November 5, 2025

How $125 Billion in AI Spending Becomes 75,000 Layoffs

Who Stays, Who Goes: The Real Story Behind Tech's Biggest Cutbacks

By Jay Cadmus

Here's the Kicker

On October 28, Amazon announced 14,000 job cuts with CEO Andy Jassy blaming "culture." Meanwhile, on OpenAI's careers page, "Forward Deployed Engineer" postings sat unfilled at $200,000-plus. Both are happening at the same time. Amazon is cutting because its infrastructure bill hit $125 billion this year. OpenAI is hiring because enterprises will drop serious money for humans who can actually get AI working in the real world. The tech industry isn't in crisis over AI replacing people. It's in crisis because the compute cost more than the people, and that math forces brutal choices about who stays and who goes.

Top Takeaways

• Amazon cut 14,000 corporate jobs (possibly 30,000 total) October 28 while raising 2025 infrastructure spending to $125 billion, up 30% just in one year. That money has to come from somewhere, and it's coming from the payroll
• In October and November alone, over 75,000 tech workers got pink slips (Amazon, UPS, Accenture, Microsoft, Google, Meta, IBM, Salesforce, Dell). Every company says it's about "efficiency." Every CFO report shows the real story: infrastructure costs are eating the budget alive
• Anthropic spent $2.66 billion on AWS compute through September while bringing in only $2.55 billion in revenue. That means they're spending more on keeping the lights on than they're making. That doesn't work, and when your biggest customer can't make the math work, Amazon has to find the money somewhere
• Accenture cut 11,000 people in 90 days (spent $865 million on severance) while raking in $2.6 billion from AI consulting work in just six months. The pattern is clear: get rid of the generalist, hire the specialist who can actually sell AI projects
• OpenAI and Anthropic are now hiring "Forward Deployed Engineers" at $200–400K each, embedding them with clients for months. This is the counter-move: while companies cut, the AI companies are hiring people who can turn AI into revenue
• Anthropic went from 1,000 to 300,000+ enterprise customers in two years. They're not just selling API access. They're opening offices in Dublin, London, Tokyo, and hiring across the board because enterprises will pay a lot for people who know how to implement AI
• AI job postings jumped 56% in 2025, with roles like AI Engineer up 143% and AI Solutions Architect up 109%. At the same time, 94% of leaders say they can't find AI talent while they're laying off tens of thousands. The market is splitting into two camps: people companies want and people they don't
• Data centers ate up $657 billion in spending in 2025, nearly double what was spent in 2023. McKinsey thinks we'll need $6.7 trillion for data centers by 2030. When that's the priority, everything else gets cut

The Brief

Here's what's really happening in tech right now: companies aren't cutting jobs because AI is smart enough to replace people. They're cutting jobs because the servers that run AI cost so much that people have to go. Amazon spends $28.9 billion a quarter on tech and infrastructure (up 30% in a year) while cutting corporate staff. That's not a coincidence. The GPU bills are crushing the people budget.

But on the flip side, Anthropic is hiring Forward Deployed Engineers at crazy salaries because enterprises will spend $200 million for someone who can actually make AI work in their organization. That's not API access. That's implementation, consulting, training, problem-solving. That's actual human work that AI can't do yet.

The two trends seem to contradict each other, but they're both reactions to the same economic reality: infrastructure is expensive, and everything else has to shrink to pay for it.

Deep Dive

The Infrastructure Bill Is Getting Out of Hand

Look at Amazon's Q3 2025 earnings call on October 30. Revenue went up 13% to $180.2 billion. AWS, their cloud business, grew 20.2% to $33 billion in the quarter. That's good news for investors. Then CFO Brian Olsavsky talked about spending.

Capital expenditures in Q3 alone: $34.2 billion. Year-to-date: $89.9 billion. Full-year projection: $125 billion. That's nearly double what they thought it would be at the start of the year. And for 2026? Olsavsky basically said "wait and see, but it's going up" (CNBC, 2025–10–30).

To put that in perspective, $125 billion is more than Amazon spent on salaries for its entire 350,000-person global workforce last year. But it's not just servers. It's the power to run them, the cooling systems, the real estate, the networks. All of it adds up to a number that crushes the traditional budget categories.

The really telling number sits buried in the expense report: technology and infrastructure costs jumped 30.2% year-over-year to $28.9 billion in Q3 alone. That's growing three times faster than the overall business. When one category of spending grows that fast, something else has to shrink. That something else is people.

This isn't just an Amazon problem. Microsoft dropped $80 billion on AI infrastructure in fiscal 2025 while laying off 15,000 people across May and July. Google raised its infrastructure spending from $75 billion to $91–93 billion and cut 9,000 people. Meta committed $70–72 billion (revised up from their original estimate) while cutting 600 people from their AI labs alone. IBM announced layoffs in Q4 2025 targeting thousands (WIRED, 2025–10–29; Economic Times, 2025–11–02; Reuters, 2025–11–04).

When every big tech company is following the same pattern, it's not a coincidence. It's a response to the same math problem: infrastructure is expensive, and people are negotiable.

The Numbers Don't Lie When You Look Closer

Anthropic is Amazon's major AI company investment, and their story is brutal. Through September 2025, they spent $2.66 billion on AWS compute while bringing in $2.55 billion in total revenue. They're spending more on infrastructure than they make. That's 104% of their entire revenue going just to keep the lights on, before paying a single engineer, before marketing, before anything else (Where's Your Ed At, 2025–10–19).

In June alone, Cursor (an AI coding tool that relies on Anthropic) saw its AWS bill jump from $6.2 million to $12.6 million in one month. That's a 103% increase in a single month. When your biggest customers are burning cash that fast just to stay operational, the pressure to cut costs upstream becomes intense (Where's Your Ed At, 2025–10–19).

Accenture showed exactly how companies respond to this pressure. They cut 11,000 people in 90 days and spent $865 million on severance. But in the same six months, they made $2.6 billion from AI consulting work. The strategy is transparent: kill the generalist jobs that don't directly feed AI revenue streams, and pour the money into hiring specialists who can sell AI projects at premium rates. CEO Julie Sweet said it plainly: "We are exiting people where reskilling is not a viable path for the skills we need" (Upskillist, 2025–09–28). If you can't immediately help sell AI, you're out.

The Layoffs: It's Real and It's Massive

October and November 2025 saw over 75,000 tech workers lose their jobs. This isn't scattered cuts. It's coordinated across the entire industry.

Amazon: 14,000 announced, potentially 30,000 over multiple waves. UPS: 48,000. Accenture: 11,000 in 90 days. Microsoft: 15,000 across two rounds (6,000 in May, 9,000 in July). Google: 9,000 through the year. Meta: 3,600 through 2025, including 600 specifically from AI labs in October. IBM: Thousands announced for Q4. Salesforce: 1,200+ for the year. Dell, Target, and others: Additional thousands.

Every company has a different excuse. Amazon says "culture." Microsoft cites "efficiency." Google offers "voluntary" separations. Meta calls it "removing bloat." But if you read the earnings calls and the financial reports, they all say the same thing: infrastructure spending is up 25–40% while everything else gets cut (TechCrunch, 2025–10–23; Economic Times, 2025–11–02; CNBC, 2025–10–28).

It's not a secret. They're just not advertising it loudly.

The Flip Side: AI Companies Are Actually Hiring

This is where it gets interesting. While Amazon cuts 14,000, OpenAI is hiring "Forward Deployed Engineers." This isn't a support position. These are specialists at $200–400K salary embedded on-site with clients, living with Pentagon contractors and financial firms, building custom AI systems that work. It's a new category of work, and it's expensive because it actually produces results.

Colin Jarvis runs this division for OpenAI. The early projections show $5–10 billion in revenue from this model alone, potentially scaling to $50–100 billion if it works at full capacity (Angela Stewart AI, 2025–07–07; Answer Rocket, 2025–11–03).

Anthropic is on a completely different trajectory. Two years ago they had 1,000 customers. Now it's 300,000+. In August, their revenue was running at $5 billion annually, up from $87 million at the start of 2024. They just closed a Series F funding round at $183 billion valuation, raised $13 billion, and opened offices in Dublin, London, Tokyo, and Zurich. They're hiring 100+ people in Europe alone (Anthropic, 2025–09–25).

On November 3, Cognizant, one of the world's biggest consulting firms, announced a partnership with Anthropic to build enterprise AI systems. This isn't about buying API access. This is about hiring people, teams, and implementation specialists who can integrate AI into real business operations (eWeek, 2025–11–04).

The contrast is stark. Infrastructure companies are cutting people. AI implementation companies are hiring. If you work for one, you're getting laid off. If you work for the other, you can probably name your price.

The Split Is Happening Right Now

AI job postings jumped 56% in 2025. Roles like AI Engineer are up 143% year-over-year. AI Solutions Architect is up 109%. Prompt Engineer is up 96% (Autodesk, 2025–07–08).

NVIDIA has 901 open positions in AI. OpenAI has 233 of its 312 total openings in AI-related roles. Microsoft still lists 111 AI research positions despite all the layoffs (Edison & Black, 2025–07–05).

But here's the messed up part: 94% of leaders say they can't find AI talent while they're laying off tens of thousands. That's not a contradiction. That's the market doing exactly what it's supposed to do. Companies don't need more people. They need different people. They need specialists, not generalists. And if you're a generalist, you're out.

The traditional career path in tech is disappearing. The old model was: get hired as a junior, learn the business, get promoted to manager, build a team, climb the ladder. That worked when companies needed lots of people.

Now? Companies want specialists who can start day one and immediately contribute. Gen Z job openings have actually fallen since 2022 as entry-level roles disappear. The pyramid isn't getting rebuilt; it's getting eliminated (New York Post, 2025–11–04).

Meta is throwing $100 million signing bonuses at AI researchers to poach them from OpenAI. Google DeepMind is paying $20 million a year for top AI scientists. Those aren't normal tech salaries. Those are scarcity prices, and they signal that the market knows what it needs: people who can build AI, and not many of them (Edison & Black, 2025–07–05).

What Everyone's Not Talking About Openly

The infrastructure spending isn't slowing down. Amazon added 3.8 gigawatts of power capacity in the last year. They're adding another gigawatt in Q4 and planning to double total capacity again by 2027. That's equivalent to multiple nuclear power plants' worth of electricity (Futurum Group, 2025–11–03).

On November 3, OpenAI and Amazon announced a $38 billion, seven-year contract for GPU access. That sounds enormous because it is. OpenAI makes around $4 billion a year right now. This one contract locks in $5+ billion annually in infrastructure costs. For that math to work, AI revenue has to explode (CNBC, 2025–11–03).

Here's the thing that should worry people: Amazon is making record profits ($21.2 billion net income in Q3 alone). Accenture is killing it on revenue ($17.6 billion in Q4 ahead of estimates). Google, Microsoft, and Meta are all printing cash. They're not cutting jobs because they're desperate. They're cutting because the market rewards you for cutting. Amazon's stock hit all-time highs right after announcing layoffs. Wall Street loves cost discipline, even when it means people lose their jobs (CNBC, 2025–10–31).

That sends a message: cut people, keep the infrastructure spending. The companies getting it right are the ones who figured out that AI is going to be expensive for years, infrastructure budgets always come first, and everything else has to fit into what's left.

The Risks That Nobody's Fully Thinking Through

This strategy has serious failure modes. When you cut senior people (directors, experienced managers), you lose the stuff that's hard to replace. You lose the person who knows why a system was built a certain way, which customer has weird contract terms, how the org actually works versus how the org chart says it works. New hires don't have that context.

If you're constantly cutting the bottom and middle layers, the next generation of engineers and managers is going to be smaller and less experienced. That means less mentorship, less institutional knowledge, more mistakes.

Enterprise AI adoption might actually be slowing down. RBC Capital analysts looked at this in early November and found that the percentage of U.S. businesses actually paying for AI services dropped from 44.5% in August to 43.8% in September. That's the first decline in years (Business Insider, 2025–11–03). If companies are slowing on AI purchases while speeding up on infrastructure capex, the math stops working.

Amazon's profit margins on AWS have also gotten worse. Operating margin on the cloud business dropped 3.4 percentage points even though they grew revenue 20%. When your best business is growing fast but making less profit per dollar, that's a warning sign (Sergey CYW, 2025–10–30).

What to Watch

Q1 2026 earnings will tell you whether the capex binge is justified. If revenue growth doesn't match spending growth, expect more layoffs and potentially the company admitting they invested too much too fast.

Pay attention to attrition in January and February. If the best people leave after seeing layoffs, the strategy backfires. Companies need some continuity or things break.

Watch AI deal sizes. If companies are doing more AI business but for less money per deal, the economics get worse fast.

Most importantly, watch how fast OpenAI and Anthropic actually grow their Forward Deployed Engineer teams. If hiring slows down, it means enterprises are getting skeptical about the ROI. If it accelerates, the market is saying the model works.

Playbook

If you're an executive

Do a real audit of your AI infrastructure spending versus the business value you're actually getting. If capex is growing 30% while revenue grows 13%, you're headed for trouble. Start justifying it now or start cutting.

If you're a mid-level manager or coordinator

Your role is exposed. You've got maybe six months to figure out if you can transition into something that directly touches AI or revenue. If not, start looking elsewhere. Internal retraining isn't moving fast enough to save these roles.

If you're a software engineer or technical person

This is your best moment since the dot-com crash. Roles in AI implementation, forward deployment, and infrastructure optimization are scarcity categories. Build actual projects that prove you can ship something.

If you're early in your career

The easy path into tech has closed. You need to show you can already do something valuable, not just that you're willing to learn. Bootcamps, side projects, demonstrated skills matter more than degrees.

What's This Mean?

The machines didn't eat the jobs. The infrastructure costs did, and now the market is brutally sorting who's valuable and who isn't.

References

Amazon.com. (2025, October 30). Amazon.com announces third quarter results. https://ir.aboutamazon.com/news-release/news-release-details/2025/Amazon-com-Announces-Third-Quarter-Results/

Angela Stewart AI. (2025, July 7). OpenAI's new weapon: Forward deployed engineers. https://angelastewartai.substack.com/p/openais-new-weapon-forward-deployed

Answer Rocket. (2025, November 3). Forward deployed engineers: The critical role that drives AI success. https://answerrocket.com/forward-deployed-engineers-the-critical-role-that-drives-ai-success/

Anthropic. (2025, September 25). Anthropic expands global leadership in enterprise AI. https://www.anthropic.com/news/anthropic-expands-global-leadership-in-enterprise-ai-naming-chris-ciauri-as-managing-director-of

Autodesk. (2025, July 8). AI job growth in design and make: 2025 report. Autodesk News. https://adsknews.autodesk.com/en/news/ai-jobs-report/

Business Insider. (2025, November 3). Ruh roh. Analysts sense a chill in enterprise AI demand. https://www.businessinsider.com/enterprise-ai-demand-chill-analysts-2025-10

CNBC. (2025, October 28). Amazon laying off about 14,000 corporate workers. https://www.cnbc.com/2025/10/28/amazon-layoffs-corporate-workers-ai.html

CNBC. (2025, October 30). Amazon (AMZN) Q3 earnings report 2025. https://www.cnbc.com/2025/10/30/amazon-amzn-q3-earnings-report-2025.html

CNBC. (2025, October 31). Amazon stock soars 11% after Q3 earnings beat. https://www.cnbc.com/2025/10/31/amazon-stock-market-amzn.html

CNBC. (2025, November 3). Amazon closes at record after $38 billion OpenAI deal with AWS. https://www.cnbc.com/2025/11/03/open-ai-amazon-aws-cloud-deal.html

CNBC. (2025, October 22). Meta lays off 600 from 'bloated' AI unit. https://www.cnbc.com/2025/10/22/meta-layoffs-ai.html

CRN. (2025, August 4). Dell Technologies layoffs hit sales, 'new logo' acquisitions team. https://www.crn.com/news/computing/2025/dell-technologies-cuts-undisclosed-number-of-sales-jobs-nationwide

Edison & Black. (2025, July 5). What AI companies are hiring the most right now. https://edisonandblack.com/pages/what-ai-companies-are-hiring-the-most-right-now.html

Economic Times. (2025, November 2). Over 100,000 job cuts rattle tech industry in 2025. https://economictimes.indiatimes.com/news/international/global-trends/us-news-over-100000-job-cuts-rattle-tech-industry-in-2025-amazon-meta-google-intel-lay-off-thousands-of-employees-check-full-list-of-companies/articleshow/125029264.cms

eWeek. (2025, November 4). Cognizant taps Anthropic's Claude to power enterprise AI. https://www.eweek.com/news/anthropic-cognizant-deal/

Fortune. (2025, July 15). Tech layoffs 2025: How Microsoft, Google, and Meta are plotting for the future. https://fortune.com/2025/07/16/tech-layoffs-2025-how-microsoft-google-meta-amazon/

Futurum Group. (2025, November 3). Amazon Q3 FY 2025 results: Strong AWS, retail, ads momentum. https://futurumgroup.com/insights/amazon-q3-fy-2025-earnings-aws-reaccelerates-retail-and-ads-grow/

McKinsey & Company. (2025, April 27). The cost of compute: A $7 trillion race to scale data centers. https://www.mckinsey.com/industries/technology-media-and-telecommunications/our-insights/the-cost-of-compute-a-7-trillion-dollar-race-to-scale-data-centers

New York Post. (2025, November 4). Gen Z job openings have dwindled since 2022 as AI slams entry-level roles. https://nypost.com/2025/11/04/business/gen-z-job-openings-have-dwindled-since-2022-as-ai-slams-entry-level-roles-study/

Reuters. (2025, November 4). IBM to cut thousands of jobs in fourth quarter amid software focus. https://www.reuters.com/business/world-at-work/ibm-cut-thousands-roles-focus-software-growth-bloomberg-news-reports-2025-11-04/

Sergey CYW. (2025, October 30). Amazon Q3 2025 earnings analysis. Compounding Your Wealth. https://sergeycyw.substack.com/p/amazon-q3-2025-earnings-analysis

Tech.co. (2025, September 29). Accenture makes room for AI with thousands of layoffs. https://tech.co/news/accenture-layoffs-ai-pivot

TechCrunch. (2025, October 23). A comprehensive list of 2025 tech layoffs. https://techcrunch.com/2025/10/24/tech-layoffs-2025-list/

TechRadar. (2025, August 14). Amazon's data center spend tops $100bn, more than the GDP of most countries. https://www.techradar.com/pro/amazons-data-center-spend-tops-usd100bn-more-than-the-gdp-of-most-countries

Times of India. (2025, November 4). IBM to cut thousands of jobs, here's what the company said. https://timesofindia.indiatimes.com/technology/tech-news/ibm-to-cut-thousands-of-jobs-heres-what-the-company-said/articleshow/125103243.cms

Upskillist. (2025, September 28). Accenture announces 11,000 layoffs amid $865M AI-driven restructuring. https://www.upskillist.com/blog/accenture-layoffs-ai-driven-restructuring/

Where's Your Ed At. (2025, October 19). This is how much Anthropic and Cursor spend on Amazon Web Services. https://www.wheresyoured.at/costs/

WIRED. (2025, October 29). Meta, Google, and Microsoft triple down on AI spending. https://www.wired.com/story/microsoft-google-meta-2025-earnings/

World Economic Forum. (2025, October 2). How we can balance AI overcapacity and talent shortages. https://www.weforum.org/stories/2025/10/ai-s-new-dual-workforce-challenge-balancing-overcapacity-and-talent-shortages/

Yahoo Finance. (2025, October 30). Amazon.com Inc (AMZN) Q3 2025 earnings call highlights. https://finance.yahoo.com/news/amazon-com-inc-amzn-q3-030438522.html