Meta Secures $100B+ AI Chip Deal with AMD Following Nvidia Partnership
- Meta Platforms is significantly expanding its artificial intelligence infrastructure with a deal to purchase billions of dollars worth of chips from Advanced Micro Devices, potentially exceeding $100 billion.
- The deal encompasses up to 6 gigawatts of hardware from AMD, with the first gigawatt of shipments slated to begin in the second half of this year.
- Beyond the chip purchase, Meta has been granted a stock warrant allowing it to acquire up to 160 million shares of AMD common stock at a price of...
Meta Platforms is significantly expanding its artificial intelligence infrastructure with a deal to purchase billions of dollars worth of chips from Advanced Micro Devices, potentially exceeding $100 billion. The agreement, announced on , comes just days after Meta unveiled a long-term partnership with Nvidia for AI data centers, signaling a broad strategy to diversify its AI supply chain.
The deal encompasses up to 6 gigawatts of hardware from AMD, with the first gigawatt of shipments slated to begin in the second half of . Meta will acquire AMD’s MI450 series of GPUs and the latest generation of CPUs, including those codenamed Venice and Verano. The MI450 accelerator is built on Taiwan Semiconductor Manufacturing Co.’s two-nanometer process and features 432 gigabytes of high-speed HBM4 memory capable of moving 19.6 terabits of data per second.
Beyond the chip purchase, Meta has been granted a stock warrant allowing it to acquire up to 160 million shares of AMD common stock at a price of $0.01 per share. However, the vesting of these shares is contingent upon the successful completion of performance milestones related to the collaboration. The full stock award is conditional on AMD’s share price reaching $600.
AMD’s stock price jumped 8% on following the announcement, reflecting investor confidence in the deal’s potential. The agreement represents a significant win for AMD as it seeks to gain ground in the rapidly expanding AI chip market, currently dominated by Nvidia. AMD CEO Lisa Su highlighted the strong demand for CPUs, stating that the market is “absolutely on fire” due to AI infrastructure deployments and the scaling of AI inference.
The move by Meta underscores the escalating investment in AI infrastructure by major technology companies. Meta CEO Mark Zuckerberg described the partnership with AMD as “an important step” in diversifying compute resources and progressing towards “personal superintelligence,” which he defines as AI systems designed to deeply understand and empower individuals.
Meta has committed to investing at least $600 billion in U.S. Data centers and AI infrastructure over the coming years, with a projected capital expenditure of $135 billion in alone. The company is also developing a new rack design, called Helios, in collaboration with AMD, featuring liquid cooling and a double-wide layout to improve maintenance efficiency. Each Helios rack is designed to accommodate up to 72 MI450 accelerators.
The agreement with AMD mirrors a similar deal struck last October between AMD and OpenAI, where equity was exchanged for a commitment to purchase chips. This trend suggests a growing willingness among AI developers to diversify their chip suppliers and potentially secure long-term supply agreements through equity stakes.
While the appetite for AI chips remains substantial, concerns exist regarding the substantial financial outlays made by companies like Meta and their ability to recoup these investments through increased profits and productivity. The deal highlights the intense competition for AI processing power and the strategic importance of securing access to advanced chip technology.
The scale of the agreement – potentially exceeding $100 billion – positions AMD as a key player in the AI revolution, challenging Nvidia’s established dominance. The 6-gigawatt commitment from Meta represents a significant portion of AMD’s potential revenue stream and could substantially impact the company’s financial performance in the coming years. A gigawatt of power usage is equivalent to that of several hundred thousand homes.
The increasing reliance on CPUs for AI inference, as noted by AMD’s CEO, is a notable trend. CPUs are seen as more efficient and scalable than GPUs for certain AI tasks, offering an alternative to complete dependence on Nvidia’s graphics processing units.