AIToday
Large Language ModelsOpen-Source AIAI Business & IndustryYahoo Finance AIPublished: Aug 12, 2026, 10:00 JST5 min read

Zuckerberg's AI Manifesto: Meta Gives Away Models, Keeps the Ad Business

Zuckerberg's AI Manifesto: Meta Gives Away Models, Keeps the Ad Business

Key takeaway

  • Mark Zuckerberg published a letter Monday committing Meta to releasing superintelligence technology openly rather than restricting it to a few institutions, backed by the immediate release of Muse Glimmer models and plans to publish the weights of Muse Spark 1.2.

  • The strategy lets Meta spend $130 billion to $145 billion on 2026 capital expenditures and data-center build-out while preserving its core profit engine—advertising—which grew 28% in the second quarter to $60.8 billion.

  • Investors remain unconvinced; the stock trades at 18.5 times forward earnings as costs grow faster than revenue.

3 Key Points

  1. What happened

    Meta CEO Mark Zuckerberg published a letter Monday titled "The Future is for Everyone," stating superintelligence should not be restricted to a few institutions. The same day, Meta released Muse Glimmer, a family of open-weight models small enough to run on a laptop, with plans to release the weights of Muse Spark 1.2 for public download.

  2. Why it matters

    Meta is spending $130 billion to $145 billion on capital expenditures in 2026, with second-quarter capex alone at $31.1 billion. The strategy bets that giving away model weights—which carry little revenue potential—while retaining control of its advertising system, user base, and data will keep the ad business profitable enough to fund the build-out. Second-quarter revenue grew 28% year over year to $60.8 billion, driven by 14% growth in ad impressions and 12% growth in average price per ad.

  3. What to watch

    The market remains skeptical; Meta shares trade at about 18.5 times forward earnings, down about 10% in 2026 and about 25% from their 52-week high. Third-quarter guidance of $61 billion to $64 billion signals growth is moderating from the first quarter's 33%, raising the question of whether ad growth can sustain the scale of capital spending implied by the letter.

In Depth

Read the full story

On Monday morning, Mark Zuckerberg posted a letter to Meta's newsroom titled "The Future is for Everyone," arguing that superintelligence should not be "centralized and restricted to a few institutions." He framed the stakes plainly: "The defining questions of our age are who will have access to superintelligence and what will we direct it toward." His answer was to distribute the technology as widely as possible. That same day, Meta released Muse Glimmer, a family of open-weight models small enough to run on a laptop, with an explicit commitment to release the weights of Muse Spark 1.2, the company's most advanced model, so the public could download and use the underlying calculations freely.

The announcement raises a natural question: is Meta giving away billions of dollars in AI development to shareholders' detriment? Zuckerberg's letter suggests not. He argued that superintelligence should not be kept behind closed doors and that "distribution is both the safety plan and the business plan." The strategy hinges on a distinction between what Meta gives away and what it keeps. The models themselves are free, but the surrounding infrastructure—computing power accessed through what Zuckerberg called a "dynamic auction mechanism," along with Meta's advertising system, user base, and proprietary data—remains private. Meta's revenue still comes "almost entirely from advertising." By giving away model weights that carry little revenue potential while retaining control of the ad stack, Meta can present itself as the champion of open AI while protecting its core profit engine.

The bill for this strategy is staggering and accelerating. Meta entered 2026 projecting $115 billion to $135 billion in capital expenditures, raised the range to $125 billion to $145 billion in April, and lifted the floor again in July to $130 billion to $145 billion. Second-quarter capital spending alone reached $31.1 billion, more than half again the $19.8 billion Meta spent in the first quarter. Free cash flow collapsed accordingly, from $12.4 billion in Q1 to $784 million in Q2, with most cash reinvested into the build-out. Second-quarter net income fell 14% year over year to $15.8 billion as total costs surged 55%; the decline also reflected $2.4 billion of legal charges and severance costs tied to May job cuts. Meta paid $1.35 billion in dividends that quarter and repurchased no stock.

So far, advertising has kept pace. Second-quarter revenue grew 28% year over year to $60.8 billion, driven by 14% year-over-year growth in ad impressions and a 12% increase in the average price per ad. Management still expects full-year operating income to exceed 2025's level. But growth is moderating visibly: the second quarter's 28% expansion decelerated from the first quarter's 33%, and third-quarter guidance of $61 billion to $64 billion suggests further slowdown ahead. The market remains skeptical. At about $597 per share as of the article's writing, Meta stock is down about 10% in 2026 and about 25% from its 52-week high. Shares trade at about 18.5 times analysts' earnings expectations for the coming year and about 22 times current earnings—a significant discount to Nvidia, which trades at nearly 34 times earnings. Investors are pricing in the risk that the arithmetic above—costs growing 55% against revenue growing 28%—cannot hold for the years of spending the letter implies are still ahead.

Context & Analysis

Zuckerberg's letter reframes Meta's enormous AI spending as a deliberate strategy rather than a cost burden. By committing to open-weight model releases, Meta positions itself as the champion of democratic access to superintelligence—a narrative that also justifies years of capital-intensive data-center expansion. The letter's governance detail—independent board approval of safety criteria—adds a layer of institutional credibility to the bet.

The economics depend on a specific assumption: that the models themselves have limited direct revenue potential, while the infrastructure supporting them (and the user data flowing through it) remains Meta's irreplaceable moat. The strategy works only if advertising can keep growing fast enough to fund the build-out. In the second quarter, that test passed—28% revenue growth on a $60.8 billion base is substantial. But deceleration is already visible: second-quarter growth of 28% slowed from the first quarter's 33%, and third-quarter guidance of $61 billion to $64 billion hints at further moderation.

Market pricing reflects this tension. At 18.5 times forward earnings, Meta trades at a discount to Nvidia (nearly 34 times), but the discount exists precisely because investors doubt the ad business can sustain the costs ahead. The arithmetic is stark: costs grew 55% in the second quarter against 28% revenue growth. Zuckerberg's letter settles the narrative—but not the market's skepticism about whether the math works over time.

FAQ

What models did Meta release, and can anyone use them?
Meta released Muse Glimmer, a family of open-weight models small enough to run on a laptop, with plans to release the weights of Muse Spark 1.2 so the public can download the calculations that determine how the model behaves and use them freely.
How much is Meta spending on AI infrastructure?
Meta expects $130 billion to $145 billion of capital expenditures in 2026, with second-quarter capex alone at $31.1 billion, up from $19.8 billion in the first quarter.
Is the strategy hurting Meta's profitability?
Second-quarter net income fell 14% year over year to $15.8 billion as total costs rose 55%, though revenue still grew 28% year over year to $60.8 billion. Free cash flow shrank from $12.4 billion in the first quarter to $784 million in the second, as most cash went into the build-out.
Yahoo Finance AIRead Original Article

Get the latest Large Language Models news every morning

AI-summarized, only the topics you pick — one digest a day via Email, Slack, or Discord.

Free · takes 30 seconds · unsubscribe anytime

Ask AI

Ask AI anything about this article. Q&As are published on this page for other readers too.

Related Articles

Next articleGlobal youth unemployment rose to 12.4%, UN agency reports

The AI news that matters, in one minute each morning.

Sign up free