Nvidia Announces $80B Buyback as Meta Raises 2026 Capex to $145B—AI Cycle Shifts to ROI Accountability

Three of the most capital-visible names in the AI trade just sent signals that point in different directions at once — which is exactly what a maturing market cycle looks like.

Nvidia announced an $80 billion share repurchase program, Meta raised its 2026 capex forecast to $145 billion, and Palantir boosted its full-year revenue guidance to $7.656 billion with 71% growth.

Morgan Stanley suggests the AI cycle is shifting from capital expenditure to ROI accountability, with capital potentially rotating from chipmakers to hyperscalers.

Meta's $145 billion capex commitment is the single most important data point. The spending is existential — if Meta stops building and a competitor doesn't, it loses the advertising advantage from AI-powered targeting and recommendation systems.

But Meta's free cash flow declined 12.4% year-over-year while revenue grew 26.2%, highlighting the structural tension of this phase.

The stock now trades at 21.6 times earnings, and the market is asking harder questions about when capex turns into profit.

"A $145 billion capex commitment is enormous, but it's being funded by a company that still trades at 21x earnings and generates real advertising revenue to justify every dollar spent. If the AI monetization story even partially delivers, that stock is priced for a fraction of what it could become." – Victor Hale, AI research agent

The market is transitioning from the capex-build phase to the ROI-accountability phase, where every dollar of infrastructure needs to generate visible revenue.

Nvidia generated $119 billion in free cash flow over the trailing twelve months at a 47% FCF margin, with a remarkable 89.4% ROIC.

The $80 billion buyback signals that Nvidia sees its cash generation as durable enough to absorb massive capital return even as the industry continues spending.

The company's ROIC sits at 89.4%, and it trades at roughly 30x earnings on 85% revenue growth.

Palantir sits furthest downstream from the compute build-out and closest to the software monetization layer, with zero infrastructure to build, 84.1% gross margins, and a 38.1% operating margin.

However, the company faces scrutiny over $43 million in insider sales amid a 60.8x price-to-sales ratio. The AI cycle ROI shift is creating a clear value chain distinction.

As Victor Hale, an AI research agent, put it: "The question is no longer whether AI spending is real. It's whether your capital is in the right part of the pipeline."

The debate isn't about whether these companies are important — it's about whether the return profile at current prices justifies the allocation.

In a cycle shifting from build to prove, the companies generating cash may look more compelling than those that still need to convince the market the spending pays off.