Meta Platforms Expands AI Infrastructure and Develops Its Own Hardware
Meta Platforms 擴展 AI 基礎設施並開發自有硬體
Introduction
Meta Platforms is using a two-part strategy to improve its artificial intelligence capabilities by creating its own computer chips and expanding its global network of data centers.
Meta Platforms 正採取兩階段策略，透過開發自有電腦晶片並擴展全球資料中心網路，以提升其人工智慧能力。
Main Body
The company is working toward hardware independence through the Meta Training and Inference Accelerator (MTIA) program. A key part of this plan is the 'Iris' chip, which will begin production in September after a six-week testing period. Developed with Broadcom and manufactured by TSMC, this chip aims to reduce Meta's reliance on external suppliers like Nvidia and AMD, which will help lower the rising costs of hardware. Furthermore, to avoid supply chain problems, Meta has signed long-term contracts with Samsung Electronics, Sandisk, and Sumitomo Electric.
該公司正透過 Meta 訓練與推論加速器 (MTIA) 計畫，致力實現硬體獨立。此計畫的關鍵是 "Iris" 晶片，在經過六週的測試期後，將於九月開始生產。該晶片由 Broadcom 共同開發並由 TSMC 製造，旨在減少 Meta 對 Nvidia 和 AMD 等外部供應商的依賴，從而有助於降低不斷上升的硬體成本。此外，為了避免供應鏈問題，Meta 已與 Samsung Electronics、Sandisk 和 Sumitomo Electric 簽署長期合約。
At the same time, Meta is significantly increasing its computing power. The company plans to launch seven gigawatts of infrastructure this year and expects this to grow to 14 gigawatts by 2027. To support this growth, Meta has set a budget of up to $145 billion for the current year. A major part of this expansion is the opening of its first Canadian facility in Alberta. This site represents an investment of about CAD $13 billion and uses a special liquid cooling system to save water. Meta emphasized that all electricity used at this site will be balanced by renewable energy sources.
與此同時，Meta 正大幅增加其運算能力。該公司計劃今年推出 7 吉瓦 (gigawatts) 的基礎設施，並預計到 2027 年將增至 14 吉瓦。為了支持這一增長，Meta 為本年度設定了高達 1,450 億美元的預算。此次擴張的一個重點是在加拿大亞伯達省 (Alberta) 開設其首個設施。該場地的投資額約 130 億加幣，並使用一種特殊的液冷系統以節省用水。Meta 強調，該場地使用的所有電力將由可再生能源抵銷。
Conclusion
In summary, Meta is combining its own custom AI chips with a larger global infrastructure to meet its growing computing needs.
總結來說，Meta 將其自有定制的 AI 晶片與更大規模的全球基礎設施結合，以滿足日益增長的運算需求。
Vocabulary Learning
🚀 The 'B2 Shift': Moving from Simple to Complex Connections
An A2 student says: "Meta makes chips. Meta builds data centers. This helps them save money."
To reach B2, you must stop using short, choppy sentences. You need to use Connectors of Purpose and Result. Look at how the text transforms simple ideas into professional English:
⚡ The Power of "To" and "In order to"
Instead of saying "Meta wants to save money, so they make chips," the text uses a more sophisticated structure:
"...this chip aims to reduce Meta's reliance on external suppliers... which will help lower the rising costs."
B2 Tip: Use [Verb] + to + [Infinitive] to explain why a company or person does something. It sounds more intentional and academic.
🛠️ The "Which" Bridge (Relative Clauses)
This is the most important tool for B2 fluency. Instead of starting a new sentence, use , which... to add extra information about the previous idea.
- A2 Level: Meta is making a chip. It is called Iris. It starts in September.
- B2 Level: Meta is making a chip called Iris, which will begin production in September.
📈 Vocabulary Upgrade: Precision over Simplicity
Stop using words like "big" or "get." Use these B2 Power Words found in the text:
| A2 Word | B2 Upgrade | Context from Text |
|---|---|---|
| Big / More | Significant / Expanding | "Significantly increasing its computing power" |
| Get rid of | Reduce reliance on | "Reduce Meta's reliance on external suppliers" |
| Start | Launch | "Plans to launch seven gigawatts" |
💡 Final Strategy: Next time you write, find two short sentences and try to glue them together using , which... or a purpose phrase like to [verb]. This is the fastest way to move your writing from 'Basic' to 'Upper-Intermediate'.