Meta develops internal chips-what the stock price means

A new and exciting report on wonderful seven shares, definition platforms Nasdak: DefinitionRecently. Reuters Thompson NYSE: TI I mentioned that dead Her semiconductor test to train artificial intelligence models.

This is a direct attempt to reduce dependence on the GPU Nasdak: nvda.

So, what are the internal Meta chips, and why the company takes this step?

How can this company help reduce the costs of artificial intelligence, and can this development benefit the shares in the long run?

Meta chips at home: What are and why Meta returns to Nvidia

Meta is now testing the first home -made chip specifically to train artificial intelligence. Artificial intelligence training indicates one time and directly to teach the artificial intelligence model how to think and make predictions. It is said that Meta is already using the company’s manufacturers to infer. Inference indicates the process of artificial intelligence model in fact a unique question or circumstance. This is what happens when someone Chatgpt asks a question, for example. Companies must train models before they can perform inference.

Both tasks require large -scale energy and calculation costs, but they occur at different times. Training requires computing and energy costs. Once the company trains a model, the costs of each individual conclusion are low. However, due to its inference, these small costs add up, It is possible to create a greater cost over time.

In general, Meta is looking at both ends of the spectrum, trying to reduce costs on both training and inference. Reducing costs in each of the key. Meta wants to create better models over time through training. They also want billions of users to interact with their models through inference, which makes the pieces cost their key as well.

The tendency to do this by creating their internal chips is logical, taking into account the tremendous cost of NVIDIA graphics processing. The total NVIDIA margin was approximately 74 % in the last quarter, indicating the power of the enormous prices. NVIDIA chips are mainly used for training and inference. Building chips at home implicitly insert more competition for NVIDIA. This means that companies like Meta do not have to accept and pay any NVIDIA price simply.

Energy plays a huge role in the internal Meta progress

It can offer the prosecutor’s AI chips like Meta, which is also used as both Faster performance and low energy consumption Compared with graphics processing units. This is true for specialized inference tasks, although graphics processing units still lead to Genai’s general performance. Amazon Nasdaq: amzn It says its customs, Tradium2, 30 % to 40 % the best price for performance From H100 NVIDIA graphics processing units.

Massachusetts Institute of Technology Sloan Management Energy is expected to be used from data centers to height. Data centers currently represent 1 % to 2 % of global energy demand. The Massachusetts Institute of Technology says this number may rise to 21 % by 2030, given the costs of artificial intelligence. Thus, reducing energy use in its chips is very important for dead. It can reduce the costs that it has to pay the costs of energy agreements to operate the huge data centers.

Meta is at different stages when it comes to using its household chips, depending on the task. Meta is currently using an inference chip known as Artemis, widely to make recommendations. This includes providing short -in -form ads and videos such as Reels and the other content recommendation on Facebook and Instagram. However, Meta has not yet adopted an artemis for Genai.

According to Reuters, the company has just started testing its training segment. If things go well, the company hopes to start using the chip for training by 2026. The company will first study how it can use the chip to train the recommendation systems, then how you will do so for Gen AI.

The internal Meta chips can help reduce the cost, which may benefit from the shares in the long run

The Meta batch can be to build its internal chips as important wind winds for the arrow. The probability of low costs can be added to the company’s margins in the long run. However, the stakeholders will have to wait and know if their training segment tests will succeed. In addition, the increase in the production of these chips will have significant costs related to them, and possibly put the clouds in the short term on the margins.

Additional developments related to the test phase in the future training and implementation segment of both Raska will be a key to watching.

Before you think about the identification platforms, you will want to hear it.

Marketbeat follows the best research analyst at Wall Street, the best performance in Wall Street and the stocks they recommend to their customers on a daily basis. Marketbeat has identified the five shares whom the top analysts quietly whispered to their customers to buy now before wiping the broader market … The definition platforms were not on the list.

While Meta platforms currently have a moderate purchase classification between analysts, higher -rated analysts believe that these five stocks buy better.

Show the five stocks here

Are you looking for the following FAANG stock before everyone hears it? Enter your email address to find out any stock of analysts who believe that the next Trillion Dollaar Tech Companyy.

Get this free report

Such an article? Share it with a colleague.

Copy to the portfolio.