Meta Muddles AI Progress: Chip Project Mishaps and Metaverse Ambitions Inflate Costs

Despite splurging on AI research, Meta Platforms, Inc META has struggled to adopt expensive AI-friendly hardware and software systems for its primary business.

By the end of summer 2022, Meta CEO Mark Zuckerberg gathered his top lieutenants for a comprehensive dissection of the company's computing capacity, focused on its ability to do cutting-edge artificial intelligence work, Reuters cites a company memo dated September 20.

"We have a significant gap in our tooling, workflows, and processes when it comes to developing for AI. We need to invest heavily here," said the memo by the new head of infrastructure Santosh Janardhan.

Meta failed to acknowledge the AI-related challenges, including capacity crunches, leadership changes, and a scrapped AI chip project.

The revamp doubled Meta's capital expenditures to ~$4 billion a quarter while it battled a severe financial squeeze, laying off employees since November at a record scale. It led to the dumping of previously planned data center builds in four locations. 

Meanwhile, Microsoft Corp MSFT-backed OpenAI's ChatGPT's popularity triggered an arms race among tech giants to release generative AI products.

Generative AI consumes reams of computing power, amplifying the urgency of Meta's capacity scramble. Meta's late embrace of the graphics processing unit, or GPU, for AI work, posed one of the reasons behind its trauma.

Also, Meta's pivot towards metaverse further slowed down its ability to deploy AI to respond to hazards like the rise of TikTok and Apple Inc AAPL-led ad privacy changes.

After shelving a large-scale rollout of its custom inference chip, Meta placed orders for billions of dollars worth of Nvidia Corp NVDA GPUs in 2022, followed by the departure of several executives left Meta trailing behind peers like Alphabet Inc GOOG GOOGL Google.

Meta started retooling its data centers to acclimate the incoming GPUs. After that, it planned to develop a new and more ambitious in-house chip, which, like a GPU, can train AI models and perform inference, likely to finish around 2025.

Chief Technology Officer Andrew Bosworth likewise said this month that generative AI was where he and Zuckerberg spent the most time, forecasting Meta would release a product this year.

Price Action: META shares traded lower by 1.66% at $209.16 on the last check Tuesday.