosed on January 6, we entered into a memorandum of understanding to settle the previously disclosed action litigation, which was originally filed in 2012. Earlier this week, we disclosed that we entered into a settlement agreement that is consistent with the MoU.
As part of this agreement, which is subject to court approval, we agreed to pay a total of $350 million to resolve the claims asserted by the class action. The settlement agreement does not contain any admission of liability, wrongdoing, or responsibility by First Solar. While we are confident in the facts and merits of our position, we believe it was prudent to end this protracted and uncertain class action litigation process and focus on driving the business forward. As a reminder, the settlement agreement does not resolve any of the claims asserted in the opt-out action against us or the derivative action. Secondly, challenges related to our systems business over the last few months has had significant impact with respect to revenue and gross margin. These challenges relate to both project sale and completion timing, as well as higher expected costs due to adverse weather impacts.
Alex will provide more detail on the impact of these challenges to 2019 results. Despite the EPS result and a near continued intense competitive pressure across the PV industry, I would like to highlight some of our notable achievements in 2019. Please turn to slide four. Firstly, the company celebrated its 20th anniversary and reached a significant milestone of 25 GW of modules shipped. We are the world's largest thin-film PV module manufacturer and the largest PV module manufacturer in the western hemisphere. Secondly, we saw strong net bookings of 6.1 GW, as well as record shipments of 5.4 GW. Thirdly, in 2019, we produced 3.7 GW of Series 6 product, a 3 GW increase over 2018. Our Series 6 nameplate manufacturing capacity increased to 5.5 GW. Our top production bin reached 435 W. In our commercial production line, we manufactured a new record 447 W CdTe module, as validated by Fraunhofer.
These remarkable accomplishments, which demonstrate the strength of the First Solar team and culture, give us confidence in our ability to continue to realize the full potential of our competitively advantaged Series 6 platform. Turning to slide six, I'll provide an update on our Series 6 capacity ramp and manufacturing metrics. Over the course of 2019, we realized significant operational improvements. Comparing December of 2019 metrics against those of December of 2018, megawatts produced per day was up 152%. Capacity utilization, adjusted for planned downtime, increased 26 percentage points to 100%. Production yield was up 32 percentage points to 94%. Average watts per module increased 20 W, and our highest volume bin increased to 435 W. Finally, the percentage of modules produced with anti-reflective coating increased by 24 percentage points to 96%. This momentum has continued into 2020.
Comparing February 2020 month-to-date against October 2019 metrics, megawatts produced per day is up 25%. Capacity utilization, adjusted for downtime, remains over 100% at 105%. Production yield is up 2 percentage points. Average modules per watt, led by our highest bin of 440 W, has increased 7 W , and the percentage of modules produced with antireflective coating has increased by 2 percentage points. This combination of our efficiency improvement program and increased ARC utilization led to a significant improvement in the module bin distribution. The ARC bin distribution from 430 W-440 W during this period was up significantly to 93% of production. Turning to slide seven, I'll next discuss our most recent bookings in greater detail. Our fourth quarter net bookings of 1.4 GW bring total 2019 net bookings to 6.1 GW.
We are off to a strong start in 2020, with 0.7 GW of net bookings since the beginning of the year. Included in our new bookings since the previous earnings call are approximately 0.8 GW of aggregate orders for deliveries in 2022 and 2023. Our future expected shipments of 12.4 GW remain strong even after a record fourth quarter shipment, which accounted for 31% of the full-year total. Our net bookings for the year included 1.7 GW of debookings, including 1.2 GW of debookings in the fourth quarter. Approximately 0.9 of the fourth-quarter debookings related to a customer in financial distress. To improve our counterparty risk, we have relieved this customer of their obligation and recontracted the majority of this volume. Note, demand for our Series 6 module remains strong, as reflected in our gross bookings since our last earnings call of 2.6 GW.
We're very pleased with our bookings performance in 2019, which exceeded our 1:1 target book-to-ship ratio. We believe our record of meeting pricing and delivery commitments for long-dated agreements enables us to contract significant module volume, not only in the near term, but also in 2021 and beyond. I'll now turn the call over to Alex, who will discuss our fourth quarter and full year 2019 results.
Thanks, Mark. Before reviewing the financials for the quarter and full year in detail, I'll first provide some context around the factors which led to the 2019 year-end results falling below our guidance ranges. Firstly, in early January, we settled our class action lawsuit for $350 million. As noted earlier, the settlement remains subject to approval of the court. Additionally, we accrued $13 million of estimated losses relating to the separate opt-out case. This represents our best estimate of the lower bound of the costs to resolve this case. These litigation losses were recorded as operating expenses in the fourth quarter of 2019. Secondly, with respect to our international systems business, we did not complete the sales of our Ishikawa, Miyagi, and Anamizu projects in Japan.
At the time of our last earnings call, we were still evaluating the impact of the then-recent Typhoon Hagibis, which passed near to our Miyagi project. We've since completed our analysis of the impact, which shows limited damage to the project site itself, which is largely expected to be covered by insurance. However, the road along which the gen-tie line is designed to run was seriously damaged, which prevented the project sale in 2019 and also impacted the structuring and timing of the private fund vehicle, which was expected to acquire all three assets. In addition, the sale of approximately 40 MW of assets in India, included in our guidance for the year, did not close. With respect to our U.S. systems business, in Q4, we completed the sale of 150 MW Sun Streams 1, 100 MW Sunshine Valley, and 20 MW Windhub A projects.
All three projects, which are being constructed by First Solar EPC, achieved substantial completion in December of 2019. During the quarter, we also completed the sale of our 160 MW Little Bear portfolio. This transaction was structured as a sale of the project entities with an attached module sale agreement, utilizing a third-party EPC provider. Previously used in the sale of our Cove Mountain and Muscle Shoals assets in Q2 of 2019, this structure is reflective of our recently announced transition to a third-party EPC execution model. Note that in our public filings, this also has the effect of removing the Little Bear assets from our systems pipeline table and adding an equivalent volume of expected future module sales.
Thirdly, as it relates to our Series 4 production in Malaysia, in December, we began the transition of one of our remaining two Series 4 plants to our second Series 6 factory there and incurred $6 million of shutdown costs. We anticipate discontinuing our remaining Series 4 production in the second quarter of this year. With this context in mind, I'll discuss some of the income statement highlights for the fourth quarter and full year 2019. Starting on slide nine, net sales in the fourth quarter were $1.4 billion, an increase of $853 million compared to the prior quarter. The higher net sales were primarily a result of our U.S. project sales and increased module shipments. For the full year 2019, net sales were $3.1 billion compared to $2.2 billion in 2018.
Relative to our guidance expectations, net sales were lower, primarily due to the aforementioned delay in the sales of our Japan and India assets, as well as lower-than-forecast percentage of completion from our U.S. asset sales and timing of revenue recognition on certain module sales. As a percent of total quarterly sales, our systems revenue in the fourth quarter was 53% compared to 32% in the third quarter. For the full year 2019, 52% of net sales was from our systems business compared to 78% in 2018, as we expanded our module sale volume in 2019. Gross margin was 24% in the fourth quarter compared to 25% in the third quarter. For the full year 2019, gross margin was 18% compared to 17% in 2018.
The Systems segment gross margin was 24% in the fourth quarter compared to - 5% in the third quarter. Fourth quarter was positively impacted by the sale of our U.S. systems assets previously mentioned, offset by two principal items. Firstly, with respect to a project we're constructing in Georgia, which is among the final projects being constructed in-house by First Solar EPC, as we transition to a third-party execution model. In late December 2019 and January and February of 2020, we experienced heavy rainfall at the site, which resulted in project delays and increased costs, impacting gross margin by approximately $12 million. Secondly, based on an ongoing dispute with a customer, we recorded a reduction to revenue and gross margin of $7 million related to certain outstanding EPC project receivables. We're evaluating our legal options with respect to this matter.
For the full year, this led to a Systems segment gross margin of 16% compared to 25% in 2018. The Module segment gross margin was 24% in the fourth quarter compared to 40% in the third quarter. Third quarter was positively impacted by an $80 million product warranty liability reserve release, equivalent to 22 percentage points of gross margin, as discussed on our third quarter earnings call in October. In the fourth quarter, module gross margin was negatively impacted by the aforementioned Series 4 shutdown costs, and $13 million of ramp costs as we continue to ramp our second Perrysburg facility. For the full year, Module segment gross margin was 20% compared to - 10% in 2018. SG&A, R&D, and production startup totaled $88 million in the fourth quarter, a decrease of approximately $9 million relative to the third quarter.
This decrease was primarily driven by a reduction in start-up expense from $19 million in Q3 to $7 million in Q4 as our second Perrysburg facility ramped. SG&A, R&D, and start-up totaled $348 million in 2019 compared to $352 million in 2018. Combined with the previously discussed litigation losses of $363 million, total operating expenses were $451 million in the fourth quarter and $711 million for the full year 2019. Operating income was - $118 million in the fourth quarter and - $162 million for the full year 2019. Compared to our guidance for the year, operating income was lower than expected as a result of the previously mentioned factors. We recorded a tax benefit of $31 million in the fourth quarter, including a benefit of $91 million related to litigation losses.
For the full year, we recorded a tax benefit of approximately $5 million, which also includes the aforementioned $91 million benefit, compared to $3 million of tax expense during 2018. The tax benefit was primarily driven by the tax effect of litigation losses, offset by return to provision adjustments for certain foreign jurisdictions, normalization of uncertain tax positions, and a change in jurisdictional mix of income, largely due to the aforementioned project entity plus module sale agreement structure we recently employed in our U.S. project sales. Fourth quarter loss per share was $0.56 on a GAAP basis, compared to GAAP earnings per share of $0.29 in the prior quarter. For full year 2019, the loss per share was $1.09 on a GAAP basis, with earnings per share of $1.48 on a non-GAAP basis, adjusting for litigation losses, compared to GAAP earnings per share of $1.36 in 2018.
In summary, relative to our guidance, our full-year earnings were adversely impacted by several factors. Firstly, not closing the sale of our Japan assets impacted EPS by approximately $0.50, the possibility of which we indicated in our third quarter earnings call. Secondly, we had an approximately $0.20 impact from a combination of the delay in the sale of our projects in India, delayed revenue recognition due to partially reduced percentage of completion of our U.S. systems assets under construction, and the timing of revenue recognition on certain module sales. Thirdly, we had an additional aggregate $0.20 of systems business impact due to adverse weather events and the reversal of an accrual related to a customer dispute. Fourthly, severance costs associated with the shutdown of our Series 4 facilities, combined with increased variable compensation and other miscellaneous operating expenses impacted EPS by approximately $0.10 in the aggregate.
Finally, increased other income from the gain on sale of certain securities associated with our end-of-life recycling obligations was offset by increased tax expense. I'll next turn to slide 10 to discuss select balance sheet items and summary cash flow information. Our cash, marketable cash, and restricted cash balance at year-end was $2.3 billion, an increase of approximately $0.6 billion from the prior quarter. Total debt at the end of the fourth quarter was $472 million, a decrease of $9 million from the prior quarter. As a reminder, all of our outstanding debt continues to be project-related and will come off our balance sheet when the projects are sold. Our net cash position, which includes cash, restricted cash, and marketable securities, less debt, increased by $0.6 billion to $1.8 billion at the end of the fourth quarter.
The increase in our net cash balance was driven by the sales of our U.S. project assets, module sales, and greater than previously forecast advanced payments received for sales of solar modules prior to the year-end 2019 step-down in the U.S. investment tax credit. Note, our year-end net cash balance does not reflect the impact of the accrued $350 million class action settlement, which was paid into escrow in January of 2020. Net working capital in the fourth quarter, which includes non-current project assets and excludes cash and marketable securities and the litigation related accrual, decreased by $0.5 billion versus the prior quarter. The change was primarily due to project development assets that were sold, advanced payments received from module sales. Cash flows from operations were $174 million in 2019, an increase of $501 million relative to 2018.
As a reminder, when we sell an asset with project-level debt that is assumed by the buyer, the operating cash flow associated with the sale is less than if the buyer had not assumed the debt. In 2019, buyers of our projects assumed $88 million of liabilities relating to these transactions. Finally, capital expenditures were $158 million in the fourth quarter compared to $183 million in the third quarter. Capital expenditures were $669 million in 2019 compared to $740 million in 2018. Our capital expenditures were primarily attributable to our Series 6 capacity expansion. I'll now turn the call back over to Mark, who will provide a business and strategy update.
Thank you, Alex. Turning to slide 12, I want to start by highlighting the strong market opportunity in front of us. In the next five years alone, as reflected in the graph to the left, the amount of PV capacity installed globally is expected to double. As shown on the graph on the right, PV in many markets is competitive with all major forms of fossil fuel generation. Market momentum for PV continues to build. Our Series 6 technology, product roadmap, and market-leading research and development are all key differentiators which we believe will enable us to meaningfully participate in this wave of demand for clean and affordable energy. Within this context of the overall market, slide 13 provides an updated view of our global potential bookings opportunities, which now total 18.1 GW of opportunities.
This includes 9.8 GW in 2020 and 2021, with the remainder, 8.3 GW, for deliveries in 2022 and beyond. In terms of segment mix, the pipeline of opportunities includes approximately 15.4 GW of module sales, with the remaining 2.7 GW representing potential systems business. In terms of geographical breakdown, North America remains the region with the largest number of opportunities at 14.8 GW. Europe represents 2.4 GW, with the remainder in other geographies. A subset of this opportunity set is our mid- to late-stage bookings opportunities of 8.2 GW, which reflect those opportunities we feel could book within the next 12 months. This subset is approximately 72% module-only, 70% North America-based, with 43% of the deliveries anticipated in 2022 and beyond. This opportunity set, combined with our contracted backlog, gives us confidence as we scale our manufacturing capacity.
Turning to slide 14, with the addition of our second Perrysburg factory during the fourth quarter of 2019, we exited the year with a nameplate Series 6 manufacturing capacity of approximately 5.5 GW. This includes increasing the nameplate capacity of our second factory in Perrysburg from 1.2 GW to 1.38 GW, enabled by optimizing tool performance, identification, and alleviation of bottlenecks, and optimizing work in process across the broader Perrysburg complex. Through 2020, we will roll out similar throughput improvements across the three operating facilities in Vietnam and Malaysia, with limited capital expenditures will enable these factories to end the year at higher than 1.3 GW nameplate capacity, an increase of 0.3 GW of aggregate nameplate capacity. In addition, we will continue factory optimization in Ohio and expect to increase nameplate capacity there by an additional 0.2 GW, resulting in a fleet-wide year-end 2020 nameplate capacity of 6 GW.
Continuing into 2021, we expect the combination of improved throughput, yield, and efficiency to increase nameplate capacity at our international factories to 1.4 GW. With the addition of the second Series 6 factory in Malaysia, this implies total international manufacturing capacity of 5.6 GW. In Ohio, through the installation of additional tools and optimizing the two Perrysburg factories into one consolidated platform, we expect to increase nameplate capacity to 2.4 GW by the end of 2021, resulting in anticipated fleet-wide nameplate capacity of 8 GW by the end of 2021. Turning to slide 15, this capacity expansion will have a meaningful impact on our production capabilities. In 2019, we produced approximately 3.7 GW of Series 6 and 2 GW of Series 4.
As previously discussed, we expect our remaining Series 4 capacity to be wound down in the second quarter of 2020, with total production in the year to be approximately 300 MW. Series 6 production is expected to increase significantly due to the start of production of Perrysburg 2 and the implementation of the aforementioned manufacturing and module efficiency improvements. In 2020, we expect approximately 5.7 GW of Series 6 production. With the second Series 6 factory in Malaysia expected to start in the first quarter of 2021, and with anticipated increased nameplate capacity in Perrysburg, we expect 2021 production of 7.3 GW- 7.7 GW. With regards to bookings, we are effectively sold out through 2020 and are approximately 2/3 sold out to the midpoint of the expected supply in 2021. In addition, we have approximately 2 GW sold into 2022 and beyond.
Turning to slide 16, I will now discuss our module efficiency improvement roadmap. On our 2017 guidance call in November of 2016, and updated at our Analyst Day in December of 2017, we provided an expectation of near and midterm efficiency goals. As shown by the purple dot and the yellow line on the graph, we expected to launch in 2018 at 420 W-430 W per module. We set out a midterm target of 460 W per module. At the end of 2018, despite a high bin of 425 W, our average watts per module was only 411 W as we faced challenges in the initial ramp of our Series 6 product.
Through continued operational improvements, increased ARC penetration, and the execution of our efficiency improvement roadmap, by year-end 2019, our average watts per module has increased to 430 W on a fleet-wide basis, with a high bin of 435 W. Today, our highest volume bin is 435 W, and we are consistently producing 440 W modules, as mentioned previously. We have certified a record production module of 447 W, which requires no significant technology changes and thus represents a near-term production target. As we look forward, we see a clear line of sight to achieving the target stated at the December 2017 Analyst Day of 460 W per module, as well as significant opportunity to go beyond that with a new midterm goal of 500 W per module. Note, unlike recent increases in crystalline silicon module sizes, the watts increase will be achieved using our current module form factor.
As previously discussed on the prior's earnings call, the key driver to achieving this efficiency increase is our Copper Replacement or CuRe program. Structured in three phases, the initial phase I work, combined with other ongoing R&D programs, is expected to lead to an approximately 20 W improvement, bringing to us a 400 W per module goal, which we expect to achieve on our lead line in the second half of 2021. After the launch of CuRe, there will be further optimization in two additional phases that will be the main drivers behind our new 500 W midterm target. As shown recently through our R&D efforts, replacing copper in the thin-film device not only serves to increase module wattage but also dramatically improves energy delivery.
This program is expected to increase the Series 6 energy advantage by improving our temperature coefficient advantage relative to crystalline silicon modules, as well as significantly reduce long-term degradation in a predictable and quantifiable manner, and thereby increase life cycle energy. Turning to slide 17, I'd like to compare the value proposition of our new CuRe Series 6 module relative to a crystalline silicon mono PERC bifacial module. While the potential energy advantages of bifacial modules are often touted, the increased costs are often overlooked. As PPA and merchant energy prices continue to decline, the ability to increase energy output with little to no increased cost is critically important. Designing a solar power plant with bifacial modules is a trade-off of cost for energy, as it typically adds incremental capital and operating costs compared to a monofacial plant.
Among others, these costs include the requirement for additional steel to enable elevated structures, additional land and development costs to accommodate increased row spacing, and increased O&M and vegetation management costs to allow for diffuse light reflection. Turning to slide 18, I'll provide some context around our module cost per watt. As presented on our 2017 guidance call in November of 2016, over a year prior to the production of our first Series 6 module, we forecasted a Series 6 cost per watt approximately 40% lower than that of Series 4, while at the same time eliminating any significant form factor difference and associated cost penalty. At the start of Series 6 commercial manufacturing, we have faced challenges with regard to certain aspects of the overall cost per watt, in particular related to glass and frame costs, compounded by tariff gyrations and uncertainty.
Offsetting these, we have seen significant improvements in throughput and efficiency, especially in our high-volume international manufacturing locations. As mentioned on our third quarter 2019 earnings call, these international facilities have consistently been producing above 100% of nameplate, reaching a recent high of approximately 120% of original nameplate. We're on a plan to achieve our year-end cost goals for these international facilities. However, in Perrysburg, the earlier production ramp of our second factory, together with the challenges related to the bill of materials, labor, and sales rate costs, created significant headwinds. With this backdrop, at Q3, we forecasted our fleet-wide cost per watt to end the year approximately half a penny higher than the internal target we set at the beginning of the year, which is where our fleet cost actually ended the year.
Based on our 2019 exit point and forecasted throughput yield and efficiency improvements in 2020, we are expecting to exit 2020 at our low-cost, high-volume manufacturing sites, having achieved the original cost per watt target that we set out in November of 2016. Throughout 2020, the module cost per watt of Perrysburg is expected to improve as we ramp our significantly larger second facility and we drive throughput improvements across the two factories. We do not anticipate to fully overcome the cost challenges experienced in 2019. Across the fleet in 2020, Perrysburg, representing 1/3 of the production, will create a headwind of approximately $0.01 per watt. Looking beyond 2020, I would like to discuss five key levers that we believe will enable us to reduce cost per watt in the midterm.
Relative to these levers, it is important to note the significant impact improved efficiency and throughput have on cost per watt. Firstly, efficiency improvements generally have little, if any, impact on the cost of producing a module. Therefore, in general, the percentage improvement in watt per module can be directly translated into a reduction in cost per watt. Secondly, throughput improvements, essentially by definition, are leveraged against fixed costs, which results in the incremental volume above nameplate capacity being the variable cost of production, or typically, the module bill of materials. Now, looking at the slide and starting on the left, the blue bar represents the original cost per watt target communicated in November of 2016, which we anticipate achieving at our high-volume international manufacturing sites by the end of 2020.
Beginning with watts per module, increased module wattage through our previously discussed R&D efforts in the CuRe program leads to a significant cost per watt reduction. Secondly, over the midterm, we see the potential to increase throughput by approximately 30%-35%, which provides a fixed cost dilution benefit. Thirdly, we are targeting an increase in manufacturing yield from approximately 95% today to a midterm run rate of approximately 98%, which provides a direct benefit to fixed and variable costs. Fourthly, we see midterm opportunities to reduce variable bill of material costs by between 20% and 30%, primarily across glass and aluminum. Finally, we believe the combination of increased watts per module and transport optimization can lead to a 10%-20% reduction in sales freight costs. Note, for comparison purposes, please remember, unlike our competitors, we include sales freight and warranty in our cost per watt.
Combined with the benefits of our CuRe and other R&D work with the aforementioned cost levers, we believe we are strongly positioned to continue to drive Series 6 cost per watt efficiency and energy improvements over the near and midterm. Relative to our commitment to technology leadership, as I mentioned previously, we have recently re-energized our advanced research team. While there is still tremendous headroom in our Series 6 platform, we continue to challenge ourselves on commercializing the next generation disruptive thin-film technology. It is exciting to see what the team has accomplished so far and the extraordinary potential there is for thin-film CdTe PV beyond Series 6. Finally, before turning the call over to Alex, I would like to provide an update on the internal review discussed on the third quarter earnings call.
As we reflect over the less than two years since our first Series 6 production module came off our initial line in Perrysburg, we are extremely pleased with the progress we've made. We've created a position of strength with our multi-year backlog and our module wattage, energy, and cost per watt roadmaps. However, as we look across the next decade, we need to challenge our business vertical strategy to assess if our product offerings are at a position of strength that can leverage point differentiation to create value for our customers in an attractive profit pool. We have been conducting an evaluation of the long-term sustainable cost structure, competitiveness, and risk-adjusted returns of each of our product offerings, including the module, development, and O&M business. At our core, we are a technology and manufacturing company.
Over time, we have added to this core competency in order to address unmet needs within the market, optimizing around and enabling the delivery of our products, and capturing an incremental profit pool. These capabilities have included, among others, project development, EPC, and O&M. As discussed in our previous earnings call, we have made a decision to transition to a third-party EPC execution model. We originally entered into the EPC business to enable cost-effective installation of our smaller form factor modules and to fill a capability gap in the PV market. Over time, market participants increased, with many having economies of scale leveraged across multiple market segments. The external ecosystem of EPC capabilities improved and risk-adjusted returns diminished at the same time as our product evolved to be more compatible with market balance of system offerings.
Consequently, the premise for us maintaining an internal EPC competency was no longer justified, and hence, we made the decision to transition to a third-party EPC execution model. The U.S. development business has likewise experienced a significant evolution, and the business that we entered into in 2008 is dramatically different today. Originally viewed as a channel to market for our smaller form factor modules, the development business initially saw PPA size in the hundreds of megawatts in a handful of markets, providing certainty of offtake for a significant portion of our manufacturing capacity. We also benefited from a first-mover advantage, enabling us to capture a profit pool incremental to our module sales. Today, we are significantly expanding our manufacturing capacity with a more advantaged Series 6 product.
Competition within the development market has increased, project sizes have decreased, and the risk-adjusted returns have reduced as aggressive pricing has resulted in benefits of the projects flowing to declining LCOEs rather than to increased development margins. At the same time, the capabilities required to be successful in the development have changed. The historic pillars of solar project development include siting, permitting, interconnection, and securing a creditworthy PPA. These skills remain fundamental. However, successful project development at a meaningful scale today requires a broader geographical market presence as well as additional competencies such as battery storage, power trading, the ability to manage increased offtake complexity, and financial structuring complexity, as well as asset ownership. In this more competitive environment, there remain opportunities for project developers to make acceptable margins.
However, for us to remain competitive in the long term, we would need to invest in enhancing our capabilities and offerings to the market to reflect this new development paradigm while maintaining a competitive cost structure. Any such investment needs to be compared with our primary investment thesis to increase modular R&D and add manufacturing capacity and improvements. Importantly, our focus is not to create internal capabilities that already exist externally. As a result, we are working with an advisor to evaluate strategic options to best position our U.S. development business with a mandate to position the business to succeed in a continuing evolving market for solar generation assets while maximizing value for First Solar shareholders.
While we are open to partnering with a third- party who possesses complementary competencies and capital to further scale the business, the pursuit of a partnership could potentially result in a complete sale of the U.S. development business. Turning to O&M. We entered the business at the same time as we entered into utility scale development and EPC in order to satisfy another unmet need in the PV market and take advantage of another profit pool within the utility scale space. Our O&M business was a natural extension of our position as one of the largest developers and EPC contractors in the PV industry, allowing us to maintain a long-term relationship with the counterparty and the project after it was developed, sold, and constructed by us.
Over the last several years, we have expanded our O&M business beyond our captive development pipeline to third-party developed projects with and without our modules. We have created a formal position as the largest O&M provider in the U.S. our economies of scale largely have created a competitive advantage and allowed us to maintain a profit pool in an aggressive pricing environment. However, beyond scale, additional value-added services and cycles of innovation are needed to enhance our O&M value proposition and deliver services in a more cost-effective manner. We continue to evaluate our O&M strategy in light of these requirements. For clarity, through our ongoing evaluation, the objective is to ensure our O&M business is able, without constraints, to achieve its full enterprise value potential and continued market leadership.
The consideration of strategic options for our U.S. development business is at the preliminary stage and may not result in any transaction being consummated. We do not intend to disclose further developments with respect to this evaluation process, except to the extent the process is concluded or is otherwise deemed appropriate. I'll now turn the call back over to Alex, who will provide 2020 guidance.
Thanks, Mark. Turning to slide 21, I will begin by discussing the assumptions included in our 2020 guidance. Given the uncertainty around any outcome from the evaluation of strategic options for our development business, our 2020 guidance assumes no change to existing lines of business. Starting with production, our Series 6 volume is expected to increase to 5.7 GW, with an additional 300 MW of Series 4 prior to shutting down our remaining Series 4 capacity in the second quarter. As a result of this transition, we expect to incur approximately $20 million of severance, decommissioning, and other shutdown costs in 2020. 2020 volume sold is expected to be 5.7 GW-5.9 GW. As a reminder, in 2019, we structured our Cove Mountain, Muscle Shoals, and Little Bear projects as sales of their project entity with an upfront development fee and an associated module supply agreement.
In 2020, we expect to continue to structure U.S. assets sales under a similar structure, including the sale of our American Kings and Sun Streams 2 assets. Optimized for our new approach to EPC execution, this structure will have the effect of moving approximately 900 MW of sales from our Systems segment to the Module segment. The mix of 2020 net sales is anticipated to be approximately 70% module and 30% systems. Included in the systems net sales in the United States is the residual revenue recognition associated with the GA Solar 4, Sun Streams 1, Sunshine Valley, Seabrook, and Windhub A projects. Additionally, our guidance includes the sale of our Ishikawa and Anamizu assets in Japan, which may be sold together or individually. Due to the uncertainty relating to the cost and timing of the construction of the gen-tie, we have excluded Miyagi from our 2020 guidance.
Our ongoing Series 6 capacity expansion is expected to impact 2020 operating income by $55 million-$75 million. This is comprised of $50 million-$60 million of start-up expense incurred by our second Malaysia factory and $5 million-$15 million of ramp costs associated with our second Perrysburg factory. We anticipate our second Perrysburg factory will exit the ramp period by the end of the first quarter of 2020. While we're not providing specific guidance around the Series 6 module cost per watt for 2020, we do anticipate continuous improvement over the course of the year. Despite an increase in the proportion of module volume coming from our higher-cost Perrysburg facility in 2019 relative to where we ended, sorry in 2020 relative to where we ended 2019, we expect our fleet-wide cost per watt to decline approximately 10% over the year. A brief word on the coronavirus outbreak.
While we have a geographically diverse supply chain that does include partners in China that supply us with raw materials and commodities. To date, we've managed the impact of the coronavirus outbreak, and it has not had any material impact on our operations. Our guidance accordingly assumes we will continue to be able to mitigate any such impacts on our supply chain and operations without the incurrence of material costs. Finally, in addition to the previously mentioned Series 4-related shutdown costs, as part of the strategic review and cost structure analysis that Mark discussed earlier, we've recently effected a reduction in force. Although we expect this to lead to $25 million-$35 million of long-term run rate savings, in 2020, we expect to see severance-related impacts of approximately $10 million from these actions. I'll now cover the 2020 guidance ranges on slide 21.
Our net sales guidance is between $2.7 billion and $2.9 billion. Gross margin is projected to be between 26% and 27%, which includes $5 million-$15 million of ramp costs. Operating expenses are expected to be between $340 million and $360 million, which includes $50 million-$60 million of production startup expenses, primarily for our second Malaysia factory. We anticipate core R&D and SG&A costs, excluding startup, of $290 million-$300 million. Operating income is expected to be between $360 million and $420 million and is inclusive of between $55 million and $75 million of combined ramp costs and plant startup expenses, $20 million of Series 4 shutdown costs, and $10 million of severance costs. In terms of non-operating items, we expect interest income, interest expense, and other income to net to zero.
Full-year tax expense is forecast to be $15 million-$25 million, which includes the benefit of approximately $60 million in the fourth quarter associated with the closing of the statute of limitations on uncertain tax positions. We expect no contribution from equity in earnings. This results in full-year 2020 earnings per share guidance range of $3.25-$3.75. Earnings are expected to be back-end weighting with approximately 20% in the first half of the year and 80% in the second half as a result of several factors. Firstly, although ASPs are expected to remain relatively flat, cost per watt is expected to decrease throughout the year. Secondly, we expect to recognize revenue on lower-margin systems business, including our remaining U.S. EPC projects, as well as our India asset sales in the first half of the year.
Conversely, our Japan assets are expected to be sold in the second half of the year. Thirdly, Series 6 ramp and startup costs, Series 4 shutdown costs, and severance charges are all weighted towards the first half of the year. Capital expenditures in 2020 are expected to range from $450 million-$550 million as we convert one of our remaining two Series 4 factories in Malaysia into our sixth Series 6 factory, invest in expanding capacity at our existing Series 6 facilities, and begin the implementation of our CuRe program. Our year-end 2020 net cash balance is anticipated to be between $1.3 billion and $1.5 billion. The decrease from our 2019 year-end net cash balance is primarily due to payment of the $350 million class action lawsuit settlement, capital expenditures, and deliveries against module safe harbor prepayments in 2019, offset by cash flows from module and project sales.
Finally, we expect module shipments of 5.8 GW-6 GW in 2020. Turning to slide 22, I'll summarize the key messages from today's call. We continue to make significant progress on our Series 6 transition, both from a demand and supply perspective. On the demand side, we ended 2019 with net bookings of 6.1 GW and a current contracted backlog of 12.4 GW. Our opportunity pipeline continues to grow, going into 2020 with a global opportunity set of 18.1 GW, including mid- to late-stage opportunities of 8.2 GW. On the supply side, we continue to expand our manufacturing capacity and expect to increase our nameplate Series 6 manufacturing capacity to 6 GW by year-end 2020 and 8 GW by year-end 2021. In 2020, we expect to produce 5.7 GW of Series 6 volume, a year-over-year increase of over 50%.
We see a significant midterm opportunity for improvements to our module efficiency, cost, and energy metrics. Despite a challenging end to 2019, we recorded non-GAAP EPS of $1.48 and are forecasting full-year 2020 earnings per share of $3.25- $3.75. Finally, following a review of our cost structure, risk-adjusted returns, and strategic value, we are exploring strategic options for our U.S. development business. With that, we conclude our prepared remarks and open the call to questions. Operator?
As a reminder, to ask a question, you only press star one on your telephone. To withdraw your question, press the pound or hash key. Please stand by while we compile the Q&A roster. Your first question comes from Philip Shen with Roth Capital Partners. Your line is open.
Hey, guys. Thanks for the questions. The first one is on your definition of midterm. I was wondering if you could provide a little bit more detail on that? Mark, I think you mentioned that the lead line with Copper Replacement technology would be starting in the back half of 2021. Is that midterm target a back half 2021 or early 2022 type timeframe? As it relates to the systems business, I was wondering if you could walk us through how we should be modeling going forward? Historically, you guys have talked about a gigawatt of system sales per year. I think that's probably what's baked into everybody's model. Should we start to feather that back or just remove that completely? Any thoughts on that would be great. Thanks, Mark.
On the midterm, if you think about the CuRe program, we'll start the initial production on our lead line in the second half of 2021, and then start to see it really realized across the entire fleet in 2022. If you think about even that, let's say the 2022, we originally sort of set the midterm goal in 2017 or so. It also gives you some indication of a horizon which we may be looking towards for the 500 W module that we've indicated as well. We're very obviously pleased with the launch of our Copper Replacement program, and we're also going to couple that with where our backlog position is right now. It really hits the window where we want it to hit. It's the window we need to sell through into in 2022 and 2023.
You should look to the majority of that volume in that window. We'll be able to have our Copper Replacement program out there and competitively pricing into the marketplace and capturing the full value of the energy yield that we would realize from that. Alex can give you some insight around the modeling piece, but I want to make sure it's clear as well, is we have committed to a safe harbor investment, and we've talked about that. We've got capability of safe harboring a couple of gigawatts. We have a mid to late-stage pipeline of close to 2 GW here in the U.S. of opportunities that we're actively engaging in. We have purposely looked to try to monetize those projects into a 2022, 2023 window.
It also somewhat ties in nicely to where the 201 tariffs start to wind down, plus your value of your safe harbor investment is most accretive in 2022 and 2023. You'll see, as we continue to build out that pipeline and monetize and contract, most of the volume is going to be out in 2022 and 2023. I think the best way to think about it right now, Phil, is not to assume any changes, because we're going down two paths. Look, I think when you position us into a utility-owned generation space, which we're seeing a lot of that happening in the market right now and a pretty significant inflection point of that happening in a big portion of our 2 GW, that's a sweet spot for us, and we'll hit home runs there all day long.
It's exactly where we want to be because we don't want to own generating assets, and we want to work with great counterparties. The problem we have a little bit of where I want to see how we can further enhance our capabilities is more complex transactions, merchant risk exposure, hedge contracts, basis risk, basically block power shape, storage. That's a space that we, I don't think yet, are where we need to be relative to the capabilities of the marketplace. We need to challenge ourselves and how do we best accomplish that, and one of the paths to do that is can we find a partner or somebody else we can work with that has those types of capabilities that are complementary to where we are?
We also highlighted to the extent we go down that path, it could result in a sale of the business through a partnership structure. Obviously, we'll look to do what's in the best interest of our shareholders. If it resulted in someone paying maximum value for the platform that we have, we may look to that as the best possible outcome if we feel we're uncomfortable with getting the partnership capabilities that we think we would need to best compete over the next decade. As you may remember, this is the objective we set out for, is how do we position not only our module business but our energy services business and our development business to be able to thrive in the upcoming decade. We need to make sure there's a path to do that, and that's what we're exploring right now.
Yeah. Phil, the only thing I'll add to that is on the near term, in the guidance, we said about 70% of the revenue line is going to be on the module, about 30% on the systems. That reflects only a pretty small portion. There's only somewhere around 300 MW-400 MW going through that line. However, as I mentioned in the remarks, I want to make sure it's clear. In the last year or so, we've been structuring deals differently as we've been looking at our EPC capabilities and looking to exit our internal EPC and go to a third-party model. What we've done is we've changed how we sell projects to selling a project SPV or entity and then entering into a module sale agreement.
If you look at all the deals we've done that with over the last year, the impact that means there is about 900 MW of volume that is going to go through the Module segment this year that had it not been for that new change in structure, would have gone through the Systems segment. We've historically guided to somewhere around a gigawatt a year of volume. If you look at this year, you are going to be somewhere around 1.2 GW, 1.3 GW, 1.4 GW of volume generated by the systems business. That volume was originated through the systems channel, although you are not going to see it flow through the Systems segment this year based on those deal structures. I think in the long term, for modeling purposes, stick to that roughly a gigawatt a year of systems business that we've guided to.
Absolutely, any changes that we guide to later in the year pending the outcome of our discussions with the market around the systems business.
Your next question comes from Brian Lee with Goldman Sachs. Your line is open.
Hey, guys. Thanks for taking the questions. I had two here. First, if I look at the module gross margin for Q4 here exiting 2019 at 24%, you assume flattish ASPs, which I think you mentioned on the call, and a 10% cost decline for 2020. It seems to imply gross margins for modules in 2020 will be high 20% or so, 28%, let's say. First, is that the right read here? I guess that's also assuming no mix shift impact from Series 6 either, given 2020 will be almost all Series 6, and you still had a meaningful amount of Series 4 in Q4. Would be curious if you could provide some color around the module margins this year. Just secondly, on the strategic review for the systems business. Just trying to understand the thought process here.
Is taking a partner, potentially, if that's an option or outcome of this review, does that lower the OpEx? Can you give us some sense of how much of your OpEx is tied to that segment versus modules? If you just end up divesting the segment in one transaction, what would be the motivation of that versus simply selling down the systems over the course of the next few years implied in the pipeline COD dates? Thanks, guys.
Yeah, Brian, I'll give you a little bit of color on the gross margin. We haven't broken out by module and system, but you can see in the guidance, we're guiding to a 26%-27% on a consolidated basis. You're right that there's limited reduction on the ASP side as we go from 2019 into 2020, although we are seeing some. We are seeing a cost per watt drag largely associated with Perrysburg. We talked about ending the year about half a penny higher than our expectations on a fleet-wide basis, largely driven by cost per watt at Perrysburg. If you think about the mix shift, although we'll get some benefit from moving Series 4 to Series 6, as we ramp Perrysburg 2 this year on a mix basis, we're going to have more relative volume coming from our higher-cost factories than we did last year.
That drag now across the fleet is going to be around $0.01 cost per watt in 2020. On the system side, we've got lower volume, hence you're seeing lower revenue on the revenue line. On the margin side, there's three other things I'd point to that are dragging down consolidated gross margin through the year. You've got startup and ramp costs that are coming in at about $60 million-$75 million coming through. That's associated with bringing Perrysburg 2 up and Malaysia 2. You've got a pretty significant drag there. We've also got about $20 million of shutdown costs associated with closing the Series 4 factory in Malaysia, and you're going to see that coming through the gross margin line as well. Finally, about $10 million severance costs.
As you look through the gross margin line this year, just bear in mind, you've got somewhere close to $100 million relative to that 2.8-ish midpoint of revenue line that's impacting gross margin on a negative basis.
Yeah. I think when you normalize for those items that are impacting the Module segment, Brian, you're going to get to a number that's in the range that you referenced from that standpoint. The systems business, first off, the way I look at this is that there's the market need and then there's internal capabilities. We have to understand, given the market needs, how do we best address that, and then what's the most efficient way and OpEx way of doing that? To sort of replicate or to invest in certain capabilities, let's say the power trading capability as an example, I don't know if we want to step into that space. To me, a partnership can bring a lot of value to us in the fact that we can create a complementary offer.
We've got a great development team with great development sites, interconnection positions, and capability with safe harbor that we've made the investment in. The real question is how do you monetize that and capture the optimal value with it? For me, it's that I'd rather, instead of internally create something that maybe is externally already in the marketplace and is already performing well, it makes more sense from my standpoint to say, how do we engage with those types of partners and then create the synergistic impact versus trying to invest heavily and create maybe not as strong of a market capability we would with otherwise with a partnership. That's kind of the motivation.
As it relates to the OpEx, look, there's a meaningful portion of OpEx that it not only resonates with just the direct, let's say, the customer-facing team from development, but it's my project finance team, it's the legal structuring cost around these deals. It's the complexity around the accounting. It drives tremendous tax-related activities and separation of new entities and setting them up and manage through that. There is a pretty significant OpEx impact. If you look at our K, we disclose that we have about 500 heads across the companies, north of 6,000 that are related to our systems business. That also includes our energy services business, which is a good portion of that total. You can tell that there's a pretty significant headcount resource intensity associated with our systems business.
Again, on some segments of the market and solutions that are required, I'll use UOG as a great example, I think we do very well there, and we'll continue to excel there. No different than our module business or no different than our energy services business, which I indicated we've created a tremendous amount of scale advantage and being a market leader. When I look at our development business, I have to be comfortable that we can create scale there as well, because that infrastructure-related cost is going to be there. You've seen it happen over the years because the cost to develop and the resources to develop a 500 MW project like we did in the early days is really no different than to develop a 75 MW or 100 MW project. Project sizes have come down, and therefore you're actually losing the leverage of scale.
Those are the things that we're looking at, and we're trying to figure out what's the right path forward to enable what we think is a great platform. We're not diminishing the platform at all. As we think through how do we make sure we can thrive and excel through this upcoming decade, there are certain capabilities we think partnering with someone else could bring to us that would further enhance the value proposition of our development business.
Your next question comes from Paul Coster with JP Morgan. Your line is open.
Yeah. Thanks for taking my question. It looks like something in the region of a dollar of the shortfall in the fourth quarter was attributable to Japan, India, et cetera. How much of that dollar approximately carries over into 2020?
I mean, it involves some of the Japan business, right?
Yeah. If I look at it, you've got about a dollar of shortfall. About $0.70 of that is related to timing. You've got Japan, India, U.S. projects, and U.S. module. There's also another, call it roughly $30 million of true cost increases. Impacts from U.S. project weathering issues. We had an accrual change relating to a dispute with a customer. We have some severance and other miscellaneous costs. If I look at those, you've got, call it $0.70 of the roughly dollar is associated with timing versus true cost impact. When I roll that forward into 2020, about $0.50 of that is going to roll into 2020. The breakdown there is in Japan, two of the three assets are being pushed into 2020. Our Miyagi asset, however, is not, just based on where we see the timing of construction and the gen-tie today.
If that changes, that could get pulled in later, but as of now, that's not in the guidance for 2020. The other piece is that previously we had assumed the structuring of our Japan assets would go through this private fund that I mentioned in the prepared remarks. Based on pulling the Miyagi asset out and the complexity we've seen, I think we're targeting now selling those assets on a bilateral basis versus in a fund with a small impact to value there. Of that $0.50 of Japan, you're going to pull about $0.35 through. The other timing piece, you're going to pull about $0.15 of the $0.20.
That's a function of, in the U.S., although we hit substantial completion on the projects that we were targeting by year-end, we had some small cost increases to do so, as well as the fact that on the India assets, we've just had some diminishing of value as we've been negotiating those sale contracts. If you look at it, you can assume about $0.50 gets rolled 2019 to 2020.
Your next question comes from Ben Kallo with Baird. Your line is open.
Hey, guys. I guess, can you talk about your low cash balance point, what you think it is with all your CapEx? My second question is just on, I guess we're all trying to figure out cost per watt. We're using $0.21 or something like that. How off are we on the mark there, going forward? Thanks.
Yeah. Ben, on cash, you've got two big things this year. You've got the remainder of the Series 6 CapEx. We talked originally about $2 billion of capital associated with what was then 6.6 GW of capacity. We are largely through that. About $250 of the midpoint $500 guidance announced for the year is the finalization of that initial capacity. Of the remaining, there's about $100 that is associated with increasing Perrysburg's capacity, and that takes us from the current 1.9 up to the long-term run rate by year-end 2021 of about 2.4. That extra $100 million is getting you about half a gigawatt of capacity. Then in the year, there's about another $150, which is other miscellaneous capacity expansion plus other spend.
If you think about that, by the end of this year, we're largely through not only the initial CapEx program, but a lot of the CapEx that's going to take us through the increase in capacity that we showed on the slide. That takes us up to a nameplate of 8 GW by the end of 2021. The other piece that you've got to remember is, you're starting the year out by immediately pulling $350 million of cash out when we settled the class action lawsuit. When you think about the low point, this should be the low point, we think, by the end of the year. If you look at anticipated CapEx going beyond, assuming no incremental greenfield or brownfield expansion, that isn't counted in these numbers today, we should be at a high point of spend.
We're through a lot of that CapEx by the end of the year. We should start to build cash thereafter.
Yeah, I think the thing about it is when you go into, especially in 2021 CapEx, burn rate's down significantly. You've got, as we show with the production plan, a supply plan that we anticipate to have about 2 GW of incremental shipments in 2021. You've got 5.7 GW relative to a high end of 7.7 GW in 2021. That is going to drive significant incremental cash flows because that contribution margin largely is going to flow through to cash. As it relates to cost per watt, Ben, look, I think the, We haven't given a discrete number, but there's many numbers that are right around that range, and those are numbers that we have said before that we're comfortable with.
We've got a near-term issue we're still working through with Perrysburg, and we've got a headwind against the fleet, which I highlighted in our remarks, of about $0.01. We expect to work that out over time, but not in 2021. Excuse me, 2020. We've got ramp-related costs and other things that are flowing through, and severance as the decommissioning costs and other things like that are starting to flow through the results for 2020. The other thing I want to make sure that we don't miss is that we have many levers to go, let's use your leaping-off point just as an example. The levers of which we can continue to drive costs down are significant and highlighted in the slide that we showed during the call. Just the increase.
Today, we're at an average of 435 type number, slightly lower than 435 W, and if we take that up to 500 W, that's tremendous increase in watts, which largely scale and correlates specifically to a reduction in cost per watt. Throughput that we have, the team's done a tremendous job of putting forth a roadmap that can take our original nameplate capacity of a factory and increase it by 1/3. That's another significant lever. I think there's a near-term issue that we're dealing with. When we look across the horizon and where we can ultimately go from a cost and performance standpoint of our product, we're extremely happy with where we are and what the potential is in front of us.
Your next question comes from Michael Weinstein with Credit Suisse. Your line is open.
Hi, guys. Hey, on slide 18, is it your intention to show that costs are coming down to about half of the starting point? Is that diagram to scale, or is that not intentional?
No, Mike, it may not be clear, but if you look at the footnote there, it says, not to scale. That's not meant to be.
Okay. Is there any particular reason why you don't give an exact number? Is it a competitive reason for something, or is there some other reason?
Yes, it's exactly right. Look, we are out selling the value of the product, right? I'll give you an example. If you look at our 2.6 GW kind of growth spaces that we booked within the quarter, the average ASP across that 2.6 GW, which includes volume that goes out into 2022 and 2023, is slightly down from the average that we reported in the last Q, which I think if you did the math on the last Q, was somewhere around $0.34 or something like that. We're selling through out into a horizon that's 2022 and 2023, and we're still holding very strong ASPs. The value of CuRe hasn't been captured in that horizon yet.
Our contractual structures, as we go that far out, will allow us to capture the value of the energy that the product can ultimately deliver on a long-term degradation benefit, temperature coefficient benefit, spectral response, efficiency, everything. It all can accrete value as we move forward. If we were to provide a discrete cost that gave you a number that's out into 2021, 2022, and 2023, then my customer starts to hold me accountable to a cost-plus model. That's not what we want to do. We want to be out there selling the full entitlement of the value that we create and not get stuck on a cost-plus. We have purposely moved away from giving a discrete cost per watt. If you are adept at modeling, you can easily take the inputs that we've given to give you an indication.
There is room still to go as we move forward. I think that's really what's most important. As you think through the window, this business is going to continue to scale. We're going to maintain and hold a relatively tight fixed cost structure, and we'll leverage that and drive incremental operating margin. Right? That's what we've been saying since day one. The more transparent I am with a cost number, the more vulnerable I am to really realize the full potential of the business model that we've created and the technology, and we want to capture the value of that technology.
Your next question comes from Julien Dumoulin-Smith with Bank of America Merrill Lynch. Your line is open.
Hey, can you hear me?
Yes.
Hey. Excellent. Hey, guys, just wanted to talk a little bit about the systems business. Again, sort of status quo, independent of monetization here, what systems volumes are we talking about in nominal terms? I know things are moving around here. Do you recognize in 2020, and then more on an ongoing basis as you think about the size and scale of your operations today? Again, this is also with the thought process of what is this business worth in a monetization scenario as we think through these peak systems years coming up here. You guys have previously talked about edging upwards that gigawatt size number in terms of annual systems business. Where does that stand now, 2020, 2021, and sort of go forward, if you will?
Yeah, Julien, we previously guided to a gig and sometimes up to a gig and a half. That gig and a half a year was also partly EPC projects. As we've now moved away from internal EPC to a third-party execution model, the best we can give you is to anchor around that gigawatt a year of volume. As I mentioned, in 2020, you're not going to see that relatively flow through the Systems segment, just given how we structured some of these deals. The same volume of that was approximately originated through the Systems segment, through that channel. On the accounting side, by virtue of how we structured those deals, you're going to see that mostly flow through the Module segment in 2020. Absent any strategic change here that we've been discussing, continue to model around that gigawatt a year.
The other thing I'd say is, as you go further out, we have safe harbor 2 GW of capacity. We would like to try and use that out in more in 2022 and 2023. We've largely sold through a lot of our capacity in the near term, and we've done so capturing full value entitlement for the module. Being able to do that without losing much money relative to a systems sale, but without taking the risk of a systems deal. I think from a relative risk perspective, we've captured full value in the next couple of years. We think if you look through into 2022 and 2023, when the relative delta between a safe harbor 30% ITC project relative to going down then towards 10% is greater, that's when we'll be looking to deploy that product.
In terms of value, there's significant value creation out in 2022 and 2023.
Your next question comes from Moses Sutton with Barclays. Your line is open.
Thanks for taking my question. For the systems business, assuming you find a partner that, let's say, solves the basis and related risks, how might cost need to come down, the core cost itself, you're moving to third-party EPC, to, let's say, allow you to bid at PPA prices of $30 a megawatt hour while still maintaining, say, around a 20% margin as a developer?
Cost as it relates to our own development costs to achieve that margin? Given where we are right now in where PPA prices or module prices or anything's moved towards, I'm not sure that a margin percent is always necessarily the best way to look at it. Partly because the development revenue stream on number from a cents perspective is relatively low. I would actually like to ensure that we can capture at least 30%-40% margin on our development activities in order to say that it's sustainable and in a position that we want to maintain. You have to look at the risk profile that you're taking. This is why our preference is to do more UOG, and that's where we're trying to position ourselves.
Every time there's a change in a merchant curve that gets published every six months, there's a risk that you're taking because you bid a merchant curve two years ago, a year ago, whatever it is, and every time that gets updated, especially with shorter tenor PPAs, I've got a risk for every time a merchant curve moves one way or the other. We prefer to try to find long tenor PPAs. We also prefer to look to UOG, which I just have to worry about either providing a site with a module agreement or building a power plant and transferring that to the long-term owner. There's others that are willing to take those other types of risk, and it's the risk they're comfortable with taking.
We just would like to see if there's a path out there that we can partner with someone that is comfortable with those risks and has ways to manage those types of risks that I would say we're not in a position to do today. If you look at Texas as a market, Texas as a market's a very strong market. We do extremely well in Texas from a module standpoint. We just haven't been successful doing development because the hedge contract structure that you see in Texas, the merchant exposure you see in Texas, the basis risk that's in Texas, those are not things that we're good at managing or hedging those types of risk profiles, but others are.
Do we find a path that we can be complementary, we can continue to develop and provide great module product, and build, if need be, a power plant, but somebody else is willing to step in and take those other risks. Those are paths we're looking at. We're also looking at would somebody be willing to team with us at the time of bidding into a PPA, where they're willing to provide underwriting assumptions. We lock in on how we underwrite a PPA. Then I hedge my exposure from the time of award versus carrying risk profile forward until time of sell down our COD or whatever point in time that may be.
That is all the time we have for questions. This concludes today's conference call. You may now disconnect.