Meta conference call: Full-year capital expenditure raised by another $10 billion, reaching a record high; commercialization path for AI agents remains unclear

Meta raises its capital expenditure to as much as $145 billion, reaching a historic high, and for the first time ever, daily active users see a quarter-over-quarter decline.

After the US market closed on April 29, Meta released its Q1 results. Although revenue exceeded expectations, the company raised its full-year capital expenditure cap to $145 billion, $10 billion higher than its guidance three months earlier, sparking sudden investor concerns about the ROI prospects of AI investment.

This is the largest capital expenditure plan in Meta's history. CFO Susan Li said the increase was mainly due to rising hardware component costs and data center expansion.

Meanwhile, Meta’s daily active users saw the first quarter-over-quarter decline ever, dropping from 3.58 billion last quarter to 3.56 billion. The company attributed this to internet outages in Iran and restrictions on WhatsApp access in Russia.

In the subsequent analyst conference call, CEO Mark Zuckerberg focused on the value AI brings to the advertising business and the long-term commercial potential of AI agents but did not give a specific timetable for ROI.

First Ever Decline in Daily Active Users—Main Causes: Iran and Russia

Meta's daily active users decreased by about 20 million quarter-over-quarter, marking the company's first such decline.

Susan Li pointed to two external factors: Iran’s internet outage and Russia’s restrictions on WhatsApp access. She said that excluding these impacts, user growth would be positive, but did not disclose exact numbers.

Meanwhile, advertising performance was relatively steady. The average price per ad increased 12% year-over-year for Q1.

Susan Li noted that total global video watch time on Facebook increased over 8% quarter-over-quarter, the largest single-quarter growth in four years. Instagram Reels also saw an increase in watch time. The main driver behind this improvement is the AI ranking model.

Ongoing AI Spending Expansion; Major Layoffs in May

Meta raised the upper range of its capital spending for 2026 to $145 billion, $10 billion higher than the guidance three months ago.

Zuckerberg and Susan Li both cited higher component prices and data center expansion as main reasons. Meta has also signed multi-year cloud service agreements to speed up capacity deployment.

Susan Li said Meta has begun planning for its 2027 compute needs and admitted the company has historically underestimated the scale of compute required, forecasting compute will play a more central role in future business.

On cost offsetting, Meta recognized $8.03 billion in income tax benefits for Q1, partially offsetting the $15.93 billion non-cash tax expense booked for Q3 2025 due to “One Big Beautiful Bill Act” legislation.

To offset massive AI spending and increase efficiency, Meta is making internal structural adjustments.

The total number of employees this year has already dropped by about 1%, and a 10% layoff is planned for May. Based on the current total personnel of 77,986, this affects about 7,800 people.

On further layoff concerns, Susan Li said the company is unsure of the optimal future headcount and will continue to assess its structure in coming years.

Zuckerberg pointed out that AI is enabling employees to build out products in weeks that previously took months. He said:

We hope to streamline teams; in the future, the key is to build products that are much more numerous and much higher quality than at any time ever.

AI Agents Become Strategic Focus, Commercialization Still Unclear

In the analyst conference call, AI agents were the topic Zuckerberg and Susan Li discussed most.

Zuckerberg stated, Meta aims to create a personal AI assistant for every user. Zuckerberg emphasized:

Others in the industry say AI will replace humans. But I think it will augment human capability, helping people achieve their ambitions. The future will see people becoming more important, not less.

He believes that as AI agents proliferate, entrepreneurship will see “massive” growth. On current AI products, Zuckerberg admitted:

Many agents on the market aren’t good enough—I wouldn’t recommend them to my own mom.

He also revealed:

We’re already training more advanced models.

On hardware, Meta is working to upgrade AI glasses from a simple “answering questions” tool to “portable AI agents.”

Susan Li confirmed demand for new AI glasses with longer battery life and better video capture capabilities is strong, and sales are achieving “robust growth.”

Internally, Susan Li said employees are now using AI agents to assist in product development and code writing, with tasks that previously took months or years now being completed in weeks.

Moreover, weekly conversations enabled by business AI now exceed 10 million, a huge leap from 1 million at the start of the year.

However, when analysts pressed for a commercialization timetable, Susan Li said there are clear future pay-to-use business opportunities for agents, but “there’s nothing more to share today.”

Bloomberg Intelligence’s Mandeep Singh noted Meta’s in-house model Muse Spark still lags other cutting-edge labs in user engagement, with most compute resources still consumed by internal tasks.

Pressed by analysts about when AI investments will yield healthy ROI, Zuckerberg said the current focus is on building leading models and products. He said:

As long as Meta builds great products, ROI will naturally happen.

Full transcript of Meta's Q1 2026 conference call (AI-assisted translation):

Meta management present at this conference call:CEO Mark ZuckerbergCFO Susan Li

Zuckerberg spoke first, opening with Iran’s network outages and Russia’s WhatsApp restrictions.

He described the launch of Muse Spark as Meta's “most important” release so far this year, and stated: "We're on track to build a world-class lab." Muse Spark is the first model from Meta Superintelligence Labs.

"We’re already training more advanced models."

Meta's goal is to create a personal assistant for everyone. Zuckerberg said his vision for AI differs from others in the industry.

He said some believe AI will replace humans, but he believes AI will augment human capability and help people achieve their ambitions. "In the future, people’s importance will only grow, not diminish."

Zuckerberg believes entrepreneurship will see "significant" growth as people gain AI agents, and AI-enabled conversations continue to rise.

He also said AI is boosting Meta's advertising business.

Meta is working to deeply understand users’ interests and each content piece's attributes in the system, then match them accurately. Zuckerberg added that Meta can also proactively generate content for users.

He stressed that AI is improving content recommendation and algorithm performance.

Zuckerberg said upgrades in content algorithms and ad targeting have driven Meta's increased investment in AI infrastructure.

However, he also noted costs have risen, partly due to higher hardware component prices.

Zuckerberg expressed excitement about glasses evolving from simple Q&A tools to portable AI agents.

Zuckerberg finished, and CFO Susan Li took over.

Zuckerberg mentioned AI is helping employees complete work in weeks that used to take months. He hopes to streamline teams, motivate top talent, and increase overall efficiency, believing that the future core is to build more products with higher quality, exceeding previous limits.

Susan Li said, excluding the impact of Iran and Russia, user growth would have been positive, but provided no specifics.

Susan Li stated that average ad price rose 12% year-over-year in Q1.

Meta’s employee count has dropped 1% this year. Also, the 10% layoff planned for May is proceeding.

Susan Li referred to tax information in the earnings release:

Meta recognized $8.03 billion in income tax benefits in Q1 2026, partially offsetting the $15.93 billion non-cash tax expense booked in Q3 2025 due to the “One Big Beautiful Bill Act.”

Susan Li pointed out two core factors driving revenue: increasing user engagement, and monetizing that engagement.

She noted ongoing improvement in content recommendation. Facebook global video watch time rose over 8% quarter-over-quarter—the highest single-quarter growth in four years. Instagram Reels also registered higher viewing time. She credited AI ranking models—helping users quickly see content they’re interested in—for the results.

Susan Li said 30% of recommended Reels are “same-day” content, showing Meta is identifying, categorizing, and pushing content to users faster than ever.

Susan Li stated that Muse Spark delivered significant engagement increases even during pre-launch testing. Since Muse Spark launched, Meta AI chatbot sessions have shown “double-digit percentage growth.”

This data shows the model is performing strongly and driving up user activity.

Susan Li said hundreds of millions of users browse ads every day in the “Status” feature—ads placed in WhatsApp.

She said over 8 million advertisers have used at least one of Meta’s generative AI tools, especially prevalent among SMBs.

Business AI-enabled conversation volume currently exceeds 10 million per week, up from only 1 million at the start of the year.

Susan Li said employees are using AI agents internally for product development and brainstorming.

Meta plans to expand data center scale to secure future compute reserves, also signing multi-year cloud service agreements to speed up deployment and product launches. The goal is rapid development of AI agents for individuals and businesses.

Susan Li spoke on the upcoming May layoffs, stating it would help boost operational efficiency.

Susan Li said rising hardware component prices and increased data center operational costs are main reasons for higher capex guidance.

Susan Li finished. The call moved to analyst Q&A.

Q1 (about spending): What metrics will you focus on in the next 12-24 months to judge whether these investments yield good ROI?

Zuckerberg replied that Meta wants to stay on track building leading models and products—first build the products, then scale, then monetize. He admitted there isn’t a precise monthly growth plan for every product, but he has a clear sense of overall direction.

Zuckerberg said Meta will track the expansion effect of the next batch of training runs, comparing with industry benchmarks.

Q2: After Muse Spark’s launch, how will the team balance focus on current versus future products? Also, any outlook on 2027 capex?

Zuckerberg said the strategic focus remains: launch the first model, then train successive, more advanced models, repeating the cycle.

At the product level, stronger models allow teams to develop in-house, without relying on external models/APIs, greatly expanding product scope.

Susan Li said Meta has started planning for 2027 compute needs. The company has often underestimated the compute required, predicting compute will become more central to future business ops.

Q3: What opportunities do AI agents create in business and SMBs?

Susan Li replied that the top priority is deepening interactions with existing users; helping individuals and businesses improve efficiency is another core direction, but some goals will take time.

She said future agent monetization opportunities—paid services—will become clear, and Meta will consider more value-added services for enterprises, but has no further details at present.

Q4: EssilorLuxottica owns many brands. Beyond Ray-Ban and Oakley, will there be more smart glasses product lines? Also, can you elaborate on the rationale behind the 10% layoff?

Susan Li: The company doesn’t yet know the optimal headcount. Meta hopes to harness AI tools to boost productivity, develop more products/services, and will continue to evaluate its structure over the next few years.

Sales are “robust,” demand for new products with longer battery life and better video quality is strong, and interest in the Meta Ray-Ban display model is high. This remains a key focus area with high expectations.

Q5: What new possibilities has Muse Spark brought to product development?

Zuckerberg said the field is developing rapidly. Meta Superintelligence Labs (MSL) produced quality models impressively quickly, proving the big bets are on the right track.

On the pace of product iteration, Zuckerberg said it's hard to comment: part is competitive sensitivity, part is the team’s focus on quality over deadlines.

He said many AI agents already exist, but few meet the quality standard he’d recommend to family and friends—that’s the target Meta aims for.

Q6: With new consumer applications emerging in the industry, how do these trends affect Meta’s AI agent strategy?

Zuckerberg replied that agents like OpenClaw show vast possibilities, but are still rough and need more operational steps. Meta needs to refine these into “sleek, smooth” products truly fit for the mass market. Meeting this goal will bring agents into the daily lives of billions.

Zuckerberg went on, saying Meta wants products to understand user goals, then help them realize these goals. Everyone in the world will want some form of such service—agents as assistants supporting life tasks.

"Can Meta build the ‘out of the box’ agent?" This is the core challenge now.

Zuckerberg said without leading models, it’s impossible to build leading AI products. Models still learn from human data, but one day will achieve self-improvement.

Q7: Are personal AI agents a short-, mid-, or long-term goal? When will users experience them?

Zuckerberg said if agents land successfully, initial products will appear in the short term, and as technology iterates, more advanced agents will have huge room for growth. Each model generation unlocks more capabilities; this is now the industry’s most exciting time.

This year lays the foundation, and ongoing model evolution will be a long-term process.

Susan Li commented on ranking/recommendation technology improvements, saying Facebook and Instagram have room to grow in this area.

Meta is working on better content search tech to show users broader content, covering more interest dimensions.

Final question: What opportunities do new AI models and agents offer in shopping/commercial fields? What growth areas exist for the core ad business?

Zuckerberg responded: Meta's direction differs from others—optimizing the entire tech stack leads to better agent performance. Meta’s core logic is: personal interests/preferences and humans as such will become more important in the future.

He added, societal progress comes from everyone pursuing their ambitions—some grand, some ordinary. Meta’s positioning of AI as “assistant” sets it apart: not to replace humans, but empower them.

Conference call ended.

Risk Warning and DisclaimerThe market carries risks; invest cautiously. This article does not constitute individual investment advice, nor does it take into account the unique investment goals, financial status, or needs of individual users. Users should consider whether any opinions, viewpoints, or conclusions herein are suitable for their specific situation. Investing based on this article is at your own risk.