AI "chasers" encounter obstacles: Zuckerberg delays new model, Musk restructures xAI "from scratch"

```

Two tech giants that entered the AI arms race in a challenger role are simultaneously experiencing internal turmoil.

On Thursday, March 13, Musk admitted on the X platform that "the initial build of xAI was flawed and is being completely rebuilt from the ground up." According to the Financial Times, the company is also undergoing a new round of large-scale personnel restructuring, with only a few members of the original founding team remaining.

Meanwhile, according to The New York Times, Meta has quietly postponed the release of its next-generation core AI model until at least May, as the model underperformed Google’s Gemini 3.0 in crucial benchmarks such as reasoning, programming, and writing.

Both companies are lagging in the critical commercial track of AI programming tools, a field now seen as the most core revenue source for AI labs.

Meta’s capital expenditure guidance this year is as high as $115–135 billion, nearly double last year’s, raising doubts about whether its core bets will pay off. xAI faces pressure regarding its prospects for listing after merging with SpaceX, and a faltering AI division is not the story Musk wants investors to see.

Meta’s core foundation model delayed, internal testing behind Google’s Gemini 3.0

Meta’s newly established "TBD Lab" is responsible for developing a range of fruit-codenamed AI products: Avocado as the core foundation model, Mango focused on image and video generation, and a larger-scale "Watermelon" project under planning.

According to The New York Times, the AI model "Avocado" had originally been scheduled for release this month, but due to unsatisfactory internal test results, has been delayed until at least May.

Testing showed the model performed poorly in key benchmarks such as reasoning, programming, and writing—better than Meta's previous-generation product and Google's Gemini 2.5 released in March, but at a disadvantage compared to Google’s Gemini 3.0 released last November.

According to sources cited in the report, Meta is even discussing temporarily licensing models from competitors like Google to support its own AI products and maintain competitiveness, though no decision has been finalized.

If Meta ultimately chooses to incorporate Google’s Gemini to support its products, it would diverge somewhat from Zuckerberg’s previous public emphasis on independent R&D and attaining AI leadership.

This year, Meta’s capital expenditure guidance is $115–135 billion, most of which is devoted to AI data centers, computing clusters, and infrastructure.

Reports say the company has also made a near $600 billion long-term commitment to future U.S. market investments, spent $14.3 billion to acquire a stake in Scale AI, and appointed its CEO Alexandr Wang as Meta’s Chief AI Officer.

Zuckerberg has publicly promised these investments will push Meta "to the frontier" and toward superintelligence.

xAI "rebuilding from scratch", co-founders leave one after another

Wallstreetcn noted that of xAI’s original 11 co-founders, only Manuel Kroiss and Ross Nordeen remain.

This week, co-founders Zihang Dai and Guodong Zhang left one after the other. Both are Chinese; Dai had publicly admitted xAI lagged in programming ability, while Zhang was responsible for pre-training the Grok model and is considered mainly responsible for programming shortcomings.

Previously departed co-founders include Greg Yang, Tony Wu, Jimmy Ba, among others, and the personnel upheaval reflects systematic restructuring.

According to reports, senior executives from SpaceX and Tesla have entered xAI as "fixers", conducting employee audits focusing on data quality issues in model training, and dismissing those deemed underperforming.

At the same time, Musk himself is proactively expanding recruitment channels. He publicly apologized on X to candidates previously rejected, stating that historical rejection records will be reviewed and promising candidates contacted.

Currently, xAI has over 5,000 employees, fewer than OpenAI’s more than 7,500 but more than Anthropic’s roughly 4,700.

Programming tools become a key battleground, business pressure drives xAI restructuring

In the AI industry, programming tools are widely seen as the most commercially viable path at present.

Grok’s programming functionality at xAI is currently lagging significantly behind Anthropic’s Claude Code and OpenAI’s Codex. Musk has acknowledged this gap, setting "catching up to competitors" as a mid-year target at an all-hands meeting this week.

This week, xAI poached Andrew Milich and Jason Ginsberg from popular AI coding app Cursor to boost the capability of its "Grok Code Fast" product.

Musk is also betting on a longer-term vision. xAI’s "Macrohard" project aims to create AI agents that can fully replace white-collar work.

But according to Business Insider, the project is now on hold, and its initial lead Toby Pohlen has already left just 16 days after joining.

Musk revealed this week that Macrohard will be jointly advanced with Tesla’s "Digital Optimus" project, with xAI’s language models powering Tesla AI agents to perform tasks. Ashok Elluswamy from Tesla will now oversee the project’s rebuild.

On the infrastructure front, xAI’s Memphis supercomputing cluster now has over 200,000 GPUs deployed and plans to expand to 1 million, leveraging X platform data for unique scale and real-time advantages in model training.

However, external pressures are significant too. With xAI merging into SpaceX in a $1.25 billion deal and SpaceX’s potential IPO window as early as June this year, this money-burning AI division needs to show substantial Grok user growth and commercialization progress to outside investors.

Risk reminder and disclaimerThe market has risks, and investments should be cautious. This article does not constitute individual investment advice, nor does it take into account the particular investment objectives, financial situation, or needs of any individual user. Users should consider whether any views, opinions, or conclusions in this article fit their own circumstances. Investing accordingly is your own responsibility. ```