Facebook owner Meta Platforms Inc is testing its first in-house chip for training artificial intelligence (AI) systems, a key milestone as it moves to design more of its own custom silicon and reduce reliance on external suppliers like Nvidia Corp, two sources told Reuters.
The world’s biggest social media company has begun a small deployment of the chip and plans to ramp up production for wide-scale use if the test goes well, the sources said.
The push to develop in-house chips is part of a long-term plan at Meta to bring down its mammoth infrastructure costs, as the company places expensive bets on AI tools to drive growth.
Photo: Reuters
Meta, which also owns Instagram and WhatsApp, has forecast total expenses of US$114 billion to US$119 billion for this year, including up to US$65 billion in capital expenditure largely driven by spending on AI infrastructure.
One of the sources said Meta’s new training chip is a dedicated accelerator, meaning it is designed to handle only AI-specific tasks. This can make it more power-efficient than the integrated graphics processing units (GPUs) generally used for AI workloads.
Meta is working with Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) to produce the chip, this person said.
The test deployment began after Meta finished its first “tape-out” of the chip, a significant marker of success in silicon development work that involves sending an initial design through a chip factory, the other source said.
A typical tape-out costs tens of millions of US dollars and takes roughly three to six months to complete, with no guarantee the test will succeed. A failure would require Meta to diagnose the problem and repeat the tape-out step.
Meta and TSMC declined to comment.
Meta executives have said they want to start using their own chips by next year for training, or the compute-intensive process of feeding the AI system reams of data to “teach” it how to perform.
As with the inference chip, the goal for the training chip is to start with recommendation systems and later use it for generative AI products like chatbot Meta AI, the executives said.
“We’re working on how would we do training for recommender systems and then eventually how do we think about training and inference for gen AI,” Meta chief product officer Chris Cox said at the Morgan Stanley technology, media and telecom conference last week.
Cox described Meta’s chip development efforts as “kind of a walk, crawl, run situation” so far, but said executives considered the first-generation inference chip for recommendations to be a “big success.”
Meta previously pulled the plug on an in-house custom inference chip after it flopped in a small-scale test deployment similar to the one it is doing now for the training chip, instead reversing course and placing orders for billions of dollars worth of Nvidia GPUs in 2022.
The social media company has remained one of Nvidia’s biggest customers since then, amassing an arsenal of GPUs to train its models, including for recommendations and ads systems and its Llama foundation model series. The units also perform inference for the more than 3 billion people who use its apps each day.
The value of those GPUs has been thrown into question this year, as AI researchers increasingly express doubts about how much more progress can be made by continuing to “scale up” large language models by adding ever more data and computing power.
Those doubts were reinforced with the late-January launch of new low-cost models from Chinese start-up DeepSeek (深度求索), which optimize computational efficiency by relying more heavily on inference than most incumbent models.
In a DeepSeek-induced global rout in AI stocks, Nvidia shares lost as much as a fifth of their value at one point. They subsequently regained most of that ground, with investors wagering the company’s chips will remain the industry standard for training and inference, although they have dropped again on broader trade concerns.
STRONG INTEREST: Analysts have pointed to optimism in TSMC’s growth prospects in the artificial intelligence era as the cause of the rising number of shareholders The number of people holding shares of chipmaker Taiwan Semiconductor Manufacturing Co (TSMC, 台積電) hit a new high last week despite a decline in its stock price, the Taiwan Depository and Clearing Corp (TDCC, 台灣集保) said. The number of TSMC shareholders rose to 2.46 million as of Friday, up 75,536 from a week earlier, TDCC data showed. The stock price fell 1.34 percent during the same week to close at NT$1,840 (US$57.55). The decline in TSMC’s share price resulted from volatility in global tech stocks, driven by rising international crude oil prices as the war against Iran continues. Dealers said
PRICE HIKES: The war in the Middle East would not significantly disrupt supply in the short term, but semiconductor companies are facing price surges for materials Taiwan’s semiconductor companies are not facing imminent supply disruptions of essential chemicals or raw materials due to the war in the Middle East, but surges in material costs loom large, industry association SEMI Taiwan said yesterday. The association’s comments came amid growing concerns that supplies of helium and other key raw materials used in semiconductor production could become a choke point after Qatar shut down its liquefied natural gas (LNG) production and helium output earlier this month due to the conflict. Qatar is the second-largest LNG supplier in the world and accounts for about 33 percent of global helium output. Helium is
China is clamping down on fertilizer exports to protect its domestic market, industry sources said, putting an additional strain on global markets that were already grappling with shortages caused by the US-Israeli war on Iran. China is among the largest fertilizer exporters — shipping more than US$13 billion of it last year — and it has a history of controlling exports to keep prices low for farmers. Shipments through the war-blocked Strait of Hormuz account for about one-third of the sea-borne supply. This month, Beijing banned exports of nitrogen-potassium fertilizer blends and certain phosphate varieties, sources said. The ban, which has not
AMAZING ABUNDANCE: Elon Musk has announced plans for a new facility in Texas which would manufacture chips for Tesla and SpaceX to use in robotics and AI Elon Musk said his Terafab project — a grand plan to eventually manufacture his own chips for robotics, artificial intelligence (AI) and space data centers — would be built in Austin and jointly run by Tesla Inc and Space Exploration Technologies Corp (SpaceX). Musk, the chief executive officer of the two companies, said he would start off with an “advanced technology fab” in Austin that would have all of the equipment necessary to make chips of any kind. The project would call for one day supporting 1 terawatt (TW) of computing power per year, the amount Musk expects the companies to
RSS