Meta to Procure 'Millions' of Nvidia Processors in Multi-Year Deal—Comprehensive GPU, CPU, and Networking Order Highlights AI Investment

Meta and Nvidia announced on February 17, 2026, that they have signed a multi-year agreement to procure 'millions' of Nvidia processors.This contract includes Blackwell-generation and future GPUs, Grace CPUs, and networking equipment. Meta is already Nvidia's second-largest customer, and this deal further expands that relationship.

1. Scale of the deal—Millions of units, several years, and tens of billions of dollars

1-1. The meaning of the expression 'millions of units'

According to Bloomberg Tech, the deal was announced using the term 'millions,' but the specific duration and monetary value were not disclosed.

However, since the average selling price (ASP) of Nvidia's data center GPUs is approximately $16,000, and reaches double that for top-tier models, it is highly likely that the GPU portion alone will reach tens of billions of dollars. With the addition of CPUs and networking equipment, Bloomberg speculates that the total amount 'could rival or exceed the amount Meta pays Nvidia annually.'

1-2. From Blackwell to Grace CPU—A comprehensive agreement

The contract includes the following:

• Blackwell GPUs (current generation) and subsequent generations

• Grace CPUs

• Networking equipment

Of particular note is that this is the first instance where Grace CPUs are being sold on a large scale as standalone products.Previously, Grace CPUs were integrated as part of systems like Grace Hopper or Grace Blackwell. This time, Meta is procuring Grace CPUs individually to use for general-purpose data center processing.

2. Standalone Grace CPU sales—Intrusion into Intel's territory

2-1. Entry into the 'general-purpose processor market'

A Bloomberg Tech reporter evaluated this deal as 'a step toward market expansion for Nvidia.'

'This is Intel's territory. It is the general-purpose processor market that handles many general-purpose computing tasks in data centers. Nvidia had been incorporating Grace CPUs into the core of its systems, but this time they are saying they can be used standalone. Meta is a good use case.'

Nvidia dominates the market with GPUs specialized for AI workloads, but Intel and AMD CPUs remain the mainstream for general-purpose processing across the entire data center (web servers, databases, storage management, etc.). The standalone sale of Grace CPUs signifies a full-scale entry into this domain.

2-2. Impact on AMD

Meanwhile, AMD stock fell, as the structure of Nvidia continuing to monopolize hyperscaler spending became clear.

Meta is not a hyperscaler (like AWS, Azure, or Google Cloud), but its data center footprint (scale) is on par with one. Bloomberg reported that Meta accounts for 9% of Nvidia's sales, with Microsoft being the top customer.

3. Relationship with in-house CPU development—The meaning of the commitment to Nvidia

3-1. Microsoft, Meta, and Amazon are developing their own CPUs

What is interesting is the fact that major tech companies like Microsoft, Meta, and Amazon are developing their own custom CPUs.

Nevertheless, Meta's large-scale commitment to Nvidia and AMD CPUs serves as a positive signal for investors. It suggests that their own CPUs cannot meet demand alone, or that they are adopting a strategy of using off-the-shelf products from Nvidia and AMD for general-purpose processing.

4. Morgan Stanley: 'AI Investment is a $10 Trillion Cycle'

4-1. Productivity gains as proof of ROI

Katy Huberty, Global Research Director at Morgan Stanley, stated on the same day that 'AI investment will be a $10 trillion CapEx cycle.'

'This investment cycle is based on the premise that productivity will accelerate as technology permeates the market. Over the past 20 years, labor productivity has been running at half its normal level. AI-adopting companies are experiencing margin expansion, with growth rates double those of the MSCI World Index and the S&P 500.'

She pointed out that the impact of AI investment is spreading to non-tech sectors—consumer goods, apparel, durable goods, and automotive.

'Leadership is shifting not just to AI enablers, but to AI adopters. This is a structural change in the market.'

4-2. 'Indiscriminate selling' of software stocks

Meanwhile, Huberty noted that software and service stocks have been 'sold off indiscriminately.' However, she believes there are opportunities for companies with a 'data moat'—such as credit information, market data, financial records, and customer data.

'Companies that possess data can connect large language models to their applications and unlock true value.'

5. Autodesk CEO: 'AI investment is already yielding ROI'

5-1. $200 million strategic investment in World Labs

Autodesk announced a $200 million strategic investment in AI research firm 'World Labs' on the same day.World Labs later announced a $1 billion funding round, with Autodesk accounting for the majority of it.

Autodesk CEO Andrew Anagnost emphasized that AI is already generating ROI.

'Software companies have been investing in AI for decades. Lighting simulation, wind analysis, generative models considering structural constraints, and decision support on construction sites—AI is already delivering value. Over the next five years, we will continue to invest and provide incremental value.'

5-2. AI solving the 'capacity problem'

Anagnost cited the 'capacity problem' as a fundamental issue facing the construction and manufacturing industries.

'There aren't enough materials to build and rebuild everything in the world. People are building data centers, and we also need infrastructure to support daily life. We need to execute more projects faster with the same number of people and the same capacity. This is a fundamental capacity problem unique to the design and build industry. AI can solve this.'

Conclusion: AI Infrastructure Investment is Accelerating

Meta's procurement of millions of units, Morgan Stanley's $10 trillion cycle forecast, and Autodesk's strategic investments—all of these indicate that AI infrastructure investment is in full swing.

In particular, Nvidia's standalone sales of Grace CPUs signify a full-scale entry into the general-purpose data center market, suggesting that competition with Intel and AMD has entered a new phase. Meanwhile, software companies are facing 'indiscriminate selling,' but opportunities remain for firms with data moats.

The ROI of AI investment is already becoming apparent in some areas and is expected to expand gradually over the next five years.

Recommended Articles

Next Big Wave (Growth Stocks, Seeds of Ideas, and Trend Deep Dives)