The AI boom is an environmental nightmare. Here’s how tech companies, investors and startups are trying to tackle the challenge.

Welcome back to LinkedIn News Tech Stack, which brings you news, insights and trends involving the founders, investors and companies on the cutting edge of technology, by Tech Editor Tanya Dua. You can check out our previous editions here.

Pitch me the interesting investors, founders, ideas and companies powering emerging technologies like AI to reach the inboxes of nearly 1 million subscribers plus thousands more on LinkedIn. Follow me for other tech updates. And click 'Subscribe' to be notified of future editions.

ICYMI: Catch this week's edition of VC Wednesdays with Toyota Ventures ' founder and general partner Jim Adler here.

A deep dive into one big theme or news story every week.

Artificial Intelligence comes with a hefty price tag – not just because of how expensive it is to actually train and run AI models, but also due to its considerable environmental impact.

Last week, 谷歌 revealed that its greenhouse gas emissions have jumped nearly 50% over the past five years — because the data centers it uses to power AI and other applications are using more power than ever before.

Emissions are just one part of the equation in AI's soaring environmental costs. The technology is also known to have a voracious appetite for both energy and water. AI’s rapid growth could cause U.S. electricity consumption to "outstrip current supply in the next two years," Bernstein analysts recently said. Meanwhile, a recent 高盛 report found that the demand for power driven by AI applications is poised to increase 160% by 2030 – and also estimated that using ChatGPT for an answer sucks up to 10 times as much electricity as a basic Google search.

"That kind of spike in power demand hasn't been seen in the U.S. since the early years of this century," the Goldman analysts said in the report.

With AI only growing in scale and scope, and thereby getting more resource-intensive, tech companies, investors and startups are trying to balance its promises with its environmental footprint, investing in areas like renewable energy sources and energy-efficient processing.

“It's a really important problem, given that models are only going to get bigger and how energy intensive they are,” New Enterprise Associates (NEA) partner Aaron Jacobson said in a recent edition of VC Wednesdays. “We owe it to the world to figure out how to run these as efficiently as possible.”?

More clean and efficient data centers

One way stakeholders are trying to limit carbon emissions related to training and using AI is by using or designing data centers powered by cleaner energy sources like hydropower, wind and even nuclear power — versus coal.?

OpenAI founder and CEO Sam Altman and VC firm Andreessen Horowitz seem to be part of this camp, with their backing of the startup Exowatt , which aims to meet such clean-energy needs by combining solar thermal tech with a thermal battery system in modules that can be stored and deployed near data centers.

Other tech companies are similarly trying to make their AI infrastructure more efficient by investing heavily in clean energy sources for their data centers.?

Earlier this year, LinkedIn parent 微软 signed its biggest-ever corporate power purchase agreement to run its data centers on carbon-free power. And recently, Google made a major investment in BlackRock-backed Taiwanese renewable energy player New Green Power 永鑫能源 to power its data centers.?

Just this week, 亚马逊 said that it had reached a crucial climate goal by using electricity in its operations from sources that didn't produce greenhouse gas emissions. The company claimed that it essentially offset its electricity use via its more than 500 solar and wind projects, but some critics argued that since those projects don't directly power Amazon's operations, the company may be offering "a misleading impression of its effect on the climate."

“Advances in AI have depended on exponential growth in training data and thus computing power; as these power requirements grow in the era of deep learning, AI is spurring a boom in clean electricity,” said Izzy Woolgar , director of external affairs at the non-profit energy research institute Centre for Net Zero (Octopus Energy Group) . “The race to invest in vast new data centers — and the green energy to power them — is on.”

New AI chips and cooling technologies

英伟达 ’s meteoric rise during the current AI boom may suggest otherwise, but the company’s GPUs, or specialized server chips that have become must-haves for running AI software, were not originally designed for that purpose. Plus, the computer architecture of GPUs requires far more electrical power than traditional CPUs.

That’s given rise to a slew of other players who are now taking aim at developing new processor prototypes in a bid to train AI models faster and more efficiently. One example is Celestial AI , a startup that is building chips based on energy-efficient light rather than electricity.?

Another is Etched , which recently raised $120 million in Series A funding to develop a specialized AI chip designed to run a specific type of AI model known as transformers, the same type used by OpenAI’s ChatGPT and Google’s Gemini. The singular purpose of its chip makes it less power-intensive and more than 20 times faster than an Nvidia GPU on the same power budget, co-founder and COO Robert W. Wachen said.

“The GPUs are getting better slowly, the next-gen NVIDIA GPUs will have 30% better power efficiency than the current ones. We think that’s not nearly fast enough — every big tech company is building a $100B AI chip cluster, dramatically outpacing the improvements in energy-efficient hardware,” Wachen said. “Specialized chips can get way more performance within the same power budget. That means the power of a 1-gigawatt data center running GPUs can be replicated with less than 100-megawatt of power running ASICs (application specific integrated circuits) like Etched.”

Not just startups, other cloud companies including Amazon, Google and Microsoft and chipmakers like AMD and 英特尔 are also ramping up development of new chips and alternatives to GPUs. When Intel recently announced its AI accelerator, the Gaudi 3, it touted that it was faster at training models and better at inference — the process in which the models actually use their training to respond to queries and prompts — than Nvidia’s H100.

New methods of cooling data centers are also gaining steam. Data centers are increasingly turning to liquid cooling, which use less power — and surprisingly, less water — than air cooling systems, Andrew Schaap , CEO of Aligned Data Centers , notes in this Forbes article. In fact, liquid cooling can shrink facility power by almost 20% and total data center power by 10% compared to air cooling alone, according to a recent study.

“We are seeing improvements made in cooling and energy management, both in the software and hardware layer,” said NEA’s Jacobson. “For example, Phaidra leverages AI to improve the power efficiency of data center cooling and Frore Systems has developed a solid-state chip for more efficiently cooling AI systems.”

AI is both part of the problem and the solution

Paradoxically, while AI is itself contributing to an energy crunch, some experts believe it is also the technology that can help solve it.?

“I believe the net energy impact of AI will be positive as it will enable significant advancement in energy-efficient technology such as new batteries, superconductors and power control algorithms, as well as new methods for clean energy production altogether, such as fusion,” NEA’s Jacobson said.

He’s not wrong. Companies are using AI as a tool to find solutions to climate and environmental challenges across the board, from weather forecasting and managing the energy load of the physical grid to developing research materials that recapture carbon from the atmosphere.?

Startup ClimateAi , for example, is using AI to evaluate how vulnerable crops are to warming temperatures over the next two decades. Its tool uses data on the climate, water and soil of a particular location to measure how viable the landscape will be for growing in the coming years. Meanwhile, startups like Plexigrid and Bidgely have AI software tools that automatically adjust household energy consumption based on usage at a certain time.

“AI’s potential to help or hinder the energy transition is under growing scrutiny, but it’s an incredibly powerful tool for designing intelligent future energy systems benefitting every sector,” Centre for Net Zero’s Woolgar said. “Early adoption already helps charge EV fleets overnight and optimize battery storage, and there are other exciting, emerging use cases.”

Here’s where we bring you up to speed with the latest advancements from the world of AI.

• Microsoft, Apple drop OpenAI board. 微软 will leave its non-voting observer seat on OpenAI's board, saying it is confident in the company's direction and is no longer needed, Axios reports, citing a memo. LinkedIn's parent company took up the role about eight months ago, when Sam Altman returned to OpenAI as CEO and the board was reconstituted. Meanwhile, 苹果 , which was reportedly planning to join the board in a similar observer position, will no longer do so, the Financial Times reports, citing an anonymous source. The moves come amid growing regulatory concerns over Big Tech’s tie ups with artificial intelligence.?
• Top VC firm builds AI chip stash. Andreessen Horowitz is not just providing financing to a number of AI startups — the venture capital firm has now amassed thousands of AI chips, including Nvidia H100 GPUs, and is renting them to startups, The Information reports, citing anonymous sources. The firm intends to increase its stockpile to include more than 20,000 GPUs, which are required to train and run the large language models that underpin AI. The initiative, known as “Oxygen,” lends the chips to startups at discounted prices or, in some cases, in exchange for equity.?
• Intuit is laying off about 1,800 workers, or 10% of its workforce, in a strategic shift toward AI. The maker of TurboTax will then hire 1,800 new employees primarily in engineering, product and customer-facing roles starting next month, according to a letter from CEO Sasan Goodarzi. More than 1,000 of the people being let go were not meeting performance expectations, the company said. Intuit expects “our overall headcount to grow in FY25 and beyond,” Goodarzi said. Offices in Boise, Idaho, and Edmonton, Alberta, will be closed as part of the reorganization, which is expected to cost up to $260 million.

Here’s a list of other notable AI developments from this week:??

• Chipmaker AMD is acquiring Finnish AI startup AMD Silo AI for about $665 million as the company tries to enhance its AI chip capabilities to compete against industry leader Nvidia. See Silo AI co-founder and CEO Peter Sarlin ’s LinkedIn post here for more.
• Amazon Web Services (AWS) announced a number of updates across several of its AI products this week, including the ability to fine-tune LLMs offered in its Bedrock product and new guardrails that detect hallucinations better.
• 埃森哲 has acquired two companies in the semiconductor design space including Cientra and Excelmax Technologies as it seeks to build up its custom silicon offerings for global clients. See its tech group chief executive Karthik Narain ’s LinkedIn posts for more here and here.
• Guy Podjarny , the founder of cybersecurity company Snyk , has launched a new company called Tessl to tackle software development for the AI era. See his post on LinkedIn for more.
• AI startup Patronus AI has launched a new model that specifically detects hallucinations called Lynx, which it says beats OpenAI’s GPT-4o and other state-of-the-art LLMs on hallucination tasks. See co-founder and CEO Anand Kannappan ’s LinkedIn post for more.??

Catch up on the tech headlines you may have missed this week and what our members are saying about them on LinkedIn.

• Nearly 10 billion passwords have been leaked and posted to a hacking forum, in what cybersecurity researchers are calling the largest hack of all time. The leak, known as RockYou2024, combines data from old breaches with 1.5 billion new, real-world credentials. It could enable "brute-force attacks," which use trial and error to rapidly test a large number of passwords and gain access to systems that aren't protected. Analysts at Cybernews warn that it increases the risk of "credential stuffing attacks," which have already targeted companies including Ticketmaster and Santander.?
• Tesla has lost its EV market majority. The company no longer makes a majority of America's electric vehicles, The New York Times reports, citing data from research firm Cox Automotive. Its share of the market fell to 49.7% in the second quarter, compared with 59.3% in the same period a year earlier. With more than 100 EV models on the market in the U.S., Tesla is facing increased competition from established automakers. General Motors, Ford, Hyundai and Kia have made inroads in the fast-growing segment. Total EV sales climbed 11.3% in the quarter, and 8% of all cars sold or leased were electric.
• 苹果 has agreed to open up its contactless payment technology to other providers for free for a decade, EU regulators announced Thursday — ending their investigation without a major fine. Apple will let developers use its “tap-and-go” technology for Apple Pay and Apple Wallet, after the EU formally launched an antitrust investigation in 2020, citing an abuse of market power. It comes just as Apple this week solidified its position as the stock market's most valuable company by becoming the first to exceed a $3.5 trillion market capitalization. Relatedly, the Apple Watch is getting a makeover as it nears its 10th birthday. The next versions of the smartwatch will offer larger displays and thinner designs; they'll also get a new, more AI-friendly chip, though none are yet slated to run Apple Intelligence, Bloomberg reports.?
• Meanwhile, 三星电子 unveiled its first "smart ring" on Wednesday as part of its push into health tracking. The Galaxy Ring is being marketed as a companion to Samsung's smartwatches and smartphones, which together can "offer a comprehensive picture of a person’s health" — and also, Samsung hopes, perhaps offset slow phone sales. The South Korean electronics giant is also debuting its newest Galaxy Watch models, which will be able to track the wearer's metabolic health. Both Samsung and competitor Apple are betting on health monitoring tools as a way to bring in — and retain — new customers. Meanwhile, thousands of Samsung employees are on strike for better pay and working conditions.

• Is Spotify turning into a social app? Spotify has added commenting to its podcasts, another step toward becoming "a social network centered around all things audio, not just a music-streaming app," according to TechCrunch. This follows an update last year that added a TikTok-inspired discovery feed, artist profiles with concert-ticket and merch sales, and interactive features like polls and Q&As. There's potential for comments to be added for music artists, as well, Spotify's head of podcast product told TechCrunch. With this incremental approach, Spotify is looking to increase user engagement and ultimately compete with large social networks for ad revenue.

Here’s keeping tabs on key executives on the move and other big pivots in the tech industry. Please send me personnel moves within emerging tech.

• OpenAI has hired former Meta public policy manager Martin Signoux as its AI policy lead in Europe, he first shared on LinkedIn.
• AI hardware startup Rain AI has hired senior Apple engineering exec J-D allegrucci as its head of hardware engineering, he shared first on LinkedIn.
• Palantir Technologies product exec Peter Wilczynski is leaving the company after 12 years to become the chief product officer at geospatial intelligence company Maxar, he shared on LinkedIn.
• Two former Humane execs including strategic partnerships lead Brooke Hartley Moy and head of product engineering Ken Kocienda have launched a new startup called Infactory , a fact checking-based search engine. See Hartley Moy’s LinkedIn post for more.

As always, thanks for reading. Please share Tech Stack if you like it! And if you have any news tips, find me on InMail.