In a determined push to accelerate its massive AI infrastructure and close the gap with rivals, Meta has begun erecting temporary, tent-like facilities to house parts of its new data centers.
CEO Mark Zuckerberg announced the company is moving forward with several enormous data center campuses.
One such project, codenamed "Hyperion," could eventually reach a staggering 5-gigawatt power capacity.
The accelerated timeline follows the underwhelming performance of the company's "Llama" AI model compared to offerings from competitors like OpenAI, Google, and xAI. Zuckerberg appears unwilling to adhere to traditional construction schedules any longer.
Following press reports, which Meta later confirmed, the company is using lightweight, tent-like structures to add immediate computing capacity at its sites while permanent facilities are still under construction.
The semiconductor research firm SemiAnalysis noted that this design prioritizes function over form. Its primary goal is not aesthetics or redundancy but simply "to get compute running as quickly as possible."
The strategy, seemingly inspired by Elon Musk's previous efforts to ramp up Tesla production, relies on prefabricated power and cooling units. It also forgoes some traditional safety standards, such as backup diesel generators, all in the name of speed.
Still, the approach comes with its own set of challenges.
Tents are known to heat up significantly, which could pose an obstacle to cooling the expensive equipment inside. The company might even need to halt some computing processes during peak summer heat.
Meanwhile, Zuckerberg has revealed far greater ambitions on the horizon.
In addition to Hyperion, the company is developing another project codenamed "Prometheus," a 1-gigawatt campus scheduled to go online in 2026.
Zuckerberg also mentioned that one of these massive complexes could eventually cover an area comparable to a significant portion of Manhattan.
These moves signal a shift in the tech landscape. The competition is no longer confined to algorithms and software; it has evolved into a frantic race for physical infrastructure. Data centers have now become the primary battleground in the next chapter of artificial intelligence.
