Meta Moves to Develop and Deploy In-House AI Chips, Unlocking Future Valuation Potential
Meta has announced the launch of four custom-designed, internally developed chips tailored for artificial intelligence, forming part of the company’s broader data center expansion strategy.

This week, Meta unveiled four custom-built chips specifically designed for AI tasks as part of its data center expansion plan. These chips belong to Meta’s Training and Inference Accelerator (MTIA) series, which was first publicly disclosed in 2023, with a second-generation version introduced in 2024.

Meta stated that it is working to develop and deploy four new generations of MTIA chips over the next two years to support ranking and recommendation systems, as well as generative AI workloads. This pace is significantly faster than the typical semiconductor development cycle. By designing its own chips, the social media giant aims to achieve greater cost efficiency across its data center fleet, rather than relying solely on external suppliers. This approach provides Meta with greater silicon supply diversification, reduces exposure to price fluctuations, and enhances its bargaining power.

The MTIA 300 was deployed several weeks ago to help train smaller AI models that power Meta’s core ranking and recommendation functions. These functions include delivering relevant content and online advertisements to users across platforms such as Facebook and Instagram.

The upcoming chips are designed to handle more advanced generative AI inference tasks, such as creating images and videos based on user text prompts. Meta has completed testing of the MTIA 400 and is currently deploying it in data centers, while two additional chips are expected to enter operation in 2027.

For any semiconductor company or engineering team, releasing a new chip every six months is highly unusual. This accelerated cadence reflects rapid capacity expansion and increased capital expenditure. The chips are expected to have a standard lifespan of more than five years.

Market Interpretation:

Meta’s AI-related spending includes the construction of a data center in Louisiana, along with two additional facilities in Ohio and Indiana. After OpenAI and Oracle Financial Software canceled plans to expand the Stargate AI data center project in Texas, Meta reportedly moved to lease space at that site.


Abel Gao brings over 11 years of experience as a financial analyst to TMGM, with expertise in advanced chart analysis and statistical modeling of global markets. As a Trading Strategy Team Mentor, he combines traditional charting techniques with modern analytical methods to provide insights that support traders in developing systematic strategies. In addition to analysis, Abel mentors both beginner and experienced traders, and his reports and commentary are widely used as educational resources within TMGM’s trading community.
Lire la suite

COTATIONS EN DIRECT

Nom / Symbole
Graphique
% Variation / Prix
EURUSD
Variation 1 jour
+0%
0
XAUUSD
Variation 1 jour
+0%
0
BTCUSD
Variation 1 jour
+0%
0

TOUT SUR FOREX

Explorer Plus d'Outils
Académie de Trading
Parcourez une large gamme d'articles éducatifs couvrant les stratégies de trading, les perspectives de marché et les fondamentaux financiers, le tout en un seul endroit.
En Savoir Plus
Cours
Explorez des cours de trading structurés conçus pour soutenir votre croissance à chaque étape de votre parcours de trading.
En Savoir Plus
Webinaire
Rejoignez des webinaires en direct et à la demande pour obtenir des perspectives de marché en temps réel et des stratégies de trading d'experts de l'industrie.
En Savoir Plus