Different this time? Tech Jobs (2025~) vs Manufacturing Jobs (1980~)
A shocking indeed job trends graph got me thinking about how the present assumed crash in tech jobs compares to the crash of manufacturing jobs due to offshoring that started in the 1980s.
LinkedIn is awash in tech-related people looking for jobs and complaining about how horrible the market is. Mainstream news regular reports about the large retrenchments of tech companies. The general coffee shop gossip is that things are “terrible” and “the end of the world is coming”. Tyler Cowen recently included a link to this graph in his Marginal Revolution newsletter.
The graph is stark and plays well into the vibe/feelings of the now. I forwarded it around to a few of my friends, and we got into some nice chats about the end of software engineering as a whole. And, I got to thinking “isn’t this like the massive offshoring that happened in the USA in the 2000s?”
I started researching a post that would show how the offshoring of the (corrected) 1980s, 1990s, 2000s and 2010s related to the softening in tech jobs that we are seeing. Yet, that isn’t what I’m writing about today. After digging through the USA Labour Statistics about different sectors, I increasingly feel that this time will be different even if some of the underlying factors are similar.
Data Note: I used USA data since the Indeed chart comes from USA data and that in general the USA economy is a bellwether for the rest of the world. I’d be happy to run this analysis on any other country if similar data is available.
What is similar?
There are some similarities between the dynamics of today and the 1980s offshoring boom. GenAI and other AI-driven automation and availability of massive computing power have enabled the automation and ‘outsourcing’ to computers of skills not previously possible. The opening of China, free-trade agreements and many other factors from the 1980s onwards did a similar thing with manufacturing. Suddenly, USA companies could build the ‘same’ product somewhere else for a fraction of the cost. Today, companies could build the ‘same’ report, app, picture, video, process in a computer for a fraction of the cost.
Offshoring constraints
There are also similarities in the constraints. Offshoring struggled for ~40 years and even struggles today. It isn’t as easy as just picking up manufacturing equipment/skills/delivery, put it on a boat and deliver the same product at a fraction of the cost. Offshoring specialists are quick to highlight the long list of constraints on any successful project (few examples):
What is the value to the move other than the cost of labor?
Is the supply chain in place?
Are you able to invest in the approach for years before seeing equivalent quality and volumes?
Can you protect yourself from copy-cats?
Is the origin of your product core to the brand-value?
…
It’s a long list. Any search on Perplexity will generate hundreds of supporting articles and how-to guides to make your offshoring project successful or highlighting the risks and reasons not to offshore.
AI-Automation constraints
Also a list that is very similar to the constraints on any GenAI/AI project you are considering:
What is the value to the automation beyond just doing it ‘cheaper’ than a human?
Do you have a supportive supply chain of data in place?
Are you willing to build the automation, iteratively over a long period before seeing equivalent levels of performance?
If this process is easy to copy, can you buy it from someone else? What is your moat to building it yourself?
Is an in-person or real human interaction core to the brand-value?
…
The list can keep going but I think you get the point. There are a lot of overlaps and similarities between considerations around ‘offshoring’ and ‘AI-driven automation’.
What is different?
As I researched this topic and my own desire to draw this parallel, I realized just how much is different about these two challenges. First let’s look at the data.
The impact of offshoring was slow, long
The first thing that struck me looking at the USA data was how long the impact of offshoring - it took almost 20 ~ years before job losses were really locked in and employment collapsed. The same graphs from FRED for easier comparison.
The impact of these job losses as they hit was likely much worse than job losses in tech as manufacturing jobs tend to be location, technology and skill dependent. If you are a manufacturing engineer for a specific product line that gets offshored, chances are all the plants in your community are going through the same thing and suddenly a lot of people are competing for the same jobs all around the country.
I haven’t done the research, but the popular belief is that the USA Government also failed to provide sufficient support in worker protections, reskilling, upskilling to buffer this shock. From my own work in the difficulties around rekilling/upskilling, this would have been a massive economy wide, complex challenge.
The ‘collapse’ in tech jobs isn’t real
Remember that horrible finger-graph at the top? Turns out it isn’t real. This is the sector-wide job statistics for the USA that includes the whole tech sector. While not perfect it's likely the best side-by-side comparison with the manufacturing sector graph.
Taking a step back, this makes sense. Even as certain roles within tech are declining, the sector as a whole is continuing to grow. Even without the growth of GenAI, it is possible to make the argument that the Indeed data is a Covid-driven hangover. We know that companies massively over-hired tech/engineering roles during Covid due to the sudden demand in tech-driven solutions. That collapse could be mostly tied to the economy normalizing and demand for software engineers returning to ‘normal’. While the entire sector as a whole continues to grow.
The challenge a lot of people face is that they are part of that “lump” in the Indeed graph. Most people who entered tech jobs during the last ~10 years are likely networked with and friends with other people who did the same. So it is easy to get caught up in the community of people who are directly impacted by this adjustment. And it can be equally difficult to see beyond that community especially when the general market growth is likely uneven across a wide range of industries, roles and categories across tech. Yet the growth was highly concentrated in a few large, dominate firms that benefited from Covid and pre-Covid.
Reasons for optimism
Going through this research and reflective post, has given me more reasons for optimism than I had before writing it:
Rationally, the tech sector as a whole continues to grow.
Tech-related roles seem to have more transferable, trainable skills than manufacturing jobs in the 1980s. This increases the competition but also should soften the impact of AI-Automation as people can more easily move within sub-categories.
The industry as a whole is more flexible towards remote, location-independent employment. While many dominate tech companies are enforcing various kinds of return-to-work, that isn’t a blanket statement across the sector. Clearly employment is fragmented across many subsectors and a wide spectrum of roles.
Near shoring, friend shoring may drive a more honest, structural investment in reskilling, upskilling than we have seen in the past.
Concerns
I still have many concerns with what is happening:
AI-hype has many similarities to offshoring-hype. Many companies seem to have fully bought into the AI-hype cycle happening today without bothering to fully assess or consider the limitations. This is the same as the irrational embrace of offshoring that was seen from the 1980s onwards. Over the long-term these firms will be punished by the market and smarter companies will win, but in the short-term workers will suffer.
While tech roles seem more flexible and adaptive to the changes happening in the market, we haven’t solved the reskilling / upskilling puzzle. It is really difficult to reskill into an adjacent industry/skill-set and even with massive investments around the world in these areas, we don’t have good solutions. A lot ends up falling back on the shoulders of the worker to figure out how they can best reskill and convince employers with the majority of the power/leverage to give them a chance.
Governments, employers and educational institutions continue to fail and address the complexity of reskilling/upskilling in a measurable and impactful way. This may simply be a problem that is too complex to solve in a consistent way across all sectors, situations. And it is too vulnerable to vibes and engagement concerns - “just get everyone more training!”
The geopolitical situation around the world is likely to push countries to invest more heavily in innovation topics and large companies will also leverage these conflicts/fears to try and extract more concessions, money and power from governments. Giving large corporations more power, has historically not shown a strong connection with lasting results. An innovative economy needs a mix of the strengths of mega large ‘infra’ companies, innovative startups that look at ways to push the realm of the possible and mid-sized companies in-between.
As authors like Scott Galloway have written, the majority of “innovation” large companies claim is actually the result of long-term investment in deep research and development funded by the public largely for the purpose of defense. The risk is that too much public funding is channeled to protecting present-day rent-taking corporations and too little is channeled to innovating the future. The incentives seem out of balance.
Career and hiring advice
Given these dynamics, if you are hiring or looking for a job, I’d make these recommendations:
Make a point to look outside of your existing network. There was a time where you could hire the best and find the best roles by doubling down on your existing network. With the shifts happening, these dynamics seem to be inverted.
Consider where upskilling/reskilling actually matters. A software engineer who upskills as a PMO has more value than a software engineer who upskills in sales and marketing. Adjacent, synergistic and in demand skills have the highest potential to be interesting hires and useful roles.
Be willing to take a risk on people with drastically different skills when you are hiring something new and be willing to try something drastically different than what you have done when you want to break out of your network.
The dynamics of the market are changing but the core foundation for being successful in any role remain the same: you need to be willing to communicate, work hard and apply your skills, knowledge to work well with others and solve meaningful problems for your employer, customers and partners.
Look for people who have tried to start their own companies, try to start your own company and find people you can try and start a company together with. The last 20 years have seen a steady decline in people starting companies as the effects of the dot-com boom/bust wore off. The market is uncertain, unstable and AI-Automation technologies offer a new reset of the dynamics of the last 50 years. Now is the time to give a chance to people who have courage to try and fail. And it’s the time to find those pockets where there are new and old problems to solve.
Predictions…
I just wrote about how I hate predictions.
Yet, in my own inconsistency,I do want to offer some predictions based on this research as I’m interested to look back in a few years time and see what has played out based on this data and the similarities / differences to the offshoring of the 1980s.
The net-impacts on impacted subsectors of AI-Automation will be similar to those of offshoring and take ~10 years to see a similar collapse in relevant roles.
However, the overall employment across the main category of tech will stabilize (as there is a massive readjustment within the sector) and continue to grow within ~5 years.
The kinds of social impacts we saw with the offshoring of manufacturing jobs will hit tech subsector workers but it will be less obvious and fragmented. I don’t know the political impact of this. It seems like there will be something, yet I could see it going either way.
Governments will spend massively to attempt to upskill/reskill their populations to capture the benefits of the changes and this will force further changes in education. I’m hopeful yet this is the prediction I think is most likely to be wrong.