Meta's Hundred-Billion Bet on AMD: Is the AI Chip Landscape Reassessing "Investment Certainty"?

On the evening of February 24th, the global AI community received a major news story that could reshape the industry landscape: Meta officially partnered with AMD, reaching a multi-year strategic cooperation agreement. According to the agreement, Meta will deploy up to 6 gigawatts of AMD AI chips in the coming years for data center expansion. Once the news broke, AMD’s stock price surged pre-market, soaring over 10% at one point during trading, and closing up 9.99%, with a market cap increase of nearly $20 billion in a single day.

If we analyze the logic behind AMD’s stock price surge, it can be summarized with three core keywords: revenue certainty, full-stack customization for differentiation, and rare deep capital alliance binding. Behind these three points, there is a deeper concept repeatedly examined by the capital market—“investment certainty.” Goldman Sachs pointed out in its latest research report that the core logic of this deal is Meta exchanging potential equity for AMD’s deeply customized services, while AMD, in turn, gains future equity dilution in exchange for guaranteed orders from top cloud providers. This “certainty” is currently one of the most scarce assets in the AI investment wave.

A Sudden “Timely Rain” for Revenue

To understand the significance of this deal for AMD, we need to look at AMD’s recent financial performance.

Less than a month ago, on February 4th, AMD released its Q4 2025 and full-year financial reports. Data showed that in Q4 2025, AMD achieved a record revenue of $10.3 billion, up 34% year-over-year. The full-year revenue reached $34.6 billion, also a record high. Notably, the data center business revenue in Q4 hit $5.4 billion, up 39% year-over-year, setting a new record.

Interestingly, despite this seemingly impressive report, AMD’s stock price fell more than 8% after hours, and even dropped over 17% at one point. Why? Because the market was concerned that the growth momentum of its AI chips might not meet expectations. The company’s guidance for Q1 2026 revenue was $9.8 billion (±$300 million), a 5% quarter-over-quarter decline, and revenue from China’s MI308 chips was expected to drop from $390 million in Q4 to $100 million. Some investors interpreted this guidance as a sign of slowing growth.

At this critical moment, Meta’s order arrived.

Industry analysts estimate that this multi-year agreement is valued between $60 billion and $100 billion. AMD CEO Dr. Lisa Su stated at the release conference that each gigawatt of computing capacity could generate hundreds of millions of dollars in revenue. If executed evenly over five years, this would bring AMD an incremental annual revenue of over $12 billion, directly adding more than 30% certainty to its 2025 total revenue base.

AMD CFO David Hu explicitly said, “We expect this cooperation to drive substantial revenue growth over many years and increase our non-GAAP EPS.” This gave the market confidence that AMD can deliver results in the AI arena—something that was most lacking after the previous earnings report.

From “Selling Chips” to “Building Ecosystems”: Full-Stack Customization for a Differentiation Moat

This cooperation is far from a simple chip purchase; it involves deep strategic integration of hardware, software, and systems.

According to the agreement, AMD will provide Meta with a customized version of the new-generation Instinct MI450 GPU, optimized specifically for Meta’s AI workloads. Goldman Sachs noted that AMD is designing a semi-custom MI450 GPU tailored for Meta’s workloads. The first batch, totaling 1 gigawatt, is expected to start shipping in the second half of 2026.

But that’s not all. The partnership also includes the sixth-generation EPYC CPU codenamed “Venice,” and a next-generation EPYC processor “Verano” optimized for AI workloads. Previously disclosed information indicates that the “Venice” CPU will be based on 2nm process technology, with performance and efficiency improvements of over 70%, and thread density increased by over 30%. The two companies will also develop a comprehensive solution based on the jointly developed Helios rack-scale architecture.

Lisa Su revealed during an analyst call that this customization begins with a “workload-first” approach rather than chip-first, with optimization spanning “chip-level, board-level, and system-level.” This full-stack customization capability from chip to rack precisely meets the needs of tech giants like Meta, which pursue maximum efficiency.

Mark Zuckerberg stated in a post-announcement statement, “This is an important step for Meta as we diversify our computing infrastructure. I expect AMD to be a key partner for many years to come.” Meta also explicitly said in the press release, “By diversifying our partnerships and technology stack, we are building a more resilient and flexible infrastructure.”

Research firm Creative Strategies analyst Ben Bajarin pointed out that a key aspect of AMD’s agreement, and what differentiates it from Meta’s deals with NVIDIA, is that this is the first deployment of customized GPUs. This deep customization partnership elevates AMD from a mere chip supplier to a co-builder of Meta’s AI infrastructure.

Equity Binding: Elevating Supply Chain Cooperation to a Capital Alliance

The third and most rare core logic, which has attracted significant market attention, is equity binding.

According to the agreement, AMD grants Meta performance-based warrants, allowing Meta to purchase up to 160 million AMD common shares, approximately 10% of the company. These warrants are linked to GPU shipment volumes and AMD’s stock price reaching certain thresholds—initial warrants vest upon the first gigawatt shipment, subsequent tranches vest as Meta’s procurement scales up to 6 gigawatts, and the final tranche vests if AMD’s stock price hits $600 per share.

This clause directly elevates supply chain cooperation to a capital alliance level. For Meta, if AMD’s future performance explodes and its stock price soars, Meta will share in the gains as a shareholder. For AMD, this performance-based cooperation tightly binds its long-term execution and value creation with Meta.

Market insiders call this innovative financing method “circular financing”—one company pays another, which then purchases products or services from the first. As AI investment heats up, the amount involved in such deals has surged. In October last year, AMD’s agreement with OpenAI was nearly identical in terms of structure to this Meta deal.

“Investment Certainty” Becomes a New Market Consensus

What is “investment certainty”? It’s the confidence that your invested money will be recouped with a reasonable chance of profit. AMD’s stock price plummeted after its earnings report because the market couldn’t see a clear growth path for its AI business—yet, Meta’s order provides a reassuring signal.

Goldman Sachs pointed out that the core logic of this deal is: Meta exchanges potential equity for AMD’s deeply customized services, while AMD, through future equity dilution, secures guaranteed orders from top clients. This “equity-for-order” model deeply binds the interests of both parties. Meta is not only a customer but also a quasi-shareholder—motivated to prioritize AMD’s products.

Based on this, Goldman Sachs raised AMD’s EPS forecasts for 2027 and 2028 by about 16-17%, and increased its target price from $210 to $240. However, they maintained a “Neutral” rating.

Because equity binding is only “paper wealth,” real certainty depends on product deployment. Goldman Sachs emphasized that they will become more optimistic once they see actual deployment schedules from Meta and OpenAI. This reveals the core contradiction in the capital market: the gap between expected certainty (signing contracts) and actual certainty (real shipments) that takes time to bridge.

From a supply chain perspective, KGI Securities noted that Tongfu Microelectronics (002156), a core AMD packaging and testing manufacturer, is expected to benefit from increased business volume with major clients, raising its 2027 profit forecast by 5% and maintaining a “Buy” rating.

In summary, higher certainty means the market is willing to assign a higher valuation premium. AMD’s potential future dilution of 10% of its shares for this certainty already answered whether it’s worth it—its 10% stock price increase confirms it.

Reshaping the Landscape: New Variables in the Era of One Dominance, Many Contenders

Meta’s deep cooperation with AMD also has an important background: Meta is building a diversified supply chain system.

Just a week earlier, on February 17th, Meta reached a “multi-generation” cooperation agreement with NVIDIA, committing to use millions of current and next-generation NVIDIA chips in its AI infrastructure. NVIDIA revealed that Meta will be the first platform to deploy its large-scale Grace CPU servers. This indicates Meta is adopting a “three-pronged” strategy: developing its own AI chips, purchasing NVIDIA, and increasing AMD investments to support different types of computing loads.

Meta’s demand for computing power stems from its aggressive AI expansion plans. Its Q4 financial report projected that capital expenditures related to AI this year will reach $115-135 billion—almost double last year’s amount. Mark Zuckerberg stated during the earnings call that the company will continue to heavily invest in infrastructure to train leading models and provide personal superintelligence for billions of people and enterprises worldwide. Meta CFO Susan Li further revealed that the company still faces “capacity constraints.”

It’s worth noting that many tech giants are investing heavily in AI this year. Meta, Amazon, Google, and Microsoft plan to invest a combined approximately $650 billion, sparking a new round of AI infrastructure construction.

For AMD, this is not only a multi-billion dollar revenue guarantee but also a comprehensive market position affirmation—proving its ability to compete with giants on the highest-end AI battlefield, demonstrating the feasibility of its full-stack customization approach, and showing that it can lock in key large clients through innovative capital cooperation.

As this order begins to be fulfilled in the second half of 2026, AMD’s AI business will officially move from “promising prospects” to a “performance ramp-up” stage. Lisa Su emphasized during the analyst call that the second half of 2026 will be a critical turning point, with the Instinct MI450 series AI accelerators scheduled for mass production and shipment, starting revenue growth from Q3. The next-generation MI500 series, based on CDNA6 architecture and 2nm process, is expected to launch in 2027.

Conclusion: Certainty Is the Highest Premium

In the high-stakes game of AI chips, the market’s greatest fear is not competition but uncertainty. AMD’s previous stock price plunge was due to the market’s inability to see a clear growth path for its AI business.

Goldman Sachs’ statement that “they will become more active only after seeing actual deployment” reveals the core contradiction in today’s market: the gap between expected certainty (signing contracts) and actual certainty (real shipments), which takes time to bridge. But for investors, this very gap is the space where value can be created.

The global AI chip market’s “one dominance, many contenders” landscape is becoming more solid and dynamic because of this deal. NVIDIA remains the undisputed leader, but AMD has proven in its own way that it can find its place in this trillion-dollar race.

And perhaps, this is the fundamental reason why the capital market is willing to assign AMD an “investment certainty” premium.