I was on the earnings call when the number landed: a capital-spend projection so large it rewired the room. Traders who had cheered a 33% revenue jump watched the stock exit the stage. You could feel Wall Street deciding whether faith in Meta’s AI bet was a courage play or a margin call.
I’ll walk you through why Meta could spend $145 billion (€133.4 billion) this year, what that means for the company and for you, and what to watch next. I won’t sugarcoat the risks—nor will I let you miss the few real chances that remain.
Meta’s market moment: shares dropped even after a fast revenue beat.
Investors punished Meta despite revenue growing 33%—the company’s sharpest quarterly jump since 2021—because management blew past guidance on spending.
Zuckerberg told investors 2026 capital expenditures would be at least $10 billion (€9.2 billion) higher than expected and could hit $145 billion (€133.4 billion). That number dwarfs last year’s capex of $72 billion (€66.2 billion) and turned a tidy beat into a headline risk.
Why did Meta’s stock fall despite revenue growth?
Because markets value profit now more than promise later. When management raises a capex target that large, it forces a new risk calculus: will future AI returns pay for years of upfront spending? The market answered with a sell-off.
Reality check: Reality Labs kept hemorrhaging cash in plain sight.
The company reported the Reality Labs division lost more than $4 billion (€3.7 billion) while bringing in just $402 million (€370 million) in sales this quarter.
That division has cost Meta over $80 billion (€73.6 billion) across six years—proof that big bets can become long, expensive wounds. You remember the Metaverse flop; this is the ghost from that decision reminding investors that not every experiment turns into profit.
Supply shock: memory prices are shaping a capital plan more than strategy.
Data center demand for AI models has pushed memory chips into a global squeeze, and Meta says higher component costs—especially memory—explain much of the capex increase.
The memory market has become a pressure cooker, compressing supply and inflating prices for everyone from hyperscalers to laptop buyers.
Is Meta really spending $145 billion on data centers?
Part of it is raw infrastructure: racks, networking, power and, yes, memory. But a chunk is also talent and new labs—Meta has poured billions into R&D and head-hunting (including hiring Scale AI’s Alexandr Wang to lead Meta Superintelligence Labs). This is not just pipes and servers; it’s a retooling of the company’s technical spine.
Muse Spark and the product bet: a first look that’s quietly hopeful.
Earlier this month Meta debuted Muse Spark, the first public fruit of the new lab’s work.
Zuckerberg called it the first release from Meta Superintelligence Labs and promised more product work built on a “strong model.” Muse Spark is currently proprietary with plans to open-source later, and Meta says it will power new consumer and business agents—two of them promised in the near term.
Meta is a casino, stacking chips into a single bet on general-purpose models and agent products—but unlike a gambler, management intends to build the tables and the crowd around them.
What is Muse Spark and why does it matter?
Muse Spark is Meta’s in-house AI model designed to power recommendations, translations and the new agent products. Internally, it’s already changing operations: Facebook and Instagram are reporting hundreds of millions of weekly views of AI-translated and dubbed videos, and the model is being phased into ad targeting and recommendations to boost personalization.
Workforce and efficiency: people feel the ripple before profits do.
Meta announced a 10% workforce reduction and offered voluntary buyouts to about 7% of U.S. staff—moves executives framed as building a “leaner operating model” to offset huge investments.
That’s the human side of capex: automation and generative models will cut some roles and reshape others. CFO Susan Li says the changes will help offset the cost of the AI build, but the optics and morale hit are real.
Where this leaves you and the industry: competition, costs, and a calendar to watch.
Google, Amazon and Microsoft are already sprinting with their own models and infrastructure. If Meta’s plan succeeds, it could narrow the gap—but success requires more than scale: it needs sustained product wins that monetize differently than social ads.
Keep an eye on three things: product adoption of Muse Spark and the promised agents, memory pricing and capex burn rate, and quarterly guidance for Reality Labs.
Meta just flipped a financial dial from big to gigantic, and you should ask whether the company is buying a future or covering a past mistake—what’s your read on the bet?