Meta Earnings Preview: The AI Spending Test

This Meta earnings preview begins with Meta Platforms reporting second-quarter results after the U.S. market closes on Wednesday, July 29, with Wall Street expecting another strong quarter from the company’s advertising machine. Consensus calls for approximately $60.2 billion in revenue, representing growth of roughly 27%, and earnings of about $7.19 per share. Operating margin, however, is expected to fall to around 36% from 43% a year earlier as AI infrastructure, talent and data-centre costs accelerate.

The obvious debate is whether Meta’s AI-powered advertising improvements can keep revenue growing quickly enough to absorb that spending. But the immediate market reaction may depend on something less visible in the quarterly income statement: whether management can finally explain how its enormous computing footprint becomes a cash-generating asset.

That creates an unusual timing problem. Q2 may confirm that the core business remains healthy, while the next two quarters determine whether Meta is building excessive internal capacity—or the foundations of a new cloud platform.

Meta Earnings Preview: What Wall Street Is Modelling

The base case assumes Meta’s advertising business delivers another clean quarter. In Q1, revenue increased 33% to $56.31 billion as advertising impressions rose 19% and average price per ad increased 12%. Family daily active people reached 3.56 billion, reinforcing the view that Meta still has room to expand engagement, inventory and advertiser returns across Facebook, Instagram and WhatsApp.

AI is already contributing to that engine. Improvements to recommendation systems increased time spent on Reels, while newer advertising models improved conversion rates and campaign performance. Meta is also expanding AI translation, business assistants, WhatsApp commerce tools and advertising across Threads, creating additional inventory without depending entirely on traditional Facebook Feed growth.

The market therefore expects strong revenue, healthy advertising demand and continued evidence that AI is improving Meta’s existing products. The problem is that these expectations are already familiar. Meta has exceeded Wall Street’s revenue and earnings forecasts for eight consecutive quarters, so another modest beat may do little to change the investment debate.

A good advertising quarter is now the entry fee, not the catalyst.

Consensus also expects operating margin to decline to roughly 36%, illustrating how rapidly infrastructure and operating costs are absorbing incremental revenue. Even if revenue reaches a record, nearly flat year-over-year EPS would suggest that much of the additional economic value is being redirected into servers, chips, data centres and specialised AI teams rather than near-term shareholder earnings.

That is why the conventional “beat and raise” framework is incomplete. Meta can exceed Q2 revenue expectations and still disappoint if management increases spending without improving visibility into future returns.

The Metric That Actually Matters

The most important number is Meta’s 2026 capital-expenditure guidance, currently set at $125 billion to $145 billion. Management already increased that range from $115 billion to $135 billion following the first quarter, citing higher component prices and the need to add more data-centre capacity. Investors will now be listening for another increase, any early indication of 2027 spending, and—most importantly—what Meta expects to earn from the capacity being built.

Meta spent $19.84 billion on capital expenditure during Q1 while producing $12.39 billion of free cash flow. Operating cash flow remained formidable at $32.23 billion, but the gap between operating cash generation and infrastructure spending shows why strong advertising growth no longer settles the argument. This Meta earnings preview treats that gap as the central mechanism behind the market’s changing interpretation of the company.

The true earnings pivot is whether Meta can build a credible bridge from record capex to future free cash flow.

That bridge matters because capex creates more than an immediate cash outflow. The resulting assets produce depreciation, energy costs, networking expenses, maintenance requirements and staffing costs over multiple years. If infrastructure arrives ahead of demand, the cash-flow pressure can persist even after the initial construction bill has been paid.

The mechanism cuts directly into valuation. Investors may tolerate lower free cash flow when spending has visible contractual demand, high utilisation and measurable future revenue. They become less tolerant when management repeatedly raises spending while monetisation remains conceptual.

Meta’s advertising business gives it more capacity to fund this investment cycle than most AI challengers. But the market is increasingly focused on return on invested capital, not Meta’s ability to write the cheque. If the company cannot explain how utilisation, external demand and revenue scale with infrastructure, its forward earnings multiple may compress even if reported earnings remain resilient.

The Upside Surprise

The cleanest upside scenario would combine strong advertising results with disciplined infrastructure guidance. Revenue near or above consensus, resilient ad pricing and a Q3 outlook showing continued demand would reassure investors that Meta’s core funding engine remains intact.

The more powerful surprise, however, would come from management putting commercial detail around Meta Compute. Bloomberg reported that Meta is developing a cloud business that could allow developers to access AI models hosted on its infrastructure and potentially purchase raw computing capacity. The plans remain under development, Meta has not confirmed them and the final strategy could change.

Confirmation alone would not be enough. Investors would want to hear what Meta intends to sell, who the likely customers are, when external revenue could begin and whether the company sees the offering as occasional capacity monetisation or a durable enterprise platform. In this Meta earnings preview, credible commercial detail is what separates a strategic option from a marketable business.

A reported computing agreement with Anthropic could help make that discussion more tangible. Meta is reportedly exploring a transaction worth as much as $10 billion over two years, which would suggest that at least part of the infrastructure programme could be supported by identifiable third-party demand rather than purely internal forecasts.

The bullish narrative changes when spare compute stops looking like waste and starts looking like inventory.

The recently announced El Paso venture with BlackRock offers another potential signal. BlackRock-managed funds will own 80% of the approximately $14 billion project, while Meta will retain 20% and lease capacity from the venture. The structure gives Meta access to a planned one-gigawatt campus without directly owning and financing the entire development.

That does not eliminate the economic obligation; leases still carry costs. But it demonstrates that Meta is exploring financing structures that distribute construction risk, preserve balance-sheet flexibility and align infrastructure delivery with future capacity needs.

The psychology around the stock could change quickly if management holds the current capex range while providing credible evidence of external demand. The initial report about a cloud business drove Meta shares more than 10% higher while pressuring CoreWeave and Nebius, showing that investors are prepared to reward a narrative in which Meta becomes a supplier of AI infrastructure rather than merely one of its largest customers.

In that scenario, the market would be less focused on whether free cash flow temporarily turns negative and more focused on whether Meta Compute can eventually create a second monetisation layer alongside advertising.

The Downside Surprise

The downside scenario begins with another capex increase that is not accompanied by better commercial visibility. A range moving materially above $145 billion—or an early 2027 outlook implying another step-change—would reinforce the fear that Meta’s infrastructure requirements are still being discovered rather than managed. The downside case in this Meta earnings preview becomes more severe if spending rises before external demand is clearly established.

The risk becomes more serious if advertising trends also soften. Revenue growth near consensus may not be sufficient if average price per ad decelerates, engagement improvements fail to produce corresponding monetisation, or Q3 guidance suggests that advertising demand is normalising faster than expected.

That combination would create negative operating leverage: infrastructure costs would rise while the advertising business funding them slows. Margins could fall faster than currently modelled, and free-cash-flow pressure could deepen before new AI products or external compute revenue reach meaningful scale.

The real downside is not high spending by itself; it is high spending with no observable utilisation curve.

Management could also disappoint through vagueness. If executives decline to discuss the reported cloud plan, avoid giving any indication of external demand or frame the infrastructure buildout almost entirely around long-term superintelligence ambitions, investors may conclude that commercialisation remains too distant to underwrite.

Meta’s history makes that sensitivity understandable. The company is already funding large losses at Reality Labs, which generated a Q1 operating loss of approximately $4.03 billion. Investors may tolerate one long-duration investment programme when the core business is strong; simultaneous open-ended spending across AI infrastructure, advanced models, smart glasses and the metaverse creates a more difficult valuation equation.

The market may also be underestimating the distinction between owning computing capacity and operating a competitive cloud platform. Selling raw GPU access requires pricing, scheduling, customer support and utilisation management. Selling hosted models requires developer tools, security, reliability, enterprise contracts and an ecosystem strong enough to compete with AWS, Microsoft and Google.

A rushed cloud strategy could therefore create revenue without generating hyperscaler-quality margins. Worse, external sales could be interpreted as evidence that Meta overbuilt capacity for its own needs and is attempting to recover part of the cost after the fact.

In that outcome, the stock’s forward multiple could compress even if Q2 revenue and EPS exceed consensus. Investors would be discounting a longer period of capital intensity, lower free-cash-flow conversion and greater uncertainty around terminal returns.

What Matters Over The Next Two Calls

The first signal to monitor is whether Meta distinguishes between infrastructure required for its advertising systems and capacity intended for frontier-model development. Those workloads have different risk profiles. Advertising infrastructure supports an established, measurable revenue stream, while frontier-model spending depends on products and capabilities that may take longer to monetise.

The second is utilisation. Investors need evidence that new capacity is being absorbed by recommendation models, generative advertising, Meta AI, business messaging, smart glasses or external customers. Commentary around customer interest, contracted demand and model-serving volumes would be more informative than another broad statement about the strategic importance of AI.

Over the next six to twelve months, utilisation must become a reported commercial story rather than an implied technical achievement.

The third signal is financing discipline. The BlackRock venture suggests Meta may increasingly use joint ventures, leases and external capital to secure infrastructure. Investors will need to assess whether these structures genuinely improve risk-adjusted returns or simply move obligations away from headline capex.

Depreciation will also become increasingly important. Even if annual capex eventually stabilises, prior spending will continue flowing through the income statement. That means margin pressure could persist after the peak construction period unless AI-driven revenue scales fast enough to absorb the expense.

The fourth signal is the quality of advertising growth. Meta’s case remains far stronger when higher impressions are accompanied by better conversion, rising advertiser returns and durable price per ad. Growth driven primarily by additional ad load would be less reassuring because it could face user-experience limits.

Finally, investors should watch the competitive response. A serious Meta cloud platform could pressure CoreWeave and Nebius by reducing Meta’s dependence on their capacity while introducing another large supplier. But Meta would also be entering markets where AWS, Microsoft and Google possess established enterprise relationships, mature software layers and years of operational experience.

This Meta earnings preview therefore extends beyond one quarter: the next two earnings calls should reveal whether Meta is attempting to challenge that full stack—or simply monetise selected models and surplus capacity.

Conclusion

Meta’s second-quarter results may confirm that its advertising franchise remains one of the most productive cash-generating businesses in global technology. That alone may not determine the market reaction.

The interpretation of capex will matter more than the absolute figure. Higher spending supported by improving utilisation and measurable commercial returns could be received more favourably than stable guidance accompanied by limited evidence of progress. Meta’s increase in contractual commitments primarily reflects its own cloud and infrastructure purchases rather than contracted external demand, while management has not disclosed utilisation rates or a precise return framework.

The central question is whether Meta’s AI investment is strengthening its existing advertising ecosystem or developing into a broader revenue platform. Management has outlined a progression from model quality and product adoption to monetisation and cost efficiency. Early gains in advertising performance and business-AI usage are encouraging, but standalone AI monetisation remains unproven.
Meta’s trailing valuation has already de-rated materially. As of 28 July 2026, it traded at 7.04x LTM revenue, 13.84x EBITDA, 17.08x EBIT and 21.59x diluted earnings, well below its June 2025 levels.

Greater visibility on monetisation, capacity utilisation and investment returns could support multiple expansion despite weaker near-term free-cash-flow conversion. Without it, investors may continue applying a discount for rising capital intensity and uncertain returns. The Q2 print will provide the first clues; the following quarters will show whether they develop into measurable revenue and a defensible return on the largest investment programme in Meta’s history.

Disclaimer: We do not hold any positions in the above stock(s). Read our full disclaimer here.