Meta Platforms is in advanced discussions to acquire a stake of up to 10% in Advanced Micro Devices as part of a sweeping custom chip agreement that could fundamentally alter the competitive dynamics of the artificial intelligence semiconductor industry. The deal, if finalized, would represent one of the largest strategic investments by a technology company in a chipmaker and would signal Meta’s determination to reduce its overwhelming dependence on Nvidia for AI training and inference hardware.
The negotiations, first reported by Ars Technica, center on a multibillion-dollar arrangement in which Meta would commit to purchasing custom-designed AI accelerators from AMD over several years. In exchange, Meta would receive a significant equity position in AMD, potentially reaching the 10% threshold that would make Mark Zuckerberg’s company one of AMD’s largest shareholders. AMD’s current market capitalization hovers around $170 billion, meaning a 10% stake would be valued at approximately $17 billion.
A Strategic Pivot Away from Nvidia’s Dominance
The potential transaction underscores a growing anxiety among the world’s largest AI companies about their near-total reliance on Nvidia, which commands an estimated 80% or more of the market for AI training chips. Meta spent tens of billions of dollars on Nvidia’s H100 and subsequent Blackwell-generation GPUs to power its Llama family of large language models and the AI features embedded across Facebook, Instagram, and WhatsApp. That spending has given Nvidia extraordinary pricing power, and Meta’s leadership has grown increasingly uncomfortable with the concentration risk.
By taking a direct equity position in AMD, Meta would be doing more than simply signing a supply contract. The company would be making a financial bet that AMD can become a viable second source for high-performance AI silicon, and it would have a direct incentive to help AMD succeed in that effort. This kind of customer-investor hybrid relationship is unusual in the semiconductor industry, though it has precedents in other sectors where supply chain security is paramount.
The Custom Chip Dimension
Central to the discussions is the development of custom AI accelerators tailored specifically to Meta’s workloads. Rather than purchasing AMD’s off-the-shelf Instinct MI series GPUs, Meta would work with AMD’s engineering teams to design chips optimized for the particular mathematical operations and memory access patterns that dominate large language model training and inference. Custom silicon can deliver significant efficiency gains over general-purpose hardware because it eliminates transistors and features that a specific customer doesn’t need while amplifying the capabilities that matter most.
Meta already has experience with custom chip development. The company designed its own AI inference chip, known internally as MTIA (Meta Training and Inference Accelerator), which has been deployed in its data centers for certain workloads. However, MTIA has not proven capable of handling the most demanding training tasks, which still require the raw computational power of Nvidia’s top-tier GPUs. A custom AMD chip could occupy a middle ground—offering more tailored performance than AMD’s standard products while drawing on AMD’s deep expertise in high-performance computing architectures that Meta’s in-house team cannot easily replicate.
AMD’s Opportunity and Risk
For AMD, the deal would represent a transformative moment in its long campaign to challenge Nvidia in AI. Under CEO Lisa Su, AMD has invested heavily in its data center GPU lineup, most recently with the Instinct MI300X and MI325X accelerators. These products have won meaningful design wins at Microsoft, Oracle, and other cloud providers, but AMD’s AI revenue still represents a fraction of Nvidia’s. A locked-in, multiyear commitment from one of the world’s largest AI spenders would provide AMD with the revenue visibility and scale to justify further investment in AI chip development.
The equity component, however, introduces complexity. Issuing shares equivalent to 10% of the company would dilute existing AMD shareholders, a prospect that could face resistance from institutional investors. AMD would need to demonstrate that the long-term revenue and strategic benefits of the Meta partnership outweigh the dilutive impact. Analysts have noted that the structure of the equity transfer—whether through new share issuance, secondary market purchases, or some combination—remains unclear and could significantly affect the deal’s reception on Wall Street.
The Broader Industry Context
Meta’s move comes amid a broader trend of hyperscale cloud and AI companies seeking to diversify their chip supply chains. Alphabet’s Google has long designed its own Tensor Processing Units for AI workloads. Amazon Web Services developed its Trainium and Inferentia chips through its Annapurna Labs subsidiary. Microsoft has introduced its Maia AI accelerator. Each of these efforts reflects a shared conviction that relying on a single supplier for the most strategically important component in modern computing is an unacceptable business risk.
Yet Meta’s approach differs from these peers in a significant way. Rather than building its chip capabilities entirely in-house, Meta appears to be pursuing a partnership model in which it gains both custom hardware and a financial stake in its supplier’s success. This hybrid strategy could allow Meta to move faster than it could through internal development alone while still maintaining meaningful influence over the chip’s design and roadmap. If the arrangement works, it could become a template for other large AI consumers looking to secure their hardware futures.
Financial Implications for Both Companies
The financial scale of the potential deal is staggering by any measure. Meta’s capital expenditure plans for 2025 already called for between $60 billion and $65 billion, the vast majority directed toward AI infrastructure including data centers and the chips that fill them. Redirecting even a portion of that spending toward AMD custom silicon—while simultaneously acquiring a $17 billion equity stake—would represent a significant reallocation of capital that Meta’s board and shareholders would need to endorse.
For AMD, the revenue implications could be enormous. The company reported $7.7 billion in data center revenue for fiscal year 2024, a figure that was itself a record driven by surging AI chip demand. A multibillion-dollar annual commitment from Meta could double or even triple AMD’s AI-specific revenue within a few years, assuming the custom chips meet performance targets and Meta deploys them at scale. That revenue growth, combined with the validation of having one of the world’s most demanding AI customers, could help AMD close the valuation gap with Nvidia, whose market capitalization exceeds $3 trillion.
Regulatory and Governance Questions
A 10% stake in AMD would likely trigger regulatory review in multiple jurisdictions. In the United States, the Hart-Scott-Rodino Act requires notification to the Federal Trade Commission and the Department of Justice for acquisitions above certain thresholds, which this deal would easily exceed. Depending on the structure, it could also attract scrutiny from the Committee on Foreign Investment in the United States (CFIUS), though both Meta and AMD are American companies, which would simplify that analysis.
Governance questions also loom. A 10% shareholder typically expects board representation or at least significant influence over corporate strategy. Meta would presumably want input into AMD’s AI chip roadmap, manufacturing partnerships (AMD relies on TSMC for its most advanced chips), and research priorities. AMD’s existing shareholders and board would need to balance Meta’s interests against those of AMD’s other major customers, some of whom—like Microsoft and Google—are Meta’s direct competitors in AI. Managing those conflicts of interest would require careful structural safeguards.
What Comes Next for the AI Chip Race
Neither Meta nor AMD has publicly confirmed the negotiations, and there is no guarantee that the discussions will result in a completed transaction. Complex deals of this magnitude frequently evolve or collapse during the final stages of negotiation, particularly when equity stakes and custom engineering commitments are intertwined. Both companies’ stock prices have shown volatility in recent sessions as investors attempt to assess the probability and terms of an agreement.
If the deal does close, its effects would ripple across the semiconductor and AI industries for years. Nvidia would face its most serious competitive threat not from a rival chipmaker acting alone but from the combined resources and motivation of a chipmaker and its largest potential customer working in concert. Other hyperscalers might pursue similar arrangements, further fragmenting Nvidia’s grip on the AI accelerator market. And AMD, long positioned as the scrappy challenger in every market it enters, would find itself backed by one of the most powerful and well-capitalized technology companies on the planet—a partnership that could finally give it the scale to compete with Nvidia on something approaching equal terms.
The coming weeks will determine whether Meta and AMD can bridge the remaining gaps in their negotiations. But the mere existence of these talks signals a new phase in the AI hardware competition—one in which the lines between chip buyers and chip makers are becoming increasingly blurred, and in which the companies building the world’s most advanced AI systems are no longer content to simply purchase their computing power off the shelf.