in this story
Earlier this year, Google (GOOGL-1.34%) and Microsoft (MSFT+0.18%) released separate reports showing that both companies are not on track to meet their climate goals by the end of the decade. They both blamed the same culprit: data centers.
Trump Media may just be a meme stock, says strategist
Despite the environmental damage, data and other artificial intelligence infrastructure will be the winners in the next phase of AI, experts say, as companies look to power their growing AI offerings.
While AI chips developed by companies like Nvidia and AMD are essential to the current stage of AI development, the broader data center industry is “very well positioned to be at the heart of the next phase of AI expansion.” ” said Tejas Dessai, director of the company. He spoke to Quartz about his research at Global X.
“What’s most likely to name a winner is the pick-and-shovel category: who’s building the infrastructure that powers all of this,” said Rowan Trollope, CEO of data platform Redis. he told Quartz. “No matter who wins, which app, which model wins, we will sit in the middle and make them all better.”
Data is a “winner”
As companies deploy more AI clusters, “memory solutions, storage solutions, and networking solutions are likely to become prevalent,” followed by companies specializing in comprehensive data center solutions. Desai said.
“The principle that many of these companies are working on now is that the more chips you can incorporate to train these models, the smarter results you can get from these models. ” Desai said in a separate interview with Quartz.
But he said there are still “many physical constraints,” such as GPU clusters running on hundreds of thousands of chips. Also, it can take years for data centers to come online. This means there is still a “lag” in terms of having enough capacity for AI workloads.
“We’re still early in that cycle, and companies like OpenAI are trying to access data center capacity from anywhere, from anywhere,” Desai said.
Sarah Friar, OpenAI’s chief technology officer, told the company’s shareholders that partner and investor Microsoft has been slow to provide the company with enough computing power. After the company closed a $6.6 billion funding round, the startup’s leaders told some employees that they would not be allowed access to data centers, The Information reported, citing anonymous people familiar with the matter. He said he intends to reduce dependence on Microsoft for AI chips.
Desai said he sees data centers remaining attractive to businesses in the short to medium term as construction increases and vacancy rates reach “all-time lows.”
“Companies actually want to buy almost all the capacity they can get,” Desai said.
“Data is the big winning category,” Trollope says. “I don’t know who will win in the data space, but I think the incumbents have a very good chance.”
The end of climate goals
Microsoft, which set a 2020 goal of becoming “carbon negative” by the end of the decade, announced in May that its carbon emissions were nearly 31% higher than when the goal was set. The company said the increase was mainly due to data center construction and hardware such as semiconductors and servers.
Meanwhile, Google announced in July that its carbon emissions have increased by 48% since 2019 and will increase by 13% in 2023 compared to the previous year. This is mainly due to data center energy consumption. In 2021, the tech giant set a goal to achieve net-zero emissions across its operations and value chain by 2030.
According to the Electric Power Research Institute, data centers could consume up to 9% of the United States’ electricity by the end of the decade, more than double the electricity used today.
Tech giants are looking to nuclear power as a solution
Ami Badani, chief marketing officer at British chip design company Arm, said in April that data centers running AI chatbots such as OpenAI’s ChatGPT account for 2% of global electricity consumption. . According to Goldman Sachs research, a single ChatGPT query requires almost 10 times more power than a Google search (GS-0.16%). Badani said this level of energy demand could ultimately slow progress in AI.
Big tech appears to be taking notice of this obstacle. In July, the Wall Street Journal reported that one-third of U.S. nuclear power plants were in talks about contracts with tech companies to power data centers.
“We have to make up for this energy deficit in some way,” Desai said. “We can’t burn any more coal, we can’t use any more fossil fuels, so nuclear energy is the obvious answer.”
Big tech companies appear to be moving toward nuclear energy, and “we’ll continue to see more deals in that direction,” he said.
Earlier this week, Google announced it had signed “the world’s first corporate agreement to purchase nuclear energy” from small modular nuclear reactors (SMRs) developed by California-based Kairos Power.
Google said it plans to bring Kairos Power’s first SMR online by the end of this decade, with others planned by 2035. The agreement will make 500 megawatts (MW) of carbon-free electricity available to the U.S. power grid 24 hours a day, 365 days a year.
Amazon (AMZN +0.38%) also signed a deal this week to “support the development of nuclear power projects,” including building “several” SMRs that “have a small physical footprint and can be built close to the power grid.” . said the company. Compared to traditional nuclear reactors, smaller SMRs require less construction time and can be brought online faster, Amazon said.
In September, Constellation Energy (CEG -3.03%), which owns most of the country’s power plants, announced it had signed a 20-year power purchase agreement with Microsoft. The agreement will restart Three Mile Island’s Unit 1 reactor and launch the Crane Clean Energy Center. The CCEC, scheduled to be operational by 2028, will add more than 800 MW of carbon-free power to the grid, according to a study by the Pennsylvania Building Construction Trades Council.
