Meta announced a new round of job cuts on Thursday even as the company accelerated its investment in artificial intelligence infrastructure and research, a seemingly contradictory combination that senior executives defended as a necessary realignment of resources toward the company’s highest-priority bets. The cuts, which were expected to affect several thousand employees across product management, sales, and administrative functions, came just months after a previous restructuring and reflected the intense pressure facing the company from rising AI operating costs, global economic uncertainty, and the significant legal liabilities exposed by this week’s jury verdicts.

OpenAI, in a separate development Thursday, signaled through a company blog post and statements from senior executives that it was pulling back from several product initiatives that had been in development, citing concerns about safety readiness, resource constraints, and what it described as a need to focus on getting its core products right before expanding into new and potentially higher-risk territory. The move represented a notable shift in tone from a company that had been characterized by rapid, aggressive product launches and a willingness to move fast and accept risk as inherent to its mission.

The two developments together captured the complex and rapidly evolving dynamics of the AI industry at a moment when the initial euphoria of the post-ChatGPT era was being tested by the realities of building sustainable, profitable businesses at an enormous and growing cost. Both companies had burned through tens of billions of dollars in the pursuit of artificial general intelligence and consumer AI products, and both now faced questions about whether their resource consumption could be justified by the revenues and strategic advantages those investments were generating.

For Meta, the job cuts represented the latest chapter in what chief executive Mark Zuckerberg had branded the company’s “year of efficiency” strategy, which had actually now stretched across multiple years and multiple restructuring rounds. Meta’s AI ambitions were not in doubt — the company had committed to spending between $60 and $65 billion on AI infrastructure in 2026 alone, a staggering figure that dwarfed the capital expenditure budgets of most industries — but the company was simultaneously trying to demonstrate that it could manage costs in the parts of its business that did not directly support those AI investments.

Analysts who cover Meta said Thursday that the combination of aggressive AI spending and workforce reductions was a coherent strategy even if it appeared contradictory at first glance. The company’s advertising business, which remains its primary revenue source, does not require large numbers of human employees to scale, and the AI investments were aimed at improving ad targeting and content recommendation systems that could generate significantly higher revenue per user. The restructuring was essentially a bet that the AI-driven productivity improvements would more than offset the revenue implications of the job cuts.

The AI industry’s growing cost structure was being significantly worsened by the global energy price spike driven by the Iran conflict. Data centers consume enormous amounts of electricity, and the surge in energy prices that had taken Brent crude above $119 per barrel was flowing directly into the operating cost statements of major AI companies. Several industry analysts estimated Thursday that the energy cost of running large language models and the inference infrastructure that delivers AI products to users had increased by 30 to 40 percent since the beginning of the year, a cost increase that was not fully anticipated in any of the major companies’ original financial plans for 2026.

OpenAI’s decision to pull back from certain product areas reflected a different kind of calculation. The company, which had been moving at breakneck speed to release new models and consumer applications, was facing internal concerns about whether its safety evaluation processes were keeping pace with the velocity of product development. A series of incidents with earlier model releases, in which the systems exhibited unexpected behaviors that had generated negative press coverage and regulatory scrutiny, had created pressure on the company’s safety team to slow down and conduct more thorough evaluations before new releases.

The pullback was also financial in part. Several of the products OpenAI had been developing required sustained investment without a clear near-term path to profitability, and the company’s leadership had reportedly concluded that concentrating resources on its core ChatGPT products and its API business for enterprise customers was a more defensible strategy in a constrained environment. OpenAI was expected to raise additional capital later in the year, and presenting a more disciplined product roadmap was seen as important for maintaining investor confidence at favorable valuations.

The diverging strategies of Meta and OpenAI were being watched carefully by the broader AI industry as signals about where the sector was headed. Meta was doubling down on scale and infrastructure, betting that whoever built the largest and most capable AI systems would ultimately win the market. OpenAI was signaling a period of consolidation and safety-focused development, at least temporarily. Other major players, including Google’s DeepMind and Anthropic, were navigating similar tensions between the competitive pressure to move fast and the practical and reputational risks of doing so.

The legal environment was also a factor in both companies’ strategic calculations on Thursday. Meta was still absorbing the implications of this week’s jury verdicts, which found the company liable for harms caused to teenage users of its platforms. While those verdicts related to consumer social media products rather than AI directly, they signaled a broader shift in the legal and regulatory climate toward holding technology companies accountable for the real-world effects of their products, a precedent with potentially significant implications for AI applications in the future.

For the tens of thousands of technology workers who had built their careers on the assumption of the AI industry’s endless expansion, the news from both Meta and OpenAI on Thursday was unsettling. The job market for software engineers and AI researchers had tightened considerably over the past year as the initial burst of AI-driven hiring gave way to a more sober reassessment of staffing needs. Workers who had joined AI companies in 2023 and 2024 on the expectation of job security and generous compensation were now facing a more uncertain landscape.

The broader question facing the AI industry by late March 2026 was whether the current moment represented a temporary adjustment on the path to transformative growth, or a more fundamental reckoning with the gap between the technology’s extraordinary capabilities and the practical realities of building durable businesses around it. The answer was unlikely to be known for months or years, but Thursday’s developments from two of the sector’s most prominent companies provided clear evidence that the industry’s self-conception as an inexorable force immune to the constraints facing ordinary businesses was in the process of being substantially revised.