META SHARES FALL AS INVESTORS QUESTION THE MASSIVE COST OF ARTIFICIAL INTELLIGENCE EXPANSION

Meta Platforms shares fell about 3 per cent in US premarket trading on Thursday as investors weighed the strong growth of the company’s artificial intelligence business against the rapidly rising cost of the infrastructure required to support it. The decline came after Meta reported better-than-expected third-quarter revenue and highlighted growing momentum around its new AI products, but also disclosed that free cash flow turned negative during the quarter and that annual capital expenditure is now expected to reach between US$135 billion and US$145 billion. The mixed reaction shows how investor attention is increasingly shifting from whether major technology companies can build powerful AI systems to whether they can do so without placing excessive pressure on cash generation, margins and long-term shareholder returns.

Meta’s third-quarter revenue rose to US$64.6 billion, representing a 26 per cent increase from the same period a year earlier. Advertising remained the main driver of the business, with ad revenue climbing as engagement and pricing improved across Facebook, Instagram, WhatsApp and the company’s other consumer platforms. Family daily active people reached 3.64 billion on average in September, while advertising impressions increased by 10 per cent and the average price per advertisement rose by 14 per cent. Those figures demonstrate that Meta’s core business remains strong even as the company pours unprecedented amounts of money into artificial intelligence infrastructure.

The concern for investors is that the scale of Meta’s spending is growing almost as quickly as the enthusiasm around its AI strategy. Capital expenditure reached approximately US$34.9 billion during the third quarter alone, and the company narrowed its full-year 2026 capital expenditure outlook to between US$135 billion and US$145 billion from a previous range of US$130 billion to US$145 billion. That level of spending reflects enormous investment in data centres, high-performance computing clusters, networking equipment and the specialised chips required to train and operate large artificial intelligence models.

Meta’s free cash flow turned negative by approximately US$2.12 billion during the quarter, compared with positive free cash flow of more than US$10.6 billion in the same period a year earlier. Operating cash flow remained strong at roughly US$32.8 billion, but property and equipment purchases and finance-lease payments consumed more cash than the business generated after operations. That difference is important because free cash flow is closely watched by investors as an indication of how much cash remains after a company funds the infrastructure required to operate and expand its business.

The negative free cash flow does not mean Meta is running out of money. The company ended September with more than US$82 billion in cash, cash equivalents and marketable securities, while its advertising business continues to generate enormous revenue. Instead, the figure illustrates just how aggressively Meta is investing in new computing capacity. Even a company with one of the world’s most profitable digital advertising businesses can temporarily consume more cash than it produces when capital spending reaches tens of billions of dollars every quarter.

Chief Executive Mark Zuckerberg has made artificial intelligence the centre of Meta’s next phase of development. The company has launched Muse and Muse Wonder, which it describes as its first frontier AI models, and is positioning Muse as a consumer artificial intelligence agent capable of operating across its large family of applications. Meta believes combining advanced models with billions of existing Facebook, Instagram, WhatsApp and Messenger users gives it an enormous distribution advantage compared with AI companies that must build audiences separately from their underlying technology.

Muse has already attracted more than 10 million users in the United States and Canada within weeks of launch, strengthening Meta’s argument that AI could become a major consumer product rather than simply an internal tool used to improve advertising and recommendations. The company expects the technology eventually to reach a much larger portion of its user base, and executives believe AI agents could ultimately create new revenue streams alongside Meta’s traditional advertising business. Investors, however, are now demanding greater clarity about how quickly those opportunities can produce returns relative to the billions being spent to build them.

The tension is becoming one of the defining issues across the technology sector. Microsoft, Alphabet, Amazon, Meta and other companies are spending enormous sums on data centres and AI accelerators because access to computing capacity has become a strategic competitive advantage. None wants to risk being left behind if artificial intelligence becomes the dominant technology platform of the next decade, but the scale of those investments has raised questions over whether demand will grow quickly enough to justify the infrastructure being built.

Meta’s case is particularly important because Zuckerberg has previously demonstrated a willingness to continue funding long-term technological bets even when investors become uncomfortable with near-term costs. The company spent heavily for years on virtual and augmented reality through its Reality Labs division, generating large operating losses while Zuckerberg argued that immersive computing could eventually become a major platform. Artificial intelligence has now become an even larger strategic priority, and Meta is redirecting more resources toward it.

Reality Labs itself remains deeply loss-making, reporting an operating loss of more than US$4.1 billion during the third quarter on only US$443 million in revenue. Meta now expects the division’s 2026 losses to total between US$18 billion and US$19 billion, although it says those losses should decline to around US$15 billion next year as resources are increasingly shifted toward AI. That transition shows how the company’s investment priorities have changed, with artificial intelligence now receiving the kind of aggressive financial commitment previously associated with the metaverse.

The company’s research and development expenses reached approximately US$20.1 billion during the quarter, up substantially from the previous year. Some of that spending reflects compensation for highly specialised AI researchers and engineers, who have become among the most expensive technical workers in Silicon Valley. Competition for talent has intensified as technology companies attempt to recruit experts capable of improving model reasoning, training efficiency, agent behaviour and AI infrastructure.

Meta has also been signing long-term agreements for data-centre capacity and electricity supply, creating obligations that may continue for years even if technology demand changes. Advanced AI clusters consume enormous amounts of energy, and companies are increasingly working directly with utilities, energy developers and governments to secure enough electricity for planned facilities. Building the physical infrastructure can require years of planning because transmission lines, substations, cooling systems and specialised facilities all have to be constructed alongside the computing hardware itself.

Those constraints help explain why companies are committing money so far ahead of expected revenue. If Meta waits until customer demand fully materialises before beginning construction, it could find itself without enough capacity to compete with rivals that invested earlier. The company is therefore effectively betting that future AI revenue and efficiency gains will justify building infrastructure before every commercial use case has matured.

Investors are willing to tolerate large spending when they can see evidence that the investment is improving existing businesses. Meta says that is already happening. Its AI systems are improving content recommendations, increasing user engagement and helping advertisers target audiences more effectively. Those improvements can generate immediate financial returns because even small increases in advertising performance become significant across platforms serving billions of people.

Advertising revenue reached more than US$63 billion during the third quarter, illustrating why AI improvements inside the existing business can be financially powerful. Better recommendations can keep users on Facebook and Instagram for longer, while more effective advertising systems can increase the value advertisers receive from each impression. Meta argues that those benefits already demonstrate that its artificial intelligence investment is not purely speculative.

The more difficult question concerns the next layer of investment. Building frontier AI models requires far more computing power than running the recommendation systems that have supported Meta’s advertising business for years. Each new generation of models can require larger clusters, more sophisticated networking and greater electricity consumption. Meta must therefore demonstrate that those costs will eventually produce either new revenue or sufficient improvements in existing businesses to justify the scale of spending.

The launch of Muse represents one attempt to create that new revenue. Meta wants the AI agent to become a broad assistant integrated across its apps, potentially helping users communicate, create content, shop, discover information and interact with businesses. If successful, the system could create advertising, commerce and subscription opportunities at a scale difficult for smaller competitors to match.

Meta’s advantage is distribution. Facebook, Instagram, WhatsApp and Messenger already reach billions of people, meaning the company can place new AI products directly in front of users without spending heavily to acquire them individually. That could reduce one of the biggest commercial challenges facing standalone AI companies, which must first attract and retain audiences before monetising them.

The disadvantage is that integrating AI deeply into existing platforms raises enormous infrastructure requirements because billions of potential users may begin sending queries, generating images or interacting with agents simultaneously. A successful consumer AI product can therefore increase computing costs very quickly. That means popularity itself can become expensive unless the company develops efficient models and a strong monetisation system.

The premarket decline in Meta shares suggests some investors believe this cost question remains unresolved. The stock had initially responded positively after the earnings release as traders focused on the stronger revenue figures and enthusiasm around Muse. By Thursday morning, however, attention had shifted back toward cash burn, infrastructure spending and the possibility that future capital requirements could remain extremely high.

Meta’s shares were down roughly 3 per cent before the opening bell, a move that threatened to erase tens of billions of dollars in market value. The company’s enormous valuation means even a relatively small percentage move can translate into a very large change in shareholder wealth. A 3 per cent decline is therefore significant even though it would not represent a dramatic collapse in the context of normal technology-stock volatility.

The broader market environment also contributed to the pressure. Technology shares were generally weaker on Thursday as investors became more cautious about AI-related spending and rising energy prices. Nasdaq futures declined sharply, while several semiconductor and technology companies also moved lower. Meta’s results therefore arrived during a market session in which investors were already questioning whether the enormous capital spending associated with artificial intelligence would generate sufficient returns.

Higher energy prices create an additional concern because data centres are large electricity consumers. If power costs rise significantly, companies operating massive AI infrastructure could face higher expenses just as capital requirements are increasing. Some technology companies have responded by signing long-term energy contracts or investing in dedicated generation projects to reduce exposure to volatile electricity markets.

Meta’s spending guidance suggests management does not intend to slow down meaningfully in the near term. The company said it expects to continue investing aggressively in 2027 to support both its existing business and emerging artificial intelligence initiatives. Management is still developing next year’s final budget and capacity plans, but the current infrastructure already under construction gives Meta the ability to allocate computing resources between model training, product development and its advertising platforms.

That flexibility matters because computing demand can change quickly. A new model may require enormous training resources for several months before those chips are reassigned to consumer products or advertising systems. Meta therefore wants enough infrastructure to avoid situations where one successful product delays another because the company lacks available processing capacity.

The company also announced plans to establish a new class of publicly listed non-voting shares beginning in January 2027. Meta says the new Class C shares could be used for capital raising, employee compensation, acquisitions and other corporate purposes, giving the company more flexibility to finance growth without changing Zuckerberg’s voting control. The decision has attracted additional attention because it reinforces the expectation that Meta may continue requiring large amounts of capital as the AI competition intensifies.

Meta’s balance sheet remains strong enough to support the investment for now. Its cash and marketable securities provide substantial financial flexibility, while its core applications continue to generate large operating cash flows. The company also has access to debt markets and equity financing if it chooses to raise additional capital.

Long-term investors will nevertheless be watching leverage more closely. Meta’s long-term debt had risen to more than US$80 billion by the end of September, compared with about US$59 billion at the end of 2025. The increase does not indicate immediate financial distress, but it shows that the company is using more financing as its capital programme expands.

Property and equipment on Meta’s balance sheet had climbed to more than US$253 billion by September, up from approximately US$176 billion at the end of the previous year. That sharp increase demonstrates how quickly investment in physical infrastructure is transforming the company. Meta was once viewed primarily as a software and advertising business with relatively light physical assets, but artificial intelligence is pushing it toward a far more capital-intensive operating model.

That shift changes the way investors evaluate the company. Software businesses traditionally attract high valuations partly because they can add users without building an equivalent amount of new physical infrastructure for each customer. AI weakens that advantage because generating responses, images and video requires substantial computing resources every time users interact with the system.

The difference between software and infrastructure therefore becomes increasingly blurred. Meta may still earn most of its money through digital products, but the machines behind those products now require investment on a scale historically associated with telecommunications companies, utilities or industrial businesses.

This does not necessarily make the strategy unattractive. If AI creates a new generation of highly profitable products, the infrastructure could become an enormous competitive barrier protecting Meta from smaller rivals. Few companies can afford to spend more than US$100 billion a year building computing systems, and even fewer have billions of users available to monetise those systems.

The central investor question is timing. Capital expenditures are happening immediately, while some of the expected AI revenue may take several years to arrive. That mismatch can temporarily reduce free cash flow even when the underlying business remains profitable, which is exactly what happened during the third quarter.

Meta earned approximately US$9.86 billion in net income during the quarter, up sharply from the previous year, although comparisons were influenced by unusually large tax charges in the prior period. Operating income was US$12.99 billion, while the operating margin declined to 20 per cent from 40 per cent a year earlier. A large legal charge related to a settlement with US state attorneys general also affected expenses, meaning the headline decline in operating margin does not result solely from AI spending.

The company recorded a US$9.89 billion charge connected to a settlement resolving litigation related to teen mental health and social media. Excluding that charge, Meta said operating income would have been approximately US$22.9 billion and diluted earnings per share would have been US$7.40 rather than the reported US$3.83. Investors therefore need to separate one-time legal costs from the longer-term infrastructure spending when assessing the quarter.

Meta expects fourth-quarter revenue between US$73 billion and US$76 billion, suggesting management remains confident about strong demand through the end of the year. The company also expects overall revenue growth to remain strong into 2027, supported by continued expansion in its advertising business and the growing contribution from artificial intelligence.

That guidance is one reason investor reaction remains more cautious than outright negative. The market is not questioning whether Meta has a profitable core business. The concern is whether the company will continue spending at such an extraordinary rate that cash flow growth becomes weaker even as revenue rises.

Meta argues that the spending should eventually reinforce profitability because artificial intelligence can improve efficiency as well as create products. AI systems can automate coding, moderation, customer support, advertisement creation and internal workflows, potentially allowing the company to operate more effectively with fewer employees. Meta’s headcount had fallen to approximately 66,775 by the end of September, down 15 per cent from a year earlier even as revenue increased strongly.

That combination of lower headcount and higher technology spending demonstrates the structural change occurring across large technology companies. Capital is increasingly being redirected from traditional labour costs toward computing infrastructure and specialised technical talent. Businesses may employ fewer people overall while spending far more on chips, data centres and electricity.

The trend has implications well beyond Meta. Nvidia, semiconductor manufacturers, energy companies and construction firms are all benefiting from the demand created by AI investment. At the same time, shareholders of companies funding that expansion are becoming more selective about how much capital expenditure they are willing to tolerate without visible returns.

The market reaction to Meta therefore fits into a broader debate over whether the current AI boom is producing a sustainable investment cycle or an infrastructure race in which companies feel compelled to spend simply because competitors are doing the same. No large technology company wants to be the one that underinvested if artificial intelligence transforms the economy, but overinvestment could leave the industry with expensive capacity that takes years to generate acceptable returns.

Meta believes demand will justify the buildout. Zuckerberg has repeatedly argued that possessing more computing capacity creates optionality because it allows the company to accelerate research, improve recommendations and launch new consumer products. If a particular model or product becomes successful, Meta can scale it more quickly when the necessary infrastructure already exists.

Investors are effectively being asked to accept lower near-term cash generation in exchange for that strategic flexibility. Some are comfortable with the trade because Meta’s advertising business provides a financial cushion and because its AI products are already showing early adoption. Others appear increasingly concerned that the scale of investment is becoming difficult to justify using current revenue alone.

The negative US$2.12 billion free-cash-flow figure has therefore become symbolic even though one quarter does not determine the company’s financial trajectory. It shows that capital spending has now reached a point where it can consume all the cash generated by one of the most profitable digital businesses in the world during a quarter.

If free cash flow returns to strongly positive levels as revenue grows, concerns may ease. If negative or weak cash flow becomes persistent while capital expenditure continues rising, pressure on management to demonstrate direct AI monetisation is likely to intensify.

Meta’s stock performance over the coming months will therefore depend not only on user growth or advertising demand but on investor confidence in the economics of artificial intelligence. Each quarterly report will be examined for evidence that AI products are producing enough additional revenue or efficiency to offset the cost of the machines required to run them.

The situation also underscores the financial advantage enjoyed by established technology platforms. A smaller AI company generating little advertising revenue would struggle to absorb a quarter with more than US$34 billion in capital expenditure. Meta can do so because its existing business produces tens of billions of dollars in operating cash and because it can access enormous pools of financing.

That strength may eventually help Meta outlast competitors that cannot fund similar infrastructure. However, financial capacity does not automatically guarantee attractive returns, and investors are now signalling that even the largest technology companies will be questioned when spending rises faster than cash generation.

The approximately 3 per cent premarket decline on October 22 should therefore be understood as a debate over capital discipline rather than a rejection of Meta’s artificial intelligence strategy. The company delivered stronger-than-expected revenue, reported large growth across its advertising platforms and highlighted rapid adoption of its new AI products. At the same time, it disclosed negative free cash flow, record capital expenditure and plans to continue investing aggressively.

Those two realities can exist together. Meta may be building an artificial intelligence business capable of becoming one of the world’s largest technology platforms while simultaneously creating short-term financial pressure through the cost of constructing it.

For shareholders, the calculation is whether the first outcome eventually becomes valuable enough to justify the second. That question will remain central as Meta approaches annual capital expenditure of up to US$145 billion and continues transforming itself from a primarily advertising-driven software company into one of the world’s largest owners of AI computing infrastructure.