Automation Milestones
Charting the Unsteady Rise of the Robots
TL;DR — Elon claims that there will be Billions of robots, soon. Unfortunately, that pace of adoption will be slowed by a shortage of all the other kinds of equipment, and then slowed again by a shortage of raw-material extraction-capacity. Robots won’t cause a strong decline in prices for final goods until the resource-bottlenecks are gone and robotic-automation systems handle most of their own fabrication.
Misunderstood Milestones
The Hype-sters gush about how “as SOON as we can get these robots to work, then SUDDENLY everyone will have one.” Nope. Robotic automation will trigger a number of very specific and nuanced changes to industry and the marketplace, in steady succession, and that whole process will unfold slowly. Reality-check:
Error-Rates and Viability
If I am a robot that makes mistakes 1% of the time, then every 100 events I’ll need a human to step-in and correct me. If those mistakes are costly, slow-to-fix, or catastrophic, or lead to litigation, then we won’t adopt robots while they sit at that error-rate. We wait for reliability, to get market-viability.
Looking across the full spectrum of tasks for physical labor, almost NONE of them can tolerate the rate-of-error of current robotic systems. That’s why we see such tepid adoption of robots. Yet, once that error-rate drops, then a few industries will find robots to be market-viable. When the error-rate continues to drop, adoption and demand will accelerate rapidly. Let’s look at a chart to see why:
To meet that curve of ‘demand-per-error-rate,’ then total production per year grows only slowly at first, then rapidly accelerates as error-rates continue to drop. The ‘integral’ (which is the area under the curve on our chart) shows our accumulated robotic capacity over time, and it goes through a steep ‘S-Curve’, transitioning suddenly from “nobody wants those robots, they’re too unreliable” to “everyone wants a robot, and there aren’t enough to go-around”.
The demand may transition quickly, yet the actual supply of robots will lag behind that demand, due to two successive bottlenecks:
Bottleneck #1: All the OTHER Equipment
When we just barely begin to ramp-up construction and deployment of robots, then we create additional demand for all of the other gear, sensors, and facilities that we need, to go-along-with that robot’s productive capacity. At first, that will absorb our spare equipment, increasing tools’ capital-utilization at good margins.
Yet, once robotic fabrication-rates accelerate, then we won’t be able to make enough of the tools that those robots use on the job. Instead, we will begin making substitutions, taking tooling away from other capital-intensive production lines that have lower-value outputs. That period will be brief, as fabrication of tools and equipment of all kinds ramp-up in response.
It’s important to remember, though, that buying those new robots is a sunk cost — same for all that extra equipment you have to buy, for the robots to use. Those sunk-costs normally pay for themselves years later; until then, you are burdened with debt.
So, if industry attempts to expand total throughput rapidly, then it will have spent a LOT of money on those upfront capital costs, while that capital hasn’t had enough TIME to pay-for-itself before you’re already buying even more equipment! That will increase the debt-burden on these companies, absorbing the savings that comes from automation. Prices will NOT drop yet because companies need the profits to pay their robot-roll-out debts.
Bottleneck #2: Raw Materials Extraction and Logistical Networks
Once capital-equipment has caught-up to robotic demands for ancillary tools, then the attempt to accelerate automation hits a hard constraint: there are only so much raw materials being extracted and hauled per year. Fighting for those scarce raw materials will drive-up their price. And, because it takes a LONG time for mines and processing facilities to be built, then this period of high-raw-material-prices will be the longest-lasting of the phases in our path to automation.
Competitors want to bully market-share away from each other, so they will attempt to out-bid for those scarce raw materials, which will absorb most of their cost-savings from automation; on net, total costs of final goods do NOT see a rapid decline. The slices of cost have changed: labor drops as a percentage of cost, while input-materials absorb that same portion of cost, in a push for market-share.
Only after raw-material extraction, processing, and hauling have ramped-up in response to high prices, then will robotic assembly BEGIN to lower prices of final goods. Yet, that price-drop will NOT be swift!
Gentle Price Drops make for Gently Increasing Demand
If a company lowers their prices quickly, then they shrink their margins and profits suffer. A rapid price-drop generates a large demand-spike, but capital can’t keep up! If we double our total production facilities, we still have to pay for their upfront costs, and that takes time.
So, once raw materials costs decline, and margins have paid-down most of the sunk costs of robotization, THEN we see a steady, incremental decline in prices for consumer-goods. Each company lowers price slightly, in a bid for market-share, without lowering prices so rapidly that demand outpaces capital-fabrication rates.
That pace of price-drops CANNOT get faster, until a threshold-percentage of the supply-chain is fully automated — in particular, the portion of industrial output that goes into making more automated systems. Once “robots are doing >40% of the work to make other robots” then THAT will drive self-accelerating levels of production, and the price of each robot will finally decline. Lower-cost robots feed-into more diverse and marginal-value applications for higher levels of robot-demand. That will allow all the consumer-goods industries to lower prices steeply. Only THEN will you have a robot folding your laundry.
Capital Saturation
There’s a limit to how much equipment we actually want to build, even if it’s fully autonomous. We eventually hit ‘capital saturation’, where folks see so little, marginal value from “one MORE tool” that they would prefer to spend their cash on personal pleasures! Yet, that level of Capital Saturation is actually a moving target, and it depends upon that Error-Rate, again!
See: if your robot messes up once every hundred actions, and you fixing their mistake takes some time — equal to ten of their actions, then you can only really ‘babysit’ about 10 robots, yourself. So, even if we wanted billions of robots, we would need hundreds of millions of people babysitting them!
However, if the error rate drops by just half, then each person can babysit TWICE as many robots! Most of those additional robots will be doing work of significantly lower value, so this doesn’t double our dollarized GDP. Yet, it does increase the level where Capital-Saturation occurs. We end-up building most of our robots to do really dumb, practically-worthless stuff which we have NEVER paid people to do because the work was only worth 10c per hour! And it will be many years before we get to that point.
Check-back in 2038, to see how it went!
