Vinod Khosla on the Technology Pathway to Biofuels

[pagebreak:VinodKhoslaPerspectiv]

[Editor's Note -- we published Mr. Khosla's introductory piece, What Matters in Biofuels? here. He now gets into the details of his perspective on biofuels and his firm's investment thesis.]

Part 1: Production technologies: where are we?

The financial crisis of 2008 set back a number of projects and slowed actual construction of pilot and demo plants like it did in all industries, be they biofuels or traditional fossil energy projects. That, coupled with the negative press for corn ethanol, slowed progress and funding of biofuels to a crawl, causing delays of up to two years in many business plans. But despite these challenges, entrepreneurs have persevered and in many cases are further along than expected in achieving practical economics that now justify their first commercial units (FCUs). 

FCUs are particularly hard to fund given that they are first-of-a-kind technology, and there is now a risk-averse capital funding environment.  Further, FCUs are generally of smaller size than the optimal scale commercial plant, are not “value engineered” (the goal of the FCU should be, in my view, cash break-even or better, and proof of technology, not cost optimization), and hence their output suffers from more challenging economics in commodity product markets.

Hence, some companies like Amyris, Gevo and LS9 have targeted higher-value specialty products to commercialize, until yields improve, scale increases and costs and capital raising risks decline. By starting with high value products and going to progressively lower value (per gallon) markets the addressable market size will continually expand for these companies, eventually hopefully encompassing most of the fuels market -- among the lowest price per gallon markets. Some like Coskata and Lanzatech (and others) are developing multi-product technologies, which gives them optionality. Others like KiOR have gone directly into fuels because of their technology, supply chain compatibility and current production costs, which can be competitive with fossil oil.

Below is a rough summary of the major technology pathways and our current comments on them.  We reserve the right to change them as we learn more or the technologies progress; in fact, this progression of views and the emergence of other new technologies is expected. Over the last two years I have been pleasantly surprised at the progress in direct-to-hydrocarbon renewable fuels. Finally, there are efforts that are too early in their development to characterize, such as Synthetic Genomics, Virent, Codexis, electro-fuels, etc., to estimate with any reliability, and others for which not enough information is available to us.

Incidentally, given the amount of R&D that is now done in private companies that don't publish their results, I find knowledge of status among academics to be relatively out of date, too. Unfortunately, a somewhat unreliable rumor mill appears to be the best balance of reliable/current information along with informed projections of technology pathways’ potential.

Table 1 summarizes the major technology pathways (leaving out the currently commercial starch/sugar to ethanol processes; bolded companies are part of the Khosla Ventures Portfolio). 

Table 1: Major Biofuel Technology Pathways

Table 2: (cellulose to fermentable sugars)

There are dozens of companies all taking different approaches, and vying for different markets. 

Some technologies, like Gevo, LS9 and Amyris, focus on creating a single molecule, whereas other technologies, such as KiOR, create a diverse mix of chemistries (much like crude oil has) but at lower cost that can be blended with crude oil and dropped directly into a refinery, or if hydrotreated can be used as gasoline or diesel blendstock. These processes result in a range of production costs, many of which will be acceptable, depending on the end-products. For example, thermochemical processes which yield mixtures of chemistries (such as crude replacements) should aim for costs below $20 to $25/ barrel of oil equivalent (feedstock costs including 15 percent IRR on capital investments) and be market competitive within 5 to 7 years of their launch (without subsidies).

Biology-based pathways that create precise molecules can afford more expensive production processes and feedstocks due to the ability to tune the end-products and target higher value chemicals segments with higher price points.  In many instances, product separation/ purification or post processing is a critical cost factor. In the long run, fuel processes must use low cost non-food feedstocks (for scalability and cost) and high yield, low cost production processes, have minimal post-process steps to get to marketable product, or produce a valuable co-product (it’s important that the market for this co-product is of similar size to the main product, or else it won’t scale).

For high value chemicals, these issues are less critical due to lower cost sensitivity.  Some processes, like those of Synthetic Genomics, are early enough to be difficult to assess viability in my view, though they appear to carry significant environmental risk (much like the GMO controversy) if used in open ponds or open ocean. In bioreactors, most algae-based processes carry high capital cost risks. Other processes like Virent, Qteros, Joule and numerous others are less well known to me. 

The next page covers the major technologies in more detail.

[pagebreak:VinodKhoslaPerspectiv2]

Page 2 of 3

Below are the major technologies in more detail:

Sugar Fermentation: Regardless of the carbohydrate source, a conventional yeast fermenting sugar to ethanol already operates at the highest carbon yield theoretically possible in corn ethanol plants -- i.e., 4 out of the 6 carbon atoms from a glucose molecule -- and much of the energy ends up in a molecule reduced enough to burn in a combustion engine. For example, at theoretical yield, glucose is converted into two 2-carbon ethanol molecules, or one 4-carbon isobutanol molecule, or four carbons of a multi-carbon hydrocarbon molecule.  Without the input of carbon-free energy, it is theoretically not possible to improve upon this yield.  

Ethanol plants are able to operate at these yields in part because the yeast function without oxygen, i.e., anaerobically.  Some biological processes and end-products require oxygen (i.e., are aerobic), which can limit yield as the cells “burn” additional carbohydrate feed stock to provide the cell with energy.   Generally anaerobic or near-anaerobic/ micro-aerobic fermentation technologies can approach the maximum theoretical yield and make the most efficient use of any given feedstock. However, anaerobic processes have their own drawbacks and limitations; anaerobic processes tend to achieve lower growth rates in certain conditions (which leads to higher capex), and there are several high energy molecules that are not possible to directly synthesize through anaerobic pathways. In the end, the goal is to get to a product of commerce the lowest-cost way possible. Depending on the product, one can either take an entirely biological route, or use a mixture of biology and chemistry.

Because ethanol plants already operate at high carbon efficiency on the available feedstocks, it is reasonable to compare new technologies to these existing plants as one metric to gauge their performance.  A typical ethanol fermentation continuously produces ethanol at the rate of 2 to 3 grams of ethanol per liter of broth per hour at 90 percent to 95 percent theoretical yield.  These plants are fairly low tech and inexpensive to build -- essentially, they are just large vessels with simple distillation equipment (though distillation can be energy intensive and expensive) -- yet they still have capital costs of about $2 to $3 per gallon of annual capacity. If a new process can operate anaerobically, the capital and operating costs of that plant will be directly proportional to the performance of the new process verses ethanol production (e.g., if the production rate is half that of ethanol, capital costs are roughly doubled).  If a new process requires oxygen, additional capital and operating costs will be required to deal with mass transfer issues, compressors, gas handling, and safety concerns. There are also micro-aerobic processes. New carbohydrate feedstocks, especially sugars derived from energy crops, may require additional capital, but will reduce feedstock costs.  

In most processes, the cost of feedstock is the primary contributor to the cost of production.  Some processes can tolerate a wide variety of sugar types and purities, while others can’t; still others can tolerate complicated mixtures (like dry mash ethanol), while others require pure sugar feedstocks in order to minimize separation costs, or to make the separation of the product even possible. In determining the lowest cost route to a given product, one must consider the whole of the business system and production processes from raw material to finished commercial product.  Separation, post-processing, purity of feedstock (Virent, rumor has it, requires very pure and hence expensive sugars to preserve catalytic efficiency in their thermochemical route) and feedstock costs are additional significant considerations. 

Each process requires a different quality or purity of feedstocks, and hence feedstock costs can vary even within the same technology category. Gevo, for instance, is able to use starch directly, and their organisms can tolerate a high level of impurities. Meanwhile, HCL CleanTech’s extraction-oriented approach to making sugars appears to provide sugar feedstocks of higher purity than those coming from traditional corn or sugar cane sources, and could be significant for the long-term competitiveness of sugar fermentation platforms. In addition to using yeast and bacteria, there are companies that produce various chemicals from microalgae by fermentation, using sugars, such as Martek and Solazyme. Thus the algae in this case are just another conversion engine similar to yeast or bacteria (as opposed to what many people think of as algae which converts solar energy to lipids), by transforming sugars to fuels and chemicals. Their near-term business models are for higher value products and biofuels. The rate of yield and productivity improvements should be similar to other single-celled organisms (like e. coli and yeast), though microalgae does not have as extensive a molecular tool kit for optimization as these other two organisms.

One often-expensive part of the fermentation production process, and any other biological approach, is the separation of the final product from the fermentation broth; for example, most algae projects face this substantial challenge.  Distillation and even centrifugation are often used, which is relatively expensive.  Some companies have succeeded in engineering organisms to secrete the chemicals (e.g., LS9, Amyris), which then phase-separate, enabling cheaper separation. Others like Gevo have developed proprietary low cost, low energy separation techniques. Several companies are developing the capability to convert the fermentation product into a wide variety of chemicals and fuel end-products, though each step generally adds additional expense and yield reduction. Direct single-step production can be especially important in cost-sensitive markets like fuels.

Gas-phase technology:  Cellulosic feedstocks can be gasified and either thermochemically processed or used as a fermentation feedstock for fuels and chemicals.  Capital costs for purely thermochemical technologies are much higher than first generation corn ethanol plants, but these plants are typically powered by the equivalent of the non-carbohydrate portion of the biomass feedstock, which reduce operating costs to make total costs cheaper than corn ethanol. Bio-fermentation is currently the most promising backend to syngas production, because it can co-produce precise chemistries and hence high value chemicals and cheaper fuels than chemical catalysis, and operate cost-efficiently and with higher target molecule yields.

The big challenge here is the capex for the gasification front-end, which can be anywhere from $4 to $6 per gallon (ethanol equivalent) or more of yearly capacity but one gets the advantage of much lower opex. The economics should work well for ethanol production (where the benchmark should be corn ethanol production costs -- the cash costs for this approach should be substantially cheaper than for corn ethanol, but the economics get even better when specialty chemicals are produced using this approach. Product flexibility of the facility may be another factor that could be engineered into the design to offer lower risks for the capex. As a result of this, some gasification and fermentation companies have considered partnering, to marry efficient gasification and fuel production technologies.  Lanzatech has cut out the front end by planning to bolt on to already existing sources of raw waste gases (from steel mills and coal plants or any source that includes carbon monoxide and/or hydrogen), dropping capex substantially, with low opex.  They can use biomass gasifiers as well to produce the syngas as feedstock for their process. 

In any biological process, bug productivity is important and sometimes challenging to maintain.  Many companies maintain redundant feed streams, seed reactors, and advanced monitoring and control of any inhibitory molecules to assure continuous productivity.  In our view, the traditional path of chemical catalysis of syngas to fuels (be it ethanol or Fischer Tropsch (FT) synthesis), appear economically challenging. FT is generally the domain of old-style thinking.  It is sometimes used as a code word for coal to liquids, which does not chart a sustainable path.  Besides, we believe bio-fermentation of any type of syngas is the most cost-effective approach; most FT advocates are simply not up to date on the day-to-day developments in the leading startups.

Direct to liquid thermochemical technology: A second thermochemical route, which KiOR takes, processes solid biomass direct to liquids, while catalytically removing oxygen. Their unique approach employs a FCC-like (Fluid Catalytic Cracking) process from the oil refinery world with a special catalyst for cellulosic materials.  The input is woody biomass, corn stover, sugar cane bagasse, or any other cellulosic/hemicellulosic/lignin material. The catalytic process occurs for a few seconds at moderate temperatures and pressures.  The result is a low-oxygen, renewable crude oil (the same way nature makes crude oil from biomass but takes millions of years to process), as well as some natural gas. The oil yield is nearly 1.5bbls/ dry ton today and is soon expected to rise to 2 bbls/dry ton.

Unlike sugar hydrolysis and fermentation technologies, KiOR can convert all parts of the biomass, including lignin, without the need for separation. That capability will likely result in the highest transportation fuel yields per ton of biomass. In contrast to ethanol, their product looks and behaves like a high-value crude oil, or a diesel/gasoline blendstock after hydrotreating -- superior to traditional biodiesel and superior to typical U.S. crude oils because of its higher value fuel distillate fractions and lack of sulfur. KiOR initially intends to sell a diesel and gasoline blendstock. The company currently has a 15bbl/day demonstration plant running, and has just started construction on its first commercial-scale unit in Mississippi. The oil industry is familiar with the basic FCC process, which should result in easier acceptance.  Scalability for FCC processes is also well understood and broadly characterized. Additionally, since the KiOR oil is stable and miscible in most crude oils, no new infrastructure is required, and supply chain planning and logistics should be simplified for refineries. 

Cashflow positive facilities are small scale, requiring less than $100 million to be economic scale. Most of the associated jobs created are rural: not only in agriculture like other biofuels, but also in forestry and transportation.  This, along with energy security, are critical issues for any fuel attempting to gain political and business support anywhere in the world. Familiarity to the oil industry and hence the ability to assess, characterize and help scale it is also a major consideration, and will make capital formation easier. This approach should be distinguished from pyrolysis oil production, which is a very different process.

Solar to fuels (algae):  Microalgae-based processes pose several challenges.  Enclosed culture systems (tubes, panels, bags -- generically photobioreactors or PBRs), cannot be scaled up and have high operating and capital costs, in the range of hundreds of dollars per square meter (versus under $1 per square meter for farming). PBRs have had limited commercial success even for very high value (>$10,000 per ton) nutritional supplements. 

Essentially all commercial production for such microalgal products uses open ponds, mostly raceway-type with paddle wheel mixing.  But even the costs of such simple ponds are too high for commodities, whether feeds or fuels, due in part to the relatively small scale of current production systems and in part to the small markets for these high value algae products.  Engineering designs and techno-economic analyses suggest that much lower cost systems could be built and operated, if pond designs are simplified (e.g., earthen ponds), scales are increased (to 10 acre ponds, versus about 1 acre now) and low-cost harvesting process are developed. However, even then, production of algal feeds and fuels would require achieving very high productivities, nearly 50 tons per acre and a high oil output, about 4,000 gallons per acre. Although high productivities are possible in principle and are routinely forecast, being one of the main attractions of microalgae biofuels, they still remain undemonstrated continuously at scale, and will require considerable long-term R&D. This is to say nothing of problems such as how to actually stably cultivate and cheaply harvest the algae. General Atomics currently has a project with DARPA in Hawaii to build a 1,000-acre raceway for fuels production; this will be an interesting test case, but I am skeptical.

Thus, I hold that solar algae biofuels still require a great deal of R&D and likely are ten years out, which I do not see as attractive for a venture investment.  As for the case for fermentation processes, the path from the current high cost of commercial production to biofuels most likely will pass through intermediate stages of scale and value, such as specialty animal feeds priced at $1,000 to $2,000 per ton and with markets in the hundreds of millions of dollars. Many claim that a major benefit for microalgae, namely, that they can use the CO2 in flue gases from power plants. I dismiss this out of hand: it makes no difference whether the CO2 comes from air, as for crops and trees, or from a power plant, as for algae, except that air CO2 is free, and sustainable.  The algae doesn’t “sequester” the CO2 (or limit CO2), as its source is still a fossil fuel and its destination (as a fuel) is still the atmosphere. 

One near-term application of microalgae is in wastewater treatment, and there the algal biomass can only be used for biofuels. But these would be relatively small, dispersed facilities, and any venture in this space requires a business model to capture value from the development of such technology.  The idea that co-production of algae fuels and nutraceuticals would improve the economics is a difficult one to defend.  Though co-products are valuable, the scale of nutraceuticals is much smaller than fuels.

Before leaving algae, I must also mention seaweed; also algae, but macro-, not microalgae, which are getting some interest in Asia and Europe. DOE ARPA-E recently funded a JV of Dupont with a startup, Bioarchitecture Lab, in California, to produce biobutanol from seaweeds.  A very challenging concept, and in very early days, but that is what ARPA-E is about; I will keep watching.

Other technologies: There are also a few technologies that are “too early to tell.” One set of examples are so-called “electrofuels.”  ARPA-E just supported 13 projects with small grants that use electricity, hydrogen or ammonia as an electron source to drive fuels production. The idea is to harness new microbe pathways to form fuel molecules of various types, using only electricity, CO2 and sometimes water as an input. These efforts are extremely early, and may yield something ground-breaking.  Still, the source of the electricity matters and much of ours is still from coal (and our hydrogen is from natural gas).

The next page covers the factors which affect the attractiveness of any biofuels technology

[pagebreak:VinodKhoslaPerspectiv3]

(Page 3 of 3)

Considerations for companies and technologies

There are several factors which affect the attractiveness of any biofuels technology:

Environmental factors/policy: Any successful biofuel should have a life-cycle analysis significantly better than petroleum (RFS II requires 20 percent improvement over gasoline for new renewable fuels processes (corn based), and 50 percent to 60 percent for advanced biofuels, cellulosic biofuels and biomass-based diesel[2]), though national security is a big enough reason alone to pursue biofuels.  In my view, a technology and its input feedstocks should have at least 50 percent carbon reduction initially to have a shot at 80 percent reduction as the process and supply chain matures.

Three standards of advanced biofuels:

Biomass-Based Diesel: 1 billion gallons by 2012 and beyond

–      E.g., biodiesel, “renewable diesel” if not co-processed with petroleum

–      Must meet a 50 percent lifecycle GHG threshold

Cellulosic Biofuel: 16 billion gallons by 2022

–      E.g., cellulosic ethanol, BTL diesel, green gasoline, etc.

–      Renewable fuel produced from cellulose, hemicellulose, lignin or algae

–      Must meet a 60 percent lifecycle GHG threshold

Unspecified Advanced Biofuel: Minimum of 4 billion gallons by 2022

–      Essentially anything but corn starch ethanol

–      Includes cellulosic biofuels and biomass-based diesel

–      Must meet a 50 percent lifecycle GHG threshold

Renewable Biofuel: Up to 15 billion gallons of “Other Biofuels”

–      Ethanol derived from corn starch -- or any other qualifying renewable fuel

–      Must meet 20 percent life cycle GHG threshold if produced in new facilities

–      Existing biofuels facilities not required to meet conventional GHG threshold

Section 526 of the Energy Independence and Security Act (EISA) of 2007 is a major driver for operational purchases of biofuels for the federal government.  In short, it says that any alternative fuel the government purchases must have a lifecycle greenhouse gas emissions profile equal to or less than petroleum.  Fuels that meet EPA’s RFS2 standard will comply, but so too will other feedstock/ fuel processes.  I believe a CLAW framework, as I have proposed previously, would be ideal for measuring the environmental impact of biofuels beyond just their carbon footprint. CLAW stands for 'carbon emissions, land use, air pollution and water use' rating, and should be used to assess each new and currently operating plant individually. I have written about this in an earlier paper in detail,[3] and frankly it should be applied to any new technology, not just biofuels.

Without an accurate and holistic assessment of carbon emissions, land use impacts, air pollution and net water usage, the true impact of a new technology is difficult to measure.  The Roundtable for Sustainable Biofuels, based out of Switzerland, is currently drafting a potential system. In the absence of such a standardized rating system, these factors should be qualitatively assessed, hopefully by industry analysts. The individual monitoring and assessment of each facility is critical, because general ratings based on a technology platform will result in abuse. A technology could be environmentally beneficial in one place while quite destructive elsewhere, due to implementation or simply environmental variation.

Indirect land use calculations are complex, though I believe the California “low carbon fuels standard” is a reasonable start to the effort of characterizing lifecycle carbon emissions of each technology. I recommend this approach to other jurisdictions. Water use should be measured against the water used in producing and refining a gallon of gasoline or diesel.

Thermochemical processes will, I suspect, do substantially better than fermentation technologies and even fossil oil production/refining technologies on water use.

Water use even among fermentation technologies can vary widely. I suspect that the majority of feedstock cultivation for biofuels over time will go to rain-fed non-food agriculture and forestry.

Another important aspect to consider is the production of unwanted byproducts, vs. useful byproducts.  As an example, some biodiesel processes produce glycerin, which is currently in oversupply. Definitions of good versus bad byproducts are malleable, as unexpected technologies may be created to usefully consume whatever byproducts are created.

Ability to scale economically: There are several factors that affect a company’s ability to scale quickly.  It would be naïve to believe that Government regulations do not play a role.  In addition to the team’s experience, and the ability of the feedstock to scale, which I will cover later, one needs to consider feedstock fungibility, the nature of plant manufacturing, the ability to build exact replicates, centralized plant control, personnel training, and many other factors.

Processes that can support multiple types of feedstocks, such as the gasification and thermochemical processes, will have a huge advantage as initial feedstocks like wood chips or bagasse can be supplemented by forestry waste (whole logs, tree bark, leaves and branches, the 30 percent to 50 percent of product that goes to waste in hardwood processing, sugarcane field waste and potentially bagasse, agricultural waste like corn stover or corn cobs, etc.), all of which will improve scalability and reduce feedstock costs over time, as well as improve feedstock price stability.

These feedstocks also offer significant upside, as the harvesting ecosystem for such peripheral biomass is developed by the likes of John Deere and Caterpillar, and will enjoy declining costs. The development of productive polycultures and perennials may also play a big role in net positive impact from feedstock cultivation. Personally, for fuels applications, I might avoid sugar-based fuels technologies until -uch time that HCL-like acid (non-enzymatic) hydrolysis of biomass technologies become available, for scalability and price volatility reasons. Still, sugar will likely remain a source for non-fuel renewable chemicals.

Different technologies are on very different cost trajectories due to their capital structure, materials, technology maturity and the ecosystem being built around their feedstocks.   Current costs are not always a good predictor of future costs, and cost trajectories that rely on eeking out the last few percentage points from 90 percent of theoretical yield mature technology are much harder to realize than those that aim are improving from 50 percent efficiency, as an example.

Regulations/certification: New fuel molecules require EPA approval, though there are a long list of registered fuel additives for diesel and gasoline. Even drop-in molecules require testing and requalification, but those tests are much less onerous compared to those trying to introduce new molecules. Ultimately, large differences exist in time to market of new fuels molecules.

There are extensive Tier 1 (combustion emissions for known molecules) and Tier 2 (extensive toxicology work for new molecules) tests required for approval. Hydrocarbons that are already in existing conventional fuels will likely be easily approved -- unless there are biologically active compounds still included, like hormones, etc.  As a result, process impurities need to be closely monitored. It is necessary for molecules to meet existing ASTM specifications (or go through a process of developing a new spec) to be sold as fuels.  ASTM specifications for blend stocks and final fuel blends must be completed, and continued participation is required to keep specifications current with constantly changing conventional fuel specifications.

As an example, Gevo has registered their isobutanol as a gasoline additive.  Meanwhile, Amyris attained EPA approval with Farnesane to blend up to 35 percent with ultra-low sulfur diesel, and LS9’s fuel product is registered as a biodiesel. Several companies have yet to register or complete tests on their fuels.  Jet fuel has a formal process in place for certifying new fuels, while gasoline blend stocks and diesel have ASTM specifications. A more expansive engine-testing program to ensure thorough and quick assessments is highly desirable. It appears that new molecules for jet fuels are likely to have the longest approval cycles.  These regulatory issues take time, but are not necessarily high risk in most cases.

Additional approval is required to meet the RFS II standard described above, as well as the low carbon fuel standard in California, where fuels must fall below the reference for gasoline, the latter includes a controversial indirect land-use impact estimate.  In some countries, GMO organisms need special consideration for biological processes and may be particularly problematic for some GMO algae approaches if the algae are grown in open ponds or open oceans.

FDA/USDA approval for co-product use as animal feed is required. Most yeast should face lower barriers, depending on country-specific GMO regulations. As a result, knowledge of country-specific GMO issues is necessary. Approval of organisms as Amyris has done in Brazil is an example of a good approach.

Fungibility or “drop-in” features: It is best when a new fuel functions in the existing fuel infrastructure, though ethanol is getting enough infrastructure to be viable in some parts of the U.S. If the technology has multi-product, multi-market and multi-feedstock capability, it has a lot higher possibility of success. Ideal fuels are compatible with existing pipelines, storage tanks, trucks, trains, and barges.  These fuels should also be compatible with fuel pumps at gas stations and with engines.  Finally, it is good if they are compatible with conventional additives used in fuel systems -- anti-oxidants, anti-corrosives, drag reducing agents in pipelines, etc.  Some molecules are considered conditionally drop-in, for instance FAME, which works best as a diesel drop-in in warmer climates, but is not suitable for cold weather or winter due to poor cold weather characteristics. In the end, new molecules will require extensive data sets before they’re integrated into the existing infrastructure.

Given that ethanol has gained relative acceptance and reasonable infrastructure, the next new molecule (or family of molecules), unless it is very similar to hydrocarbons or ethanol, will face a bigger hurdle to acceptance. Most common hydrocarbons, on the other hand, will have relatively low risk acceptance as they are well characterized. Drop-in products will have easier access to markets. Though many pathways can ultimately produce drop-in fuels with post-fermentation processing, the more unit operations required to go from the fermentation product to the final product, in general, the higher the cost.

Tactics for “Boot-up”: The transportation fuel market is over 14 million bbls/day in the US alone, across diesel, gasoline and jet fuel. Economies of scale (and relationships with refiners with these unit operations and logistics for moving products) are not easily achieved until operating at significant scale. So companies need to have a startup strategy to “boot-up to scale” where costs/experience mature, be it through finding niche markets (DOD jet fuel contract, drop-in chemical or fuel, meeting a regulatory need, etc. ) producing accepted blendstock and meeting customers’ regulatory requirements.

Ethanol production is ~12 billion gallons/year and infrastructure is gradually being adapted to accommodate it, as production from cellulosic feedstock is expected to dramatically increase this amount. Some companies can mitigate the startup risks; in their S-1, Amyris identified a $15 billion chemicals market which can be addressed competitively with their technology in its current state.  As a result, they have an extremely large addressable market to focus on in the meantime, while they scale up and go through regulatory approval, testing and cost reductions for broader applications.

LS9 has targets in currently existing “drop-in” markets for “FAME” and “fatty acid alcohols” and has developed significant customer relationships for their initial products.  Another route is to lock-down offtake agreements with large oil refiners before building scale, to assure demand.  KiOR, by creating a crude drop-in and a transportation fuel blendstock, may also mitigate these issues since their product should be directly compatible with most crudes, blendstocks, pipelines and refineries. Drop-in chemicals or fuels will be a significant advantage, while Coskata and LanzaTech should be cheaper than existing corn ethanol, displacing an existing product with an environmentally better and cheaper alternative.

To get to volumes, refining/iterating and optimizing technology, understanding markets and the value chain is critical. Smooth “boot-up” is critical. All stakeholder positions must be understood (who benefits and who is harmed by adoption of new fuel) and value must be provided to critical stakeholders for new products to be adopted.  For example, an ethanol producer will, to repurpose a plant, require at least similar margins and other de-risking (such as long term off-takes) to switch to a new fuel.  Don’t count on a “green” premium, other than regulatory benefits like mandates and financial incentives. For this reason, a new technology that gets market acceptance or large off-take agreements up front can significantly de-risk their company’s future.

On the flip-side, there are many multi-billion dollar markets to provide room for multiple participants. Market size, either in fuels or in specialty chemicals, will not be an issue for any of these technologies. At market competitive replacement costs for drop-in chemistries, the green products will find paths to market over their fossil competitors. Unsubsidized market competitiveness with fossil alternatives, whether through KiOR’s oil, or the higher value chemicals produced by Amyris, LS9 or Gevo, is the expected outcome over the next few years.  For example, the first commercial unit (FCU) for KiOR, which will be located in the southeastern United States, should hit competitive costs at just 500 tpd, with low capex. However, if the minimum cost of a cash flow positive facility was $500 million, the financing process for the first unit would take longer, especially without strong, committed partners.   A retrofit or bolt-on facility is a good boot-up strategy that companies like Amyris, Gevo, LanzaTech, UOP and LS9 are deploying. Department of Energy or USDA loan guarantees are often the only other bet for debt for companies with high FCU capital cost needs.

Predictability of the “boot-up” is another key consideration. Companies like Synthetic Genomics and Codexis that have laid out 2015 to 2020 timelines to first commercial units should in my view be considered too early to be predictable technology or cost metrics, even with inside information, and I don't currently consider them to be viable biofuels technology providers. And large partners like Shell and Exxon will show conservatism, further slowing down development. My experience indicates that dependence on large corporate partners will sometimes delay technologies by years as they tend to be slow and overly risk averse. In the long term, I would prefer radical technologies like Synthetic Genomics.

Capital costs: In general, the capital costs should be less than $5 to $6 per gallon of “ethanol equivalent annual gallon of fuel production” capacity, and that high only if the operating and feedstock costs are about $1 per gallon ethanol equivalent or less (all costs in 2010 dollars) at a minimum.  Even then, there is no doubt it is difficult to finance first of a kind commercial units, particularly when the fuel has not yet gained mass acceptance. In the end, $3 per gallon capex on the first few plants, declining over time, would make capital formation easier.

As I have mentioned throughout, novel business plans have cropped up as a result, to exploit opportunities or pursue capital light development. For example, LanzaTech, which has targeted both biomass syngas fermentation and waste syngas, has first gone to China to build their backend fermenters onto existing steel mills, obviating the need for investing in a front-end biomass gasifier before proving the technology at scale. Likewise, Gevo decided to retrofit distressed corn ethanol assets to scale up to commercial scale more quickly and less expensively than building from scratch. This is a good example of a smart opportunistic strategy to exploit lower value corn ethanol plants.

Amyris and LS9 are taking a similar approach with sugarcane mills, but instead of just retrofitting the entire operation, they are co-investing with the mill owners, bolting their technology onto the existing operating facility and offering higher IRRs to the mill owners as a result.  KiOR can build a basic, cash flow positive commercial unit for about $100 million to $120 million and have three-year to four-year cash paybacks with well-understood oil industry processes. This has enabled the firm to finance the first commercial unit, which is now under construction with expected operation in Q1 2012.  Coskata has announced USDA loan guarantees to provide the debt for their first commercial unit.

Until detailed commercial engineering starts on a process technology for an “at scale” FCU, capex (and opex) costs may be as much as 100 percent off (sometimes more) -- many a naïve estimate of costs is floating around from startups that have never engineered a facility, often ignoring site engineering, wood yard costs, environmental remediation, waste water, transportation facilities etc. To get estimated extrapolations from a demo facility, at least 100,000 gallons a year demo is preferred to even get to 30 percent accuracy on the estimates.

Production Costs: These are the costs that will make or break many technologies, and are only really known after the process reaches significant scale. They are also the hardest to estimate and face similar uncertainties to those in estimating capital costs and have additional ongoing uncertainty if the feedstock price volatility is high.

The major production cost contributions (generally with some exceptions) are feedstock, energy, other process inputs (chemicals, nutrients, catalyst), product extraction, waste treatment, and people. Yield is ultimately a crucial parameter for cost.  Yield can be lost in the pathways, or in subsequent purification or processing steps. Regarding energy, every additional step requires more, leading to more cost.  In turn, any nasty by-products impact energy, waste remediation costs, and people.

The more unit operations required to get from fermentation product to end product, the higher the cost   In the end, if the process hasn’t yet been demonstrated at about 100,000 gallons per year scale, a company’s estimates will be subject to volatility. For fuels, operating costs (ex-feedstock) should be no more than $20 to $25 per barrel of oil equivalent at scale to reach “safer during market volatility” unsubsidized market competitiveness. An ideal target, depending upon process and feedstock should be below $15 per barrel ex-feedstock operating costs in the mid-term.  High value chemicals allow for substantially higher costs.   Feedstock costs below $60 per barrel (oil equivalent) or roughly $1 per gallon ethanol equivalent, are often needed to be competitive and should trend down with scale and time, not up (often a problem with food based feedstocks). Ideal “safe” feedstock costs should ideally be below $40 per barrel ($1 per gallon oil) but definitively below $1 per gallon ethanol equivalent.

 
Part II will focus on feedstocks, environmental issues around feedstocks, and the key questions that I believe are worth asking when looking at a biofuels investment opportunity

***

Khosla Ventures offers venture assistance, strategic advice and capital to entrepreneurs. The firm helps entrepreneurs extend the potential of their ideas in breakthrough scientific work in clean technology areas such as solar, battery, high-efficiency engines, lighting, greener materials like cement, glass and bio-refineries for energy and bioplastics, and other environmentally friendly technologies, as well as traditional venture areas like the Internet, computing, mobile and silicon technology arenas. Vinod Khosla founded the firm in 2004 and was formerly a General Partner at Kleiner Perkins and founder of Sun Microsystems. Khosla Ventures is based in Menlo Park, California.