Advanced Computing in the Age of AI | Thursday, March 28, 2024

The Invisible Innovators: What Is the Missing Middle? And Why Do We Need Them? 

The manufacturing market is held together by a vitally-important yet often ignored segment, known as the "Missing Middle." The term refers to the roughly 300,000 manufacturers in the US (two million globally) that employ fewer than 500 people. Economically, this group is responsible for more than twice as much employment as all the "big" manufacturers combined and drives 60 percent of R&D investment.

Dark Matter Hubble Deep FieldDo you like astronomy? Of course you do. Everyone does. Doesn't matter whether you actually understand the math or just enjoy looking at the pictures, space is cool, and weird. And now we're learning that there's a lot more stuff in it than we know about — stuff we can't see, or touch, or interact with in any way. Dark Matter, assuming the latest theories are correct, makes up the vast majority of our universe. It's there, but invisible. And yet it holds galaxies themselves together.

The theory of Dark Matter has plenty to do with manufacturing, because that too is held together by an all-important, invisible, ignored segment: the "missing middle." That term has been thrown around a lot in the past few years, but really, who are these companies? Why are they so important?

Manufacturing's dark matter is the roughly 300,000 manufacturers in the US (two million globally) that employ fewer than 500 people. They are the neck that turns the head of industry. All the Boeings, GEs, Lockheeds, P&Gs; all the automakers; all of them depend on the missing middle for their operations and, arguably, their very existence. The missing middle is the foundation of production, process, design and innovation across every manufacturing sector around the world.Missing Middle Employment

Economically, it is responsible for more than twice as much employment as all those "big" manufacturers combined, the ones you've heard of, the ones with more than 500 employees. Developmentally, 60 percent of R&D investment comes from the missing middle — up more than 40 percent since 1983.

In the past 30 years, R&D investment in the larger companies has plummeted — from 72 percent of total in 1981 to 40 percent in 2007. (Source: National Science Foundation, Science Resource Studies, Survey of Industrial Research Development) The Bell Labs and Xerox PARCs of old are gone; pure research falls before corporate belt-tightening. But innovation can't stop, so the larger organizations that once dominated the R&D landscape are pushing innovation down on their suppliers. The smallest enterprises now shoulder the burden of innovation.Missing Middle Innovation-Chesborough

The missing middle is functionally capable of driving R&D — after all, it has for years now — but it lacks the infrastructure, investment and talent to maximize its opportunity. Industry as a whole is moving toward a 21st century digital paradigm, and the missing middle, the new torchbearers of innovation, need those 21st century tools to compete.

Assets and Liabilities

When we talk about digital manufacturing in terms of what the missing middle needs, it comes down to high performance modeling and simulation. Armed with this tool, missing middle manufacturers will be able to innovate faster and with less risk than traditional — soon to be antiquated — methods of R&D and design. Translated into the world of digital manufacturing, the whole of high performance computing can be simply defined by the number of users and the range of task complexity.Missing Middle Task Complexity Chart

At the low end of task complexity resides the vast number of computer users. They work at nearly every organization in the US. For this group, the desktop computer is king, and is used for such things as home and office basics, Internet, email, gaming, graphic design, and so on. While the missing middle certainly has millions of such users, this portion is well-served and does not need help.

At the very high end of task complexity reside the advanced research institutions: major university centers such as OSC and NCSA; national labs like Argonne, Oak Ridge and Sandia; the Department of Defense's six national HPC centers; and various other governmental agencies such as the IRS, NSA and the CIA. These users are who most people think of when they consider the application of high performance computing, but in truth they represent only the upper limit of task complexity. The infrastructure offered by these institutions could represent significant capacity and capability if we can improve on the existing ways to leverage them for US industry.

In between these two sets of users is an enormous void where many users are underserved by the tools they currently apply to their complex tasks. These users could be benefiting from HPC, but they are not. The missing middle, those companies with fewer than 500 employees, generally fall into this category.

Why?

Awareness, for one thing. Many of these small and medium sized organizations either do not realize the need for HPC or fear a perceived initial investment. Many also fail to understand how the power of HPC could be applied to their business activities.

The issue of awareness can be solved with comparative ease; the three major detractors are the availability of talent, investment, and infrastructure at this level. Missing middle manufacturers may or may not have the personnel to use HPC in place; they almost certainly lack the cash to invest heavily in their own, dedicated HPC network; and, being small, adopting such tools under their own roof is impractical. Addressing these three issues is the challenge of bringing digital manufacturing to the missing middle, the group that needs it most.

In Your Face, Tradition

Not all small and medium manufacturers have been left in the dust, of course. L&L Products, a custom chemical compounding firm located in Romeo, Michigan, foresaw the advantages of digital manufacturing and adopted it early on. With 99 percent of L&L's customer base firmly in the automotive industry, any tool to streamline their processes and give them a leg up on the competition would be welcome. In 1998, the company realized there was captive innovation in the form of new epoxy based materials that it could not fully capitalize on without the power of modeling and simulation.

The company had hit a ceiling: infrastructure costs were competing with turnaround time, putting L&L in the unenviable position of turning customers away due to simple lack of bandwidth. Increasing overhead costs was one option — continue the existing model and cut into the bottom line by buying more speed in the form of old-school computers and software — but stakeholders realized that the traditional approaches were simply too slow and too error-prone. L&L engineers were unable to validate models for their customers and could not design parts or materials that would result in optimal, cost-effective solutions using the old methodologies. With this in mind, the company took a financial gamble, investing in a four-core HPC system and associated tools.

It was a scary move. The computer itself wasn't cheap. Add in software and personnel capable of running it, and suddenly L&L was taking on a potentially disastrous risk.

The results, however, silenced any lingering fears. Within five years of adoption, L&L leapt from 300 to 600 employees and increased its annual sales by 57 percent, $150M to $260M. They also increased their HPC system, migrating to an ever growing cluster in order to keep up with customer demand.

L&L could have outsourced work to China, or made massive investments in its existing, 20th century infrastructure and methodologies. Instead the company went for it, adopting a different, innovative solution. And given its resounding success, people began to take notice.

In 2008, the National Center for Manufacturing Sciences (NCMS) approached L&L Products about joining a novel collaborative R&D project, funded by the U.S. Department of Energy under its Lightweight Automotive Materials Program (LAMP).

Teaming with R Systems, Nimbis and Altair Engineering, this project sought to evaluate innovative methods of accessing significant HPC resources using a new costing model based on usage rather than hardware and software licensing. By leveraging existing HPC infrastructure available at the very high end of the task complexity spectrum, companies could pay by job or core-hour rather than buy everything themselves. The project, slated to complete in September 2012, has already proven the model and returned impressive results, spurring the team to continue its collaboration. Next step goals include incorporating greater automation and practically applicable, real-world applications suitable for missing middle manufacturers.

Companies like L&L Products took the plunge nearly a decade ago, and are reaping the rewards. But most of the missing middle remains as dark matter: there, invisible, ignored, holding the entirety of global industry together. If US manufacturing is to surge ahead and regain its position as a dominant force in the global economy, the missing middle must have access to these tools; it must have access to training and awareness that will build a drive to adopt the new methods; and it must have strong support to break down the barriers — both real and perceived — that have so far prevented wide-scale adoption. Those companies that do not leverage this technology will take longer to design more at greater cost and with less guarantee of success.

There will always be the stubborn individuals in many companies who insist that HPC is only good for pure science, or who still believe that an immense investment is required to build high performance computing Dark Matter Eagle Nebulacapability. This misapprehension feeds into the ongoing challenge of adoption — if HPC is to become mainstream, we need to shatter these misconceptions and make it possible for collaborators to access HPC tools without the need for massive investment up front.

This mysterious missing middle is responsible for twice as many jobs as those manufacturers with more than 500 employees. It drives 60 percent of global innovation. Those who continue to ignore its needs do so at their own peril. Overwhelmingly, companies within the missing middle say they need new tools and processes. Like Dark Matter, the missing middle doesn't care if it gets the credit, doesn't care if the world knows it's out there, the glue holding manufacturing together. The missing middle wants the tools to do its job.

EnterpriseAI