Meta says AI agent development is not accelerating as planned
According to Techub, citing CryptoBriefing, Meta CEO Mark Zuckerberg said the company’s AI agent development has failed to speed up as expected over the past four months. The remark stands in contrast to his earlier optimism, when he described 2026 as an “acceleration year” for AI and said agents would begin to “really start working” during that period.
The update is notable because Meta has positioned AI agents as an important part of its next product cycle. A slower-than-expected development curve suggests that even the largest technology platforms are still encountering practical bottlenecks in turning generative AI capabilities into reliable, scalable tools. For markets tracking the intersection of AI, digital infrastructure, and platform economics, the message is clear: deployment timelines may remain uneven even when capital commitment is large and executive expectations are high.
$65 billion in annual AI infrastructure spending, but payoff is delayed
Zuckerberg also said Meta is investing $65 billion per year in AI infrastructure. However, the returns from that spending have not materialized on the expected timetable. This frames the issue less as a lack of commitment and more as a mismatch between the pace of investment and the pace of monetization or operational payoff.
The scale of spending matters on its own, but the more important signal is that the return profile is taking longer to emerge. In the current AI cycle, companies are under pressure not only to build models and compute capacity, but also to convert those investments into usable products, measurable efficiency gains, and new revenue streams. Meta’s disclosure indicates that this translation layer remains difficult, even for firms with major platform reach, engineering capacity, and distribution advantages.
Products are launching, but Meta still wants to reduce ad dependence
Meta has already made concrete commitments in this area. The company previously said it would introduce an internal “CEO agent” tool, and in early June it launched Meta Business Agent for business users managing customer interactions. That means the company is not standing still on product delivery. Instead, the issue appears to be that broader and more effective deployment is progressing more slowly than management originally hoped.
The report also said Meta is using a “mother test” to evaluate AI usability, an indication that the company is focusing on whether these tools are intuitive and dependable enough for wider adoption. Strategically, Meta wants AI products to support a business mix that is less dependent on advertising revenue. But if AI agent development continues to lag behind internal expectations, that diversification effort will take longer. In practical terms, advertising is likely to remain a central revenue pillar for an extended period while AI products continue moving through testing, iteration, and deployment.