Mark Zuckerberg and Elon Musk bet big on data centers as AI models lag behind

Mark Zuckerberg and Elon Musk bet big on data centers as AI models lag behind Both tech billionaires are pouring hundreds of billions into physical infrastructure while their AI models struggle to keep pace with competitors Share Add us on Google by Editorial Team Jul. 7, 2026 Meta and xAI are channeling staggering sums into data center infrastructure, betting that raw computational power will eventually close the gap their AI models haven’t managed to bridge on their own. The infrastructure arms race Meta’s marquee project is Hyperion, a facility in Louisiana that, when completed, will consume more electricity than the entire city of New Orleans.

The physical footprint is equally absurd. Hyperion is expected to cover an area comparable to lower Manhattan. Zuckerberg has said Meta plans to invest “hundreds of billions” into AI data centers.

Then there’s Prometheus, Meta’s Ohio-based facility projected to go online in 2026. Advertisement xAI’s Colossus supercomputer was constructed in just 122 days inside a repurposed factory. The plan is to scale Colossus to accommodate one million GPUs, backed by strategic acquisitions in the energy sector.

Where the models fall short For all the infrastructure muscle, the AI models themselves tell a less triumphant story. Meta’s Llama series and xAI’s Grok have been released in successive iterations, but neither has managed to consistently outperform competitors like OpenAI’s GPT series or Google’s Gemini in ways that matter to developers and enterprise customers. The challenges aren’t necessarily about the models being broken.

They’re about cost, efficiency, and the increasingly crowded field of foundation model providers. Community pushback and energy concerns Both Meta and xAI face growing local opposition in the communities where these data centers are being constructed. Residents are raising concerns about energy consumption, emissions, and the strain on local power grids.

What this means for investors Companies providing energy solutions, construction services, and specialized data center technology stand to benefit from what could be a multi-year, multi-hundred-billion-dollar buildout cycle. The demand for power, in particular, creates an interesting secondary market for energy producers and grid infrastructure companies. The key metric to watch isn’t whether Llama or Grok catches up to GPT.

It’s whether the sheer scale of infrastructure spending generates returns that justify the cost. The next 18 months, as Prometheus comes online and Colossus scales toward its million-GPU target, will offer the first real data points on that question. Disclosure: This article was edited by Editorial Team.

For more information on how we create and review content, see our Editorial Policy. AI Mark Zuckerberg and Elon Musk bet big on data centers as AI models lag behind Both tech billionaires are pouring hundreds of billions into physical infrastructure while their AI models struggle to keep pace with competitors by Editorial Team Jul. 7, 2026 Share Add us on Google Meta and xAI are channeling staggering sums into data center infrastructure, betting that raw computational power will eventually close the gap their AI models haven’t managed to bridge on their own.

The infrastructure arms race Meta’s marquee project is Hyperion, a facility in Louisiana that, when completed, will consume more electricity than the entire city of New Orleans. The physical footprint is equally absurd. Hyperion is expected to cover an area comparable to lower Manhattan.

Zuckerberg has said Meta plans to invest “hundreds of billions” into AI data centers. Then there’s Prometheus, Meta’s Ohio-based facility projected to go online in 2026. Advertisement xAI’s Colossus supercomputer was constructed in just 122 days inside a repurposed factory.

The plan is to scale Colossus to accommodate one million GPUs, backed by strategic acquisitions in the energy sector. Where the models fall short For all the infrastructure muscle, the AI models themselves tell a less triumphant story. Meta’s Llama series and xAI’s Grok have been released in successive iterations, but neither has managed to consistently outperform competitors like OpenAI’s GPT series or Google’s Gemini in ways that matter to developers and enterprise customers.

The challenges aren’t necessarily about the models being broken. They’re about cost, efficiency, and the increasingly crowded field of foundation model providers. Community pushback and energy concerns Both Meta and xAI face growing local opposition in the communities where these data centers are being constructed.

Residents are raising concerns about energy consumption, emissions, and the strain on local power grids. What this means for investors Companies providing energy solutions, construction services, and specialized data center technology stand to benefit from what could be a multi-year, multi-hundred-billion-dollar buildout cycle. The demand for power, in particular, creates an interesting