Last October, I wrote about the impact of Gen AI on developer productivity. Based on the few studies I could find, the results were a mixed bag. Any productivity improvements seemed from AI tools such as Microsoft GitHub CoPilot seemed to be balanced out, in whole or in part, by concerns over software quality, data privacy, and security. [1]
Now eight months have passed, and many observers have pointed to massive layoffs in the tech sector as evidence that generative AI—starting in 2022 and continuing to the present day—as evidence that the boost in productivity is real. Just this month, in the past two weeks, Microsoft announced it is laying off an additional 4% of its workforce, following earlier massive layoffs over the past three years. Other large tech firms, such as Amazon, Meta, and Intel also announced mass job cuts. And even experienced developers, let alone entry level candidates, are having a hard time finding work. So, wouldn’t that confirm that AI tools are having a negative impact on developer jobs?
Many observers think so.
But what if AI is not the primary factor?
I recently came across an analysis of the 2017 Tax Cuts and Jobs Act (TCJA), which changed how companies should account for research and development investments. Prior to the TCJA (since 1954!) companies were able to deduct 100% of qualified R&D spending against taxable income in the current year, instead of amortizing these costs over a depreciation schedule of many years.
For almost 70 years, American companies could deduct 100% of qualified research and development spending in the year they incurred the costs. Salaries, software, contractor payments — if it contributed to creating or improving a product, it came off the top of a firm’s taxable income.
The deduction was guaranteed by Section 174 of the IRS Code of 1954, and under the provision, R&D flourished in the U.S.
But the TCJA changed that by phasing out the immediate expensing of R&D investments starting in 2022, forcing those costs to be amortized.
Why would Congress do this? Here’s the key: Congress wanted to cut the overall corporate tax rate from 35% to 21%, which would be good for business. But under Congressional PAYGO rules, there had to be something to offset the lost tax revenue, so lawmakers looked at doing away with immediate expensing of R&D investments. But whereas the corporate tax rate cut would go into effect immediately (in 2018), the immediate expensing of R&D investments would be phased out starting in 2022—A time bomb indeed.
Is it a coincidence that the massive tech industry layoffs started in 2022? I think not.
Now when I first posted this study on LinkedIn, I got a few comments that essentially said, “Research and development is not optional. It critically important for every business, and companies should invest in technology.”
Of course, as a general principle, this is true. But it misses the point. All businesses make investment decisions based the return on investment (ROI). R&D is no exception. Some proposed R&D projects have a high projected return over a short period of time. Others have a low return over a long timeframe. And there are a whole bunch in between.
So, prior to TCJA, business decision-makers might approve, say, half of the requested R&D investments based on the ROI analysis (and other factors, of course).
But that ROI analysis is made according to the payback calculated after tax is considered. So now, with immediate expensing being phased out, some of those projects are no longer cost-justified on an after-tax basis. Remember the time value of money? A return delayed over several years is not as attractive as a return realized in the current year. Coincidentally, interest rates in the US began rising at the same time, starting in 2022. So, just as tech companies were losing the immediate expensing of R&D investments, interest rates began to rise, making the time value of money even greater.
Now, you say, “But the corporate tax rate is lower, so that makes up for it.” Yes, that may be true for your typical corporation outside of the tech sector that spends under 2% of revenue on R&D [2]. In fact, they’ll say, “No problem, we’ll gladly give up the R&D expensing and take the overall tax cut.”
But the technology sector typically spends over 10% on R&D, with the most innovative players spending far more. So, the change in the R&D deduction has a disproportionate negative impact on the tech sector—exactly where we are seeing the layoffs.
Some other commenters asked, if the tax code is to blame for all the tech sector layoffs, then why are these companies claiming that AI is the reason?
My response is, because that’s a better story for Wall Street. No tech company CEO wants to say, “We are laying off all these people because we can no longer get the tax break we were accustomed to.” A better story is to say, “Look how great we’re doing with AI! We have gotten so much more efficient, now we don’t need all these people!”
I’ll submit one personal story as evidence. In 2024 I had a casual discussion during a coffee break with a couple of senior executives from one of those major tech companies, which I won’t name. I asked them this very question. “Is generative AI investments the reason for your layoffs?”
The answer was, “No, it’s too early for that to be the reason.” They pointed to pandemic over-hiring, which was certainly one reason. But it is also likely that they weren’t even aware of the tax changes that were going into effect.
If you’re in the tech industry, the good news is that the so-called “One Big Beautiful Bill Act” (OBBBA) now restores the immediate expensing of qualified R&D investments.
According the a summary by the US Chamber of Commerce:
For nearly 70 years, American businesses were able to immediately deduct 100% of their research and experimental (R&E) expenses, which generally include all costs incident to the development, testing, or improvement of products or services. Starting in January 2022, however, businesses have been required to amortize (deduct ratably) their R&E expenses over five years. As we’ve previously written, mandatory R&E amortization reduces economic growth and penalizes investments by businesses in research-intensive industries—with disproportionate effects on smaller manufacturing and technology companies.
The tax bill will address this counterproductive policy by permanently reinstating the deduction for domestic R&E expenses paid or incurred in 2025 and beyond.
Also, notably, reading the full text of the OBBBA we see a special rule specifically defining software development as a research and experimental expenditure:
(3) Software development. For purposes of this section, any amount paid or incurred in connection with the development of any software shall be treated as a research or experimental expenditure.
And there’s more good news. As noted in this summary, the OBBBA provides an increased R&D credit for US-based labor. This provides an incentive for employers to employ R&D workers in the US, rather than offshore. Plus, there is a new “Innovation Workforce Credit,” which offers 15% credit for STEM apprentice wages, which should further boost the demand for tech labor.
How do you see things? Let me know in the comments [3].
[1] Experienced developers continue to hold a nuanced view of the role of generative AI in software development, whether it is a net plus or minus. In short, the jury is still out. For example, see this recent post by Fred Hebert, where he writes,
Being able to effectively use LLMs for programming demands a lot of scaffolding and skills. The skills needed are, however, poorly defined and highly context dependent, such that we currently don’t have great ways of improving them other than long periods of trial and error.
The problem is that while the skills are real and important, I would argue that the level of sophistication they demand is an accidental outcome of poor interaction design. Better design, aimed more closely to how real work is done, could drastically reduce the amount of scaffolding and learning required (and increase the ease with which learning takes place).
I don’t expect my calls to be heard. Selling sci-fi is way too effective. And as long as the AI is perceived as the engine of a new industrial revolution, decision-makers will imagine it can do so, and task people to make it so.
[2] Here is a source for R&D spending by industry. And here are some statistics for the tech sector.
[3] As always, nothing I write should be taken as investment, legal, tax, or accounting advice. Please consult the services of a competent professional when you need such counsel.
Please share with others to support my work.
Like this post? Browse my past posts.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.