The earnings calls are saying the quiet part loud: AI budgets are ballooning while head counts shrink, and the gap between those two lines is the future of work.
The Summary
- Alphabet added 4,000+ staff last quarter while Meta cut 2,000+, showing wildly divergent strategies as Big Tech doubles down on AI infrastructure spend
- Block CEO Jack Dorsey explicitly tied his company's near-50% workforce cut to "intelligence tools" enabling "smaller and flatter teams", the clearest CEO-level admission yet that AI is driving structural downsizing
- Companies face a messaging paradox: tell investors AI makes you lean and efficient, tell employees AI augments their work — both can't be fully true
- World Economic Forum estimates 92 million jobs displaced by 2030, 170 million created, but that 78 million net positive assumes smooth transitions nobody has figured out yet
The Signal
The earnings season data is painting two futures at once. Alphabet is hiring like it's 2021, adding thousands to build out AI infrastructure. Meta is trimming staff while pouring billions into the same infrastructure race. The divergence isn't about whether AI matters, it's about who builds it in-house versus who buys it as a service. Google needs bodies to maintain competitive moats in foundation models. Meta is betting it can do more with less by using those models.
"The intelligence tools we're creating and using are enabling a new way of working which fundamentally changes what it means to build and run a company."
Jack Dorsey's February X post about Block's cuts is the template every CEO is studying. He said what others are thinking: AI doesn't augment your team, it replaces layers of it. Block cut nearly half its workforce and framed it as evolution, not emergency. The playbook is clear — automate middle management, flatten hierarchies, call it innovation.
The tension Bloomberg identifies is the real story. How do you tell investors AI will make your company radically more efficient while telling employees their jobs are safe? The answer: you don't. You split the message. Earnings calls get the efficiency narrative. All-hands meetings get the "augmentation, not replacement" talk. Neither is a lie, but they're describing different timelines and different people.
Key split in how companies are managing the transition:
- Infrastructure builders (Google, Amazon) still hiring heavily for AI R&D and deployment
- Infrastructure buyers (Meta, Block) cutting staff as they implement AI tools built elsewhere
- Everyone increasing capital expenditure on AI while total compensation expenses trend down
Anthropic CEO Dario Amodei warned of "significant enduring job loss" while AWS CEO Matt Garman called doomsday predictions overblown. Both run companies selling AI infrastructure. Their disagreement isn't philosophical, it's tactical. Amodei needs regulators and the public to take AI seriously. Garman needs enterprise customers to not panic and freeze spending.
The World Economic Forum's numbers tell you the shape of the problem even if you don't trust the precision. 92 million jobs displaced, 170 million created by 2030. That's a 78 million net gain and a 262 million job transition. A net positive doesn't mean much if you're in the 92 million and the path to the 170 million is unclear. The churn is the crisis, not the arithmetic.
The Implication
Watch the ratio of capital expenditure to total workforce compensation in the next four quarters. When capex on AI infrastructure grows while total comp shrinks, you're seeing automation hit critical mass. The companies adding head count today are building moats. The ones cutting are buying efficiency. Both strategies work until the model providers commoditize and everyone who trimmed has to rebuild institutional knowledge they automated away.
If you're building a career in tech, the signal is simple: be close to the infrastructure or close to the customer. Middle layers — the analysts, the coordinators, the people who move information from one system to another — are getting compressed first. The jobs being created are in model training, agent orchestration, and high-touch human work AI can't fake yet. The 78 million net job gain assumes you can retrain fast enough to jump categories. Start now.