TLDR
- Nvidia and Meta announced an expanded multiyear partnership covering millions of Blackwell and Rubin GPUs, CPUs, and networking products.
- No dollar figure was disclosed, but HSBC estimates individual GPU racks at $3 million, with Rubin hardware priced ~40% higher.
- Meta will also deploy Nvidia’s Grace and Vera CPU-only servers, potentially displacing Intel and AMD in its data centers.
- The deal eases concerns about Meta shifting to Google’s TPUs, which had weighed on Nvidia’s stock in recent months.
- Arista Networks fell 3.2% after-hours after Meta announced it would use Nvidia’s Spectrum-X Ethernet switches instead.
Nvidia and Meta announced a major expansion of their partnership late Tuesday, with Meta committing to deploy millions of Nvidia’s Blackwell and Rubin GPUs, along with CPUs and networking products across its data centers.
BREAKING: Meta and Nvidia announce a multiyear partnership to expand AI infrastructure, including "large-scale deployment" of Nvidia CPUs and GPUs.
Nvidia says this partnership will deploy "millions of Nvidia GPUs" pic.twitter.com/f39VqDJ3wj
— The Kobeissi Letter (@KobeissiLetter) February 17, 2026
The deal covers both AI model training and inference — the process of running AI models once trained. Nvidia CEO Jensen Huang called out Meta’s scale directly, saying no company deploys AI at the same level.
“Through deep codesign across CPUs, GPUs, networking, and software, we are bringing the full Nvidia platform to Meta’s researchers and engineers,” Huang said in a statement.
Meta CEO Mark Zuckerberg added that the expanded partnership would help build “leading-edge clusters using their Vera Rubin platform to deliver personal superintelligence to everyone in the world.”
No financial terms were disclosed. However, HSBC estimates a single GB300 NVL72 rack — containing 72 of Nvidia’s GPUs — costs around $3 million. Susquehanna analyst Christopher Rolland estimated this week that the upcoming Rubin hardware will carry an average selling price roughly 40% higher than current systems. Meta confirmed it will deploy GB300-based systems.
Why This Deal Matters for Nvidia
The partnership directly addresses one of the bigger overhangs on Nvidia’s stock. Back in November, The Information reported that Meta was in talks with Google to use its TPU chips for AI workloads. Google’s TPUs are designed in collaboration with Broadcom, and the competitive threat had been cited as a reason Nvidia’s stock had stalled.
This deal essentially pushes that concern aside — at least for now.
Nvidia stock is still down more than 1% year to date, while AMD has dropped over 5% in the same period. AI-related stocks broadly have struggled in 2026, with Microsoft down more than 17% since January 1 and Meta off about 3.3%.
Meta is also rolling out Nvidia’s first large-scale Grace CPU-only servers, with plans to add Vera CPU-only systems in 2027. Those servers would handle tasks currently done by Intel and AMD chips — putting both companies in an uncomfortable spot.
Arista Networks Feels the Pinch
There was at least one loser in Tuesday’s announcement. Arista Networks fell 3.2% in after-hours trading after Meta said it would integrate Nvidia’s Spectrum-X Ethernet switches into its infrastructure. Arista counts Meta as one of its largest clients, so the shift toward Nvidia’s own networking products is a direct hit.
Meta will also use Nvidia’s Confidential Computing feature — which enables private data processing via GPUs — inside its WhatsApp messaging app.
Nvidia’s Cloud Partner program, which includes CoreWeave and Crusoe, will also play a role, with Meta able to access additional chip capacity through those hosted platforms.
Nvidia closed Tuesday at $184.97, up 1.20%, before rising another 1.51% in overnight trading to $187.77.
Get 3 Free Stock Ebooks
Discover top-performing stocks in AI, Crypto, and Technology with expert analysis.
- Top 10 AI Stocks - Leading AI companies
- Top 10 Crypto Stocks - Blockchain leaders
- Top 10 Tech Stocks - Tech giants
