The Pulse: Amazon layoffs – AI or economy to blame?
Amazon is doing more mass layoffs, claiming it wants to be more nimble. But are job losses really about US economic fears, and how Amazon’s retail business will be affected?
Hi, this is Gergely with a bonus, free issue of the Pragmatic Engineer Newsletter. In every issue, I cover Big Tech and startups through the lens of senior engineers and engineering leaders. Today, we cover one out of four topics from last week’s The Pulse issue. Full subscribers received the below article seven days ago. To get articles like this in your inbox, every week, subscribe here.
Many subscribers expense this newsletter to their learning and development budget. If you have such a budget, here’s an email you could send to your manager.
Online retail giant Amazon unexpectedly announced 14,000 job cuts earlier last week. The massive round of layoffs at the company follows other mass redundancies in recent years:
January 2023: 18,000 people laid off.
March 2023: another 9,000 people let go
November 2023: hundreds of people let go inside the Alexa team, as Amazon was looking to shift Alexa more toward GenAI
April 2024: hundreds of people let go from AWS
Software engineers, unfortunately, seem hit hard by the latest layoffs: of 2,300 employees laid off in Washington State, 25% are software engineers, GeekWire reports. We can only speculate about the ratio across the rest of the company, but if cuts at HQ are heavy on engineering, then things don’t look promising for other locations, sadly.
The memo from Beth Galetti, Senior Vice President of People Experience and Technology, to workers didn’t explain much:
“Some may ask why we’re reducing roles when the company is performing well. Across our businesses, we’re delivering great customer experiences every day, innovating at a rapid rate, and producing strong business results. What we need to remember is that the world is changing quickly. This generation of AI is the most transformative technology we’ve seen since the Internet, and it’s enabling companies to innovate much faster than ever before (in existing market segments and altogether new ones). We’re convinced that we need to be organized more leanly, with fewer layers and more ownership, to move as quickly as possible for our customers and business.”
The statement is utterly confusing, as encapsulated by its message that “business is great, but we need to do layoffs”. Job cuts usually mean a business is in trouble, which obviously isn’t the case for Amazon. So, why are these layoffs really happening?
Layoffs to boost efficiency?
The company’s memo states:
“We’re convinced that we need to be organized more leanly, with fewer layers and more ownership, to move as quickly as possible for our customers and business.”
If this line of reasoning sounds familiar, it’s because most of the layoffs in 2023 were justified the same way. The tech industry overhired during the pandemic in 2020-2021, making orgs more bloated and decision-making slower. In February 2023, I reported on the trend of fewer middle managers, with Meta the first Big Tech giant to reduce its management layers. In 2023, most of Big Tech followed this approach with layoffs or reorgs. Managers acquired more reports, and tech companies cut down the number of layers between the CEO and individual contributors.
Given Amazon did other massive layoffs in 2023, it’s unlikely they missed the industrywide trend for fewer managers. While the current layoffs seem to be targeting managers quite a bit – from the Washington State layoffs, 20% of those let go are managers – there are still more ICs laid off than managers, overall. So, this official explanation doesn’t pass my personal “smell test”.
Layoffs to buy more GPUs?
The day after its jobs announcement, Amazon had more big news, this time about AI: it unveiled Project Rainer, the largest AI computing platform AWS has ever built. It already has 500,000 Trainium2 chips (built by Amazon), This capacity is already 70% larger than any AI computing platform in AWS’s history, and Anthropic is using all of it (!!) to train its next models. Below is an image of one of the several Project Rainer data centers packed with Amazon GPUs:
Building data centers is incredibly capital-intensive: Amazon has spent $11B on Project Rainer alone. Even though very profitable, Amazon might want to invest more cash than it currently has into building data centers. So, one reason for job cuts could be to reallocate financial resources from paying salaries and compensation towards building more data centers.
Before doing the math, a couple of concepts are important to understand:
Free cash flow (FCF): profit after payment for things like capital expenditure (CapEx), such as financing data centers. If a company wants to operate with as little debt as possible, FCF is usually very important. If Amazon wants to avoid loans and not touch its reserves, then data center investment would come from FCF, reducing FCF further.
Cash reserves: a company’s liquid reserve investments, usually an accumulation of investments in financial instruments like bonds and securities, or cash deposits.
Let’s run Amazon’s numbers:
Cash reserves: $93B. This is how much Amazon has in reserve.
FCF: $32B. This is the rough free cash flow Amazon has currently, as per its latest quarterly report. This is after deducting current data center investments.
Savings from layoffs: $2-4B. This is my estimate of the rough total compensation of 14,000 employees.
So, the savings from these layoffs wouldn’t even pay for half of Project Rainer ($11B in total), and Amazon could easily build 3x Project Rainers in the next year, without needing to dip into its savings! Of course, Amazon has its famous frugality principle, but this massive layoff of 14,000 people won’t make a big difference to how much it can invest in data centers; It can already spend much more, if it wants!
Leanness and AI fail job cuts “smell test”
It’s not only me who doesn’t buy the explanation that these layoffs are to streamline the company, or to redirect resources to AI. Arne Knudson worked at Amazon for nearly two decades, most recently as a software development manager (SDM), before leaving the company earlier this year. He shared his analysis, with some insider detail:
“In my 18 years at Amazon, I went through a few layoffs and hiring freezes.
This is the first time I’ve seen multiple years of significant layoffs essentially back-to-back. Even in the depths after the .com bubble, it wasn’t this bad. They’ve been laying people off now for almost 3 straight years.
The explanation that this is downsizing after hiring too many at the height of the pandemic doesn’t pass the smell test, at least to me. That was 3 years ago; they’re not that dumb to keep those people around for 3 extra years. Those folks were laid off back in ‘22.
I’m also not convinced that this is optimization due to AI. My degree’s in AI, and I worked on AI stuff at Amazon; I don’t think there’s enough automation yet, and it’s not accurate enough yet, to replace 30,000 people. The cost of inaccuracies seems too high. But I could be wrong; maybe they’ve gotten their false negative & false positive rates low enough to avoid too many region-wide AWS outages. (Or not.)
One of the articles I read said this was going to be in HR, and I can tell you as a former manager, my experience working with HR had been steadily worsening over the past 5-7 years. They outsourced so much of the work, overworked the people they had, and had such high turnover that I never knew who I was supposed to work with. When I needed to put someone on a performance plan or help a new hire receive some kind of accommodation, it seemed like it was a different person each time. If they really are laying off tens of thousands more HR folks, this is only going to get worse.
And, I suspect, it means they don’t plan on hiring MORE people in any of the business units for a year or more. So, by the smell-o-meter, this seems more significant than streamlining the workforce, improved AI, and “nah, we don’t need as many HR folks.”
US economy to blame for Amazon layoffs?
It’s safe to assume AWS as a business unit is doing just fine, as suggested by Project Rainer’s existence and the agenda for building data centers. But how is the e-commerce side of the business performing, and what’s its outlook?
If one business should have its finger on the pulse of the US economy, it’s Amazon with its size and self-professed, relentless customer focus, providing a window into people’s spending habits across the country. Flashing lights on the dashboard of the national economy may signal tough times ahead in e-commerce, which could be a reason to start cutting costs early.
There are concerning signs from other sectors about the US economy. Below is the CEO of the restaurant chain, Chipotle.
“Earlier this year, as consumer sentiment declined sharply, we saw a broad-based pullback in frequency across all income cohorts.
Since then, the gap has widened, with low to middle-income guests further reducing frequency. We believe that this guest with household income below $100,000, represents about 40% of our total sales. And, based on our data, they are dining out less often due to concerns about the economy, and inflation.
A particularly challenged cohort is the 25- to 35-year-old age group. We believe that this trend is not unique to Chipotle and is occurring across all restaurants as well as many discretionary categories. This group is facing several headwinds, including unemployment, increased student loan repayment and slower real wage growth. We tend to skew younger and slightly over-indexed to this group relative to the broader restaurant industry”.
Chipotle is saying that everyone is eating out less, particularly 25-35 year olds, because of inflation. If people spend less on Chipotle because of rising prices, then they may also spend less in other areas of their lives for the same reason, including on Amazon.
In the e-commerce supply chain, there’s evidence of this trend, which would mean delivery services like UPS have fewer parcels to deliver. Speaking of UPS, two days ago, it announced massive layoffs:
“United Parcel Service (UPS) has slashed 48,000 jobs this year — one of the largest single-year reductions by a US company since the pandemic — as the package giant scrambled to contain costs and revive its lagging stock price.
The Atlanta-based delivery behemoth disclosed the reductions Tuesday while reporting third-quarter earnings that beat Wall Street expectations.
UPS said 34,000 of the cuts hit drivers and warehouse operations, while 14,000 targeted management (...)
UPS shares jumped nearly 9% in Tuesday afternoon trading, even as the company reported weaker revenue and profits”.
UPS’s revenue is down on last year, which suggests that there are, indeed, fewer deliveries (or lower value ones). As with the latest job cuts at Amazon, these drastic layoffs could be explained by a lot of things, most easily by UPS expecting reduced trade in the future.
If US consumer spending is trending down, then the e-commerce sector will be among the first to feel this. It could explain why Amazon is making these layoffs now. It can also explain why Google, Meta, and Microsoft might not be seeing their businesses impacted: they’re not involved in retail like Amazon is, and the AI sector is very much booming.
Among all of Big Tech, Amazon is best positioned to detect changes in US consumer spending. Google’s and Meta’s revenue is more dependent on advertising, and Microsoft’s more on enterprise spend. Like Amazon, Apple is well placed to feel market changes with its range of smartphones and watches, and other consumer tech.
I believe Amazon is highly commercially rational, so it’s worth understanding the actual reason for its second major mass layoffs in just two years, following deep cuts in 2023. I’d put my money on this reason being the economy, and how Amazon probably expects customers to cut back their spending everywhere, including on Amazon.
This was one out of the four topics covered in last week’s The Pulse. Read the full article.
This week’s The Pulse additionally covers:
Cursor and GitHub double down on agents. Each company is focusing on agents: Cursor with its multi-agent mode, and GitHub with its “Agent HQ.” Cursor is increasingly a direct rival to GitHub.
Industry pulse. Meta rolls out AI-assisted interviews, Cursor and Windsurf believed to be using Chinese open source AI models, South Korean government pays price of ignoring backup “101,” startups growing much faster in the US than in Europe, companies using AI tools buy more JIRA seats, a neat uptime service called Updog, and more.
OpenAI inflating the bubble? OpenAI signs another massive deal with AWS based on predicted growth, and seeks taxpayer protection to borrow more.
Large tech companies struggle to build their AI integrations. Apple admits failure to modernize Siri by paying Google $1B per year for its LLM. Perplexity to pay Snap $400M to be its AI search interface.
How much do Directors of Engineering earn at startups? Data from Carta says it’s more than any other Director: $215-230K at companies valued at $25-250M.