Meta’s $145 Billion AI Bet: What Muse Must Prove

Meta’s planned AI spending is a bet on more than better answers inside its existing apps. By building computing capacity and launching an agent that can carry out tasks for consumers, the company is trying to turn Zuckerberg’s vision of personal superintelligence into a new business. Muse makes that ambition tangible; it has not yet established a return on the infrastructure behind it. 2718

How large is Meta’s 2026 spending plan?

Meta’s latest reported guidance is $130 billion–$145 billion in 2026 capital expenditures, compared with $72.2 billion spent in 2025. At the top of the range, that would be roughly 101% more than last year—not a guarantee that spending will double. The July revision raised the lower end of the range from $125 billion to $130 billion while leaving its $145 billion ceiling unchanged. 214

That scale signals how much capacity Meta is willing to commit to its AI ambitions, not that superintelligence has been achieved. The comparison with national military budgets in the broader discussion is not verified by the supplied evidence; corporate capital investment and defense budgets would also measure different things.

What Muse changes for the business

Launched on September 8, Muse can handle tasks including sending emails, booking travel and completing transactions on a user’s behalf. It offers a free version alongside $20 and $100 monthly subscriptions. That gives Meta a potential source of consumer revenue beyond advertising, although paid plans alone say little about eventual profit. 18

Early demand is substantial. Apptopia estimated 2.8 million Muse downloads in its first 12 days and, on a comparable U.S.-and-Canada iOS measure, 1.8 million downloads for Muse versus 1.3 million for ChatGPT at the same point after each app’s launch. CNBC separately reported that Muse had overtaken ChatGPT atop the U.S. free iOS chart, citing Sensor Tower data. These are launch-period download estimates, not measures of retained users, paid subscribers or reliable task completion. 1819

Why investors are divided

Meta still has a powerful revenue engine: second-quarter sales rose 28% to $60.8 billion. But free cash flow fell 91%, from $8.55 billion a year earlier to $784 million, as the AI buildout strained cash generation. Capital spending affects cash flow immediately; it should not be mistaken for an equal-sized charge to that quarter’s earnings. 1

Muse’s reception has nevertheless changed some expectations. On September 10, JPMorgan analyst Doug Anmuth upgraded Meta from Neutral to Overweight and raised his price target from $640 to $820. A price target reflects an analyst’s view of future value, not revenue Meta has already earned from Muse or a guaranteed stock return. 313

For shareholders, the useful tests are whether people keep using Muse, whether enough choose paid plans, and whether revenue from those users can justify the cost of serving them. Strong download figures answer none of those questions on their own.

What partners and competitors should watch

Reuters identified Shopify and PayPal among the early beneficiaries of Muse’s emergence. An agent that helps users complete purchases could bring partners transactions, while also becoming the interface that decides which options shoppers see. Whether that strengthens merchants’ access to customers or weakens it will depend on how recommendations and checkout work in practice. 17

Competitors face a similar uncertainty from the other side: early popularity shows consumer interest in an AI agent, not a durable advantage over rival assistants. Meta’s defining challenge is to turn that interest into sustained, economically valuable use while carrying a much larger infrastructure commitment. Until then, its spending is evidence of conviction—not proof that the superintelligence bet has paid off. 218