Ken Crutchfield: Lessons From Steel — Why LLMs Will Become Commodities

Ken Crutchfield argues that large language models will follow steel's 19th-century path from scarce breakthrough technology to abundant commodity, shifting economic value from model builders like OpenAI and Anthropic to the businesses and legal professionals that apply the models. Crutchfield draws the parallel to Andrew Carnegie, whose involvement in the Eads Bridge across the Mississippi River in the late 1860s pushed him into steelmaking after the Bessemer process made large-scale production far more economical. Crutchfield points to distillation and open-weight models, plus more data center capacity and more efficient compute, as forces already reducing LLM scarcity.

Large Language Models LLMs are a breakthrough technology, helping fuel the AI industrial revolution. OpenAI and Anthropic are market leaders, with extraordinary private valuations. The prospect of their respective trillion-dollar public company valuations, should their initial public offerings come to fruition, has drawn attention to LLM economics. LLMs are transforming industries including healthcare and the practice of law. They will also create entirely new industries. Geopolitical considerations also surround LLMs, as stronger models have economic, cybersecurity, and national security implications. History offers valuable lessons for understanding the future, and some parallels exist between this AI explosion and the 19th-century Industrial Revolution. What if LLMs become commodities, like steel? Becoming a commodity means that something becomes more abundant and prices drop. Like steel in the late 19th century, perhaps LLMs may become commodities, too. For legal technology businesses and the broader legal profession, this matters a lot, as both would benefit greatly from commoditization. Economic value would shift from the model builders to the model users. That doesn’t mean every model becomes the same. Different grades of steel serve different purposes, and even as commodities, LLMs can and will vary in strength. Let’s look at the historical parallel in the steel industry’s development. Carnegie and the Eads Bridge Andrew Carnegie worked for the Pennsylvania Railroad and then the Keystone Bridge Company. In the late 1860s, Keystone won a contract to help build the Eads Bridge across the Mississippi River in St. Louis. The engineering requirements were substantial given the river’s width, winter ice flows, and the depth of drilling required to reach bedrock. Other obstacles included lobbying by steamboat owners for high clearances that would let them keep operating. The project was daunting, and James Eads, the engineer and namesake of the bridge, insisted on steel for key structural components. This pushed Carnegie into the emerging steelmaking business. The Bessemer process was developed in Britain and made large-scale steelmaking far more economical. Carnegie later championed it in the US as he built his massive steel empire. Steel helped make projects like the Eads Bridge over the Mississippi possible. Its physical properties helped make skyscrapers practical, and high-rise buildings in cities began to alter urban skylines. Steel improved railroad tracks, supported the development of automobiles, and became critical to machinery, ships, weapons, and almost every major industrial sector. Like LLMs today, steel was a defining technology of its era. But steel was also a commodity, and competition reduced margins. The US Government established tariffs to protect the industry and help it grow. Historically, prices have been volatile within the industry. Across administrations, the American steel industry has benefited from loans and subsidies to protect it from foreign competition or bankruptcy. Steel vs. LLMs Steel production became an abundant, general-purpose input into industrial manufacturing. Carnegie made a fortune, but so did Ford, General Motors, appliance makers, the construction industry, defense contractors, and others. Downstream businesses added value by applying steel to solve problems, creating higher-margin businesses. Consumers who used those products also benefited. LLMs are perceived as scarce and expensive resources, but techniques like distillation and open-weight models are beginning to reduce the scarcity. Add more data center capacity and more efficient compute, and we could quickly reach a point where LLMs are abundant, and choices are broad, including relative quality just as with steel . LLMs may become a general-purpose input into SaaS applications. In legal tech, every feature advancement of the “so-called” wrapper businesses makes them more valuable. And if competition drives down the cost of LLMs, legal tech may look more like the automotive industry with better margins and greater value than the raw material inputs. See Bill Henderson https://www.linkedin.com/in/wdhenderson/ ’s modern classic https://www.legalevolution.org/2021/04/the-best-metaphor-for-todays-legal-market-is-the-auto-industry-circa-1905-231/ piece on that analogy. If LLMs become commodities, legal tech will benefit, and the broader legal profession will benefit. Better products versus better steel At the turn of the century, industry wasn’t asking for the best steel. There were problems to solve and possibilities. Steel helped solve problems by enabling taller buildings, sturdier automobiles, and more reliable ships. Similarly, attorneys aren’t sitting around asking for better language models. Attorneys want solutions to legal problems. They want stronger legal arguments, more settlements, faster dealmaking, and better contract terms. They want clarity and better tracking of regulatory changes to advise clients better. The most powerful model won’t always be needed, and specialized LLMs may be just as effective. Across industries, smaller, specialized, open-weight, or privately hosted models may be more than adequate to solve problems. Protectionism No analogy is perfect, and the steel-to-LLM comparison has limits. One example is protectionism. A government can stop a shipment of imported steel at a port and impose a tariff. Controlling software or LLMs is more complicated. Business is global. Servers can be located anywhere. Digital offerings are more porous. Countries have sovereignty to choose models built within their country or from new trading partners. For example, American protectionism doesn’t stop businesses in the EU or Canada from using distilled Chinese models. Competition may force OpenAI and Anthropic to lower prices or adopt distillation techniques. The market might be disrupted by open-weight models offered through open-source licenses. That introduces margin pressure, which brings us back to their respective long-term prospects. These businesses may have wildly successful IPOs. But like steel, could they transform industries and become strategically important to the US, while also becoming commodity providers that, in the long term, require protection or subsidies to stay afloat? I think we should consider that possibility. If that happens, the biggest winners will be the downstream players within industry verticals like legal. The so-called “AI-wrappers” could become the higher-margin businesses, and the legal industry could be one of the winners. AI was used in the creation of this article Ken Crutchfield, founder and CEO of Spring Forward Consulting https://springforwardconsulting.com/ , has over 40 years of experience in legal, tax and other industries. Throughout his career, he has focused on growth, innovation and business transformation. His consulting practice advises investors, legal tech startups, firms, and others. As a strategic thinker who understands markets and creating products to meet customer needs, he has worked in start-ups and large enterprises. He has served in General Management capacities in six businesses. Ken has a pulse on the trends affecting the market. Whether it was the Internet way back in the 1980s or Generative AI, he understands technology and its impact on business. Crutchfield started his career as an intern with LexisNexis and has worked at Thomson Reuters, Bloomberg, Dun & Bradstreet, and Wolters Kluwer. Ken has an MBA and holds a B.S. in Electrical Engineering from The Ohio State University.