Startup Strategy for Commodity Products
Making ultra cheap products requires different rules and strategy.
The World of Hard Tech
"Hard Tech," startups building new hardware technology, has blossomed over the last 10-15 years. Companies like Tesla and SpaceX are at the vanguard, becoming fabulously valuable while changing the world around us. New waves of startups are following, clustered in place and culture in enclaves like "The Gundo." Defense and aerospace are especially popular sectors. The culture is a modern version of manifest destiny - techno-utopianism is our destiny, and America is the place to do it.
The process tends to look like lots of fancy vertically integrated software and hardware with rapid iteration and testing.
These products tend to be complex and valuable. A Tesla Model 3 sells for ~$20,000/ton. A Falcon 9 sells for ~$80,000/ton. But there are lots of things the world could use new processes for that cost $100-$2000/ton: cement, steel, aluminum, hydrocarbons, plastics, etc. The 10x-100x price difference imposes different constraints and creates new opportunities. I'll call startups pursuing these markets "commodity startups."
The Harsh Mistress of Profit
Hard tech startups usually take commodity materials and produce something 10x-100x more valuable. Getting it to work at all usually leads to success. Commodity startups are nearly the opposite. The products are relatively simple, and producers can make them in many ways. Making the numbers work is more important than creating the product.
This post focuses on a strategy of pursuing simplicity to reach cost competitiveness.
How Commodity Startups Work
Understanding the Algorithm
Elon Musk has an algorithm for engineering popular among hard tech startups:
Question Requirements and Every Requirement Must Be Tied to a Human
Delete Parts/Processes
Simplify/Optimize
Go Faster/Speed Up Cycle Time
Automate
Commodity startups need to use Elon's algorithm with a different emphasis. Steps 3, 4, and 5 almost become afterthoughts because the only way to reach profitability is to ruthlessly prune requirements and delete so many parts that there isn't much left to optimize or automate. And it's so simple that fast can become an integral feature.
Requirement management and deletion matter because these commodity products tend to sell in the millions or billions of tons, and someone has tried almost every alternative within the current framework. The constraints imposed by the low price are extreme. The physical limits dictate that there will only be 1-2 ways to win, and they often are not obvious.
It can't be stressed enough that one errant requirement could increase costs by 10x or block a new angle that could vastly simplify the problem.
Holding A Program in One's Head
Paul Graham has an essay titled "Holding a Program in One's Head." The gist is that the more of a program you can absorb, the easier it is to optimize and the more productive you will be.
A similar idea is global optimization versus local optimization. Changing one part of a design or program without considering how it interacts with the system can wreak havoc. Conversely, considering the entire system can unlock opportunities, often much greater in magnitude, that aren't obvious at the micro level.
The point of steps 1 and 2 of the algorithm is to understand what the system needs to do at the bare minimum and delete all the requirements, parts, and processes that aren't contributing. Naturally, this can take many iterations, trips down blind alleys, and many steps back to go forward. Holding the whole program in your head makes it much easier to do.
Graham usefully provides the conditions where this works best:
Avoid Distractions
Work in long stretches
Use succinct languages
Keep re-writing your program
Write re-readable code
Work in small groups
Don't have multiple people editing the same piece of code
Start small
And we can see that this mostly means a few very productive people working hard. Having too many workers or dividing the work too much can be counterproductive for finding that ultimate solution. Too many workers also increase the burn rate, reducing the time available.
Most problems in this space seem daunting, but continuous exploration and refactoring can cut them to the bare bones and unlock new, non-obvious solutions. Then, it is easier to scale and add new employees and capacity.
Taking Advantage of Orthogonal Opportunities
The old VC adage of "Why Now?" still applies. Any new process competing against a scaled incumbent has to cheat to win. Solar PV is now cheap and might help out energy-intensive commodities. Some products, like Portland cement, have weaknesses. Half of the limestone mass that cement producers spend money and effort mining goes out the stack as CO2 instead of product. Maybe there is a new performance-based standard for a product. Or a new processing technology has matured.
We can probably summarize most opportunities into:
Input price and availability changes (energy, materials, transportation, equipment, technology, labor, or capital)
Removing waste/inefficiency
Regulatory/standards/taste changes
Timelines
Many technology markets have race aspects where delivering a product first is extremely valuable, and many teams can get across the line given enough time. Commodity startups tend to be the opposite. The question is if any new technology can succeed. It can take decades for the technology to develop and deploy in the worst cases.
Cement (~$100/ton) is one of the most challenging markets to crack. Startups say things like: "We can be cheaper than traditional cement plus carbon capture" or "The government buys half of the cement, and it is only 10% of the cost for roads and bridges, so ...."
Fortera is a startup that is actually trying make money by using the CO2 byproduct to produce an alternative cement recipe. It originally started as the carbon capture company Calera in the early 2000s. The original CEO quit, it was defunct for four years after 2015, restarted as a green cement company, and is now working on contracts to deliver millions of tons of material.
The option value of time is helpful for commodity startups because inputs might continue to improve, someone might develop better ideas, or markets might change. Concepts should be able to grow fast after they pass a viability threshold, which we've seen with electric arc furnace steel, shale gas, and shipping containers.
Do Things That (Already) Scale
Scale is a massive disadvantage for commodity startups. Many technologies require billion-dollar facilities to reach competitiveness. Avoid those situations. Finding niche markets is one approach, but they are often scarce in commodities, or the price is high because of the difficulty (like serving customers in rural Alaska).
Some techniques offer scale and flexibility out of the box, which can allow small incremental CAPEX deployment. Construction and large steel assemblies are two examples. These technologies have extremely low idiot indexes, even on custom projects. Shipbuilding, chemical plants, and industrial-style construction or earthworks are examples of their efficiency.
Factories can be the wrong choice of scaling architecture and are over-indexed because of solar PV and battery learning curves. A few kilograms of solar semiconductor material in a solar panel can displace 10,000 kilograms of coal over its useful life. Lithium-ion batteries are uniquely efficient, energy and power-dense, and have a minimal balance of plant. The performance of these technologies justifies a relatively high price and factory assembly.
Most commodities are one-time use and 10x less expensive. Trying to build a giant factory with its upfront capital and engineering that isn't the final product adds risk. Road width, height, and weight restrictions also limit scale and design flexibility for factory products.
A site-built facility can be much larger than a factory-produced machine with no initial factory overhead and can use scaled technologies like steel and bulldozers. Chemical processing or desalination facilities produce the cheapest products of any manmade processes, and they are built onsite with mostly special-ordered components.
The Simple/Complex Barbell
There seems to be a barbell in commodities technologies. Winning technologies are on the extremes of simple and complex (or both extremes coexist). Middling technologies often aren't not good for much.
Information processing is an example where complexity is dominant. Computing devices like slide rules struggle to compete against the absurdly low cost of transistors produced in a chip fab.
The simple Ordinary Portland Cement process dominates cement production. A competitor must have comparable simplicity because there is only marginal performance to gain.
Technologies coexist in grid-scale electricity storage. Pumped hydro storage is a classic simple technology. It uses two water reservoirs at different heights to store electricity. Cement, steel, water, and earth are its primary materials. Pumped hydro reservoirs are constructed with highly optimized construction equipment and techniques. On the flip side are rapidly growing deployments of lithium-ion batteries, which are the highest performance of battery chemistries. We would only use pumped hydro for electricity storage if water was denser, large cliffs were everywhere, or gravity was stronger. But, in our universe, lithium-ion batteries are ~100x more energy dense than pumped hydro and fill many gaps that pumped hydro cannot.
There are a few takeaways for when one extreme will win:
Physics Matters
Basic physics handicaps many technologies. Compressed air is a potential "simple" competitor in electricity storage. Air is a compressible fluid that makes increasing pressure significantly more expensive and less efficient than an incompressible fluid like water. There is barely any compressed air electricity storage versus hundreds of gigawatts of pumped hydro.
Minimizing the Balance of Plant
Both simple and more complex technologies still need simplification and optimization to succeed. The ultimate goal is to have only the magic materials/parts in the product and no excess. Lithium-ion batteries have had incredible success reducing the amount of non-active material. Lower-performance competitors, like flow, metal-air, or molten salt batteries, struggle to remove kludge.
Opportunities for Dematerialization or Process Removal
Some technologies allow for extreme dematerialization, like computer chips or solar panels, and these are prone to go the complex path. Similarly, onerous processes with many steps are vulnerable to collapse by processes with fewer steps.
Some technologies have much better potential than others. The winners will likely be on the extreme ends of simplicity or complexity. The features of any application matter for which side might win. Being an advanced abacus maker in the face of competition from integrated circuits isn't a great idea. Nor is creating a Rube Goldberg machine to produce steel or cement. Simple technologies are more interesting (to me) and seem underexplored compared to complex ones.
When Idealism Hits Hard Reality
Trying to attack a problem through simplification, cheap materials, etc., is not a novel concept. The real world is unforgiving, and issues immediately crop up. Reliability, soft costs, material purity, side reactions, quality, and balance of system bloat are common issues.
The pressure to take on more complexity is extreme. The burn rate is always hovering. Team members specializing in a path favor continuation. "A little more" complexity might make it actually work, and it can always be costed out later. The "technical debt" comparison from software is good. Many hard tech startups can absorb some technical debt, but the budget is often near zero for commodity startups nearing commercialization.
There needs to be a push to stick to pursuing simplicity, which often means seemingly lateral or backward moves to make the puzzle pieces fit. Higher burn rates and more employees vested in specific paths increase the pain of finding a good solution.
Capital, Business, and Growth
Market Niches and Business Models
There are two main possible strategies.
The first is rapidly scaling in-house production. In-house scaling works well when the technology is capital-efficient and further returns to scale are available.
The other is to move to a services-based model. Selling high-margin services to others who want to use the technology becomes attractive if the technology is more capital-intensive, has a massive market that will take time to conquer, is difficult to patent or keep secret, or may already be reaching diminishing returns to scale.
The strategy starts by building several early commercial projects to prove the technology and unlock debt finance. The company can market design and construction management services once it becomes plausible to sell to run-of-the-mill developers or be a developer. These services will be a small portion of the project cost but add significant value from risk reduction and ease of finance. The provider can charge a healthy margin that makes the package attractive to developers but challenging for new entrants to justify a development, testing, and validation phase. A technology like pumped hydro storage is an example of one that would require this playbook.
Another consideration is market segmentation. There may not be a viable high-price entry point, but there are often markets that are much easier to satisfy customer needs than others. Steel mini-mills started with beams and rebar, only later graduating to fancier products. Usually, this niche minimizes capital expenditure and allows the deletion of many subsystems.
The ideal scenario is finding a market segment that the new technology can handle and tackling it with capital-efficient early facilities. Market share can expand through continued scaling and/or service offerings while eating more difficult market segments over time.
Assessing the Capital Stack
There are several financing options due to the depth of US capital markets.
The most obvious are the classic VC dollars or grants early-stage companies use.
A nice feature of building facilities that produce the commodity directly is that product liquidity and long-term off-take agreements make financing more practical. The primary goal becomes "bankability" to reduce the need for dilutive capital. A development process that focuses on simplification can make bankability easier by reducing the number of items that can fail and can make the system easier to evaluate for outsiders.
Bankers might finance a project once it is complete and generating revenue even if it isn't de-risked enough to finance pre-construction. Other options include selling equity in the project rather than in the parent company, a practice common in the real estate and oil and gas industries.
A rough order of operations is VC -> project equity -> project debt refinance -> upfront project debt in whatever combination is necessary.
Verticalization, Centralization, and O-Rings
Many hard tech companies lean heavily towards the o-ring spectrum of business, especially in aerospace. That encourages significant vertical integration and centralization of authority to ensure a good and reliable product. These choices bring a lot of (necessary) overhead.
Commodity startups usually aren't severe o-ring businesses. Mistakes can happen, and there is time to correct them. That can create an opportunity for less overhead, especially for construction-oriented technologies that don't have massive intermediate factories.
Intellectual Property?
Commodity startups might not produce a massive patent portfolio because it is challenging to patent deleted items. Pursuing many patents can be extremely harmful if it detracts from simplification and global optimization. I've seen real-life examples where a startup rejects a hard-to-patent but likely to work technique to pursue unique processes. Then they have to come back to it after wasting time and money. Use the best available methods and patent where it makes sense.
It is easy to underestimate how challenging execution is. A competitor won't know the many details that led to a specific design, making it harder to copy and increasing their risk of errors and failures upon deployment. There will be time before competitors can catch up.
The goal should be to win and find the lowest energy state possible. Any flourishes are just costs that prevent positive unit economics and scale while allowing future competitors an opportunity to prevail.
Founders vs. VCs
One seemingly innocuous detail deep in the exploration can kill a commodity startup's concept. Many of these details should pop up, leading to further iterations and adjustments to find the low-energy state.
Venture capitalists rarely have the expertise or time to research and anticipate these roadblocks. The classic reliance on having 1-2 winners return the fund remains.
The founder's bet is different because they are staking a significant chunk of their life and reputation in one startup. Just because you can raise a lot of money for the idea does not mean you should. Again, few VCs care about a deep technical analysis because of the cost. Founders ideally have the experience and ability to do deep techno-economic exploration quickly to filter viable ideas.
It can be extremely challenging to quit after many promises are made, especially for a founder who likely has previous high-status achievements (prestigious college, etc.) and no failures. Extreme caution should preceed any promises, hires, or other actions that create barriers to necessary architectural changes. The pressure can easily convert a founder into a Randian villain because lobbying for subsidies or kneecapping competitors is the only path to success if technical barriers arise that prevent profitability.
Flexibility and deep investigation are critical for maintaining sanity and not becoming a villain.
Growth and Dilution
There are probably dozens of software-oriented companies worth more than the market capitalization of all the players in a vertical such as cement, chemicals, steel, etc. Founders need to own more of the company to make the risk worth it. A healthy return on capital means the founders might own a valuable piece of the company instead of seeing their stake diluted in endless capital raises until it isn't worth continuing.
Real Life Examples
One of the best examples of this playbook is steel production with electric arc furnaces (EAF). The first electric arc furnaces were available in the late 1800s but were not practical outside of niche cases because of electricity costs and electrode cost/quality. By the 1960s, electricity prices had fallen relative to coal, labor was more expensive, steel scrap was more available, and electrodes had become practical. Nucor was the first in the US to build a "mini-mill" in South Carolina, and today ~70% of US steel production comes from electric arc furnaces.
EAF ultimately prevailed because a mini-mill is much simpler than traditional integrated blast furnaces, requiring less labor and allowing a much smaller minimum scale. The first US mini-mill cost ~$50 million in today's dollars, a fraction of a big mill. The lower labor and capital burden makes their production much more flexible if prices are low (the industry is cyclical). The first mills were basic but grew in size while gaining capabilities like continuous casting and sheet production. In the early 1980s, Nucor's return on capital was 25%-30% as this scale allowed them to eat more of the market at high margins.
Nucor's culture is also unusual, especially before long-serving CEO Ken Iverson died. The company was lean and flat. It had 22 corporate HQ employees out of >7000 total and only four layers of management in the late 1990s. Their impressive technical accomplishments came without any corporate engineering or R&D. Several features stand out:
Organizational Simplicity
The flat organization chart meant sites ran themselves and had free reign to optimize their operations to fit customers' needs. The corporate-level data gathering was very minimal, focused on the highest signal reports.
Traditional steel companies were the opposite of this, and most are now defunct.
Smallness
The company believed in the power of small groups, which was why each site was autonomous. Many engineering projects or new facility builds were ad hoc and completed by small groups working on them in addition to their regular jobs.
The smallness and other cultural features like flat organization charts helped avoid unionization, which was a massive drag on traditional steel makers' pace of innovation.
Risk Taking/Employee Autonomy
Iverson loved weird experiments and was constantly pushing sites to try new things, even if they were often failures. Perfection was not a goal. One of his sayings was: "If it's worth doing at all, then it's worth doing poorly." One example he gave from his pre-Nucor experience was building a pipe casting machine for $5000 compared to buying one for $150,000 (it later fell apart during a demonstration for the board of directors, shooting a molten metal dart through the walls, scattering the crowd).
Simplicity, smallness, and risk-taking were more feasible because steelmaking is not an extreme o-ring business.
Utilization of Vendors to Reduce Overhead and Cost
Nucor leaned heavily on its vendors to complete projects. Vendors could build what they needed since 99% of process equipment is custom-built/ordered. The real job of the Nucor employees was to use their day-to-day knowledge to get the correct specifications for the purchased equipment. They relied on the experience of vendors to complete the project details. Typically, this method means increasing time while decreasing capital and labor usage.
One of Nucor's main mantras was "Build facilities economically and operate them efficiently." A small team approach to projects and extensive utilization of vendors helped them do that.
Willingness to Improve and Invest
The bureaucracy of Nucor's competitors rendered them unable to make many investments. One example was when Nucor entered the fastener business, bought new equipment, and was able to produce 400 pieces per man minute. The traditional steel companies could only do 50 pieces per man minute on their old equipment and closed facilities instead of investing.
The Rolling Mill Saga and Time
By the early 1980s, Nucor was close to saturating markets like rebar. It needed to make steel sheets to grow, but these took absurdly large and expensive facilities. One possible answer was thin slab casting, which the traditional steel makers had written off as impossible. One Nucor site tried to build a machine, but it failed miserably. They worked with several potential vendors, including one from Germany, as part of the exploration. Years later, the German vendor got a working prototype and approached Nucor because Nucor was the only company crazy enough to try it. The rolling mill eventually came online in the late 1980s with much success. It took seven years for a competitor to field a similar mill, and by that time, Nucor had made many improvements, maintaining their lead.
Time and persistence paid off to finally reach commercial deployment.
Conservative Finance
20%+ return on capital means financing can be basic. There is a flywheel effect where trying new things becomes easier because the projects are funded internally or backed by existing assets instead of new project income.
Other successful examples might be shipping containers and shale gas/fracking.
The number of cases is relatively small because there are only so many massive commodity markets, and it is challenging to find these breakthroughs.
Conclusion
Breaking into commodity markets is incredibly challenging but not impossible. Simplification, requirement pruning, and careful adoption of technologies are key to achieving positive unit economics and reasonable incremental capital expenditure. Small initial teams are critical for agility and sampling large portions of the available design space. The extreme cost pressure can cause time to yield more than in higher-value hardware technology startups.
Healthy returns on capital are possible despite the challenges because successful technologies tend to be simpler than incumbents, more productive, utilize already scaled techniques, or allow for more direct financing.
The commodity startup space has been relatively quiet for ~100 years, but rapidly changing energy, transportation, and materials technologies could spark new action.