Striving toward the development of superintelligence, Meta has expanded its partnership
with chip manufacturer Nvidia -- the world's most valuable company -- announcing plans to power its data centers with new stand-alone central processing units (CPUs) and next-generation graphics
processing units (GPUs) as well as privacy-centric data processing on WhatsApp.
The multiyear partnership follows Meta's recent pledge to “build tens of gigawatts this decade,” with the technology giant
committing to spending over $600 billion in the U.S. alone throughout 2028.
advertisement
advertisement
In 2026, Meta plans to spend about $135 billion on artificial intelligence (AI).
The majority of Meta's AI
investment is expected to focus on building data centers intended to power the company's pursuit of “personal superintelligence” -- AI that exceeds human cognitive abilities.
Overall, Meta's plans include the development of 30 data centers, which will require the large-scale deployment of graphics processing units. Prometheus and Hyperion -- Meta's two largest centers
-- are currently under construction in Ohio and Louisiana, requiring a total of 6 gigawatts of power.
According to Nvidia's announcement, its partnership with Meta is designed to “enable
the large-scale deployment of Nvidia CPUs and millions of Nvidia Blackwell and Rubin GPUs, as well as the integration of Nvidia Spectrum-X Ethernet switches for Meta's Facebook Open Switching System
platform.”
As Nvidia's Blackwell GPUs become more difficult to acquire, Meta's deal with the chip manufacturer guarantees that the company will receive an adequate supply for its data center
needs.
Per the deal, Meta will become the first company to utilize large-scale Grace CPU-only servers, which are used to run AI models instead of train AI models.
The company will also
use Nvidia's Confidential Computing, a system that facilitates private data processing through its GPUs in the WhatsApp platform.
Both companies are working together to expand Confidential
Compute capabilities beyond WhatsApp to other Meta-run social platforms in attempts to support “privacy-enhanced AI at scale,” per the announcement.
Meta will also use Nvidia's
Spectrum-X Ethernet switches to link GPUs together at its massive data centers.
In addition to its partnership with Nvidia, Meta is in a multiyear deal with British semiconductor company Arm, a direct Nvidia
competitor that is expected to help Meta scale its AI systems to its 3 billion social media users across the globe.
As the AI race intensifies, more technology giants are building their own
in-house chips.
To compete with Google, Amazon, and Microsoft, Meta is in the process of developing its own in-house AI chip called “Artemis,” which is expected to launch sometime
this year.