How Forbes Compiled The 2025 AI 50 List
Forbes' AI 50 list spotlights the most promising privately-held companies applying artificial intelligence to solve real-world challenges. Created in partnership with data partners Sequoia and Meritech Capital, the seventh annual installment reflects how companies across industries are using AI tools to streamline and even fully take over tasks previously done by humans.
The selection process begins with both a public call for nominations and targeted outreach to thousands of eligible startups by our partner, Meritech Capital. There is no fee to apply. Nominees are asked to provide both qualitative insights into aspects such as their business models, technical talent and how they're building and using AI-enabled technology—as well as quantitative metrics including valuation, revenue and fundraising history. Companies may opt to submit some data confidentially. While optional, we also encourage submission of diversity-related metrics.
Forbes
In parallel, Forbes, Meritech and Sequoia identify standout AI companies that may not have applied but merit evaluation. For these, we rely on publicly accessible sources such as PitchBook, Crunchbase and LinkedIn.
Since 2023, eligibility has extended beyond North America to include privately-held companies worldwide. This year, Forbes received 1,861 submissions—making this year one of our most competitive yet.
An initial quantitative review of nominees is conducted using a proprietary algorithm developed by Sequoia Capital partner Konstantine Buhler. This model considers a mix of company-submitted data and external signals (including employee satisfaction scores, funding history and customer traction) to assess three core dimensions: financial strength, organizational culture and diversity. While withholding data does not harm an application, submitting comprehensive details can improve visibility in the review process.
From there, the top 100 companies are reviewed by a panel of expert judges with backgrounds in academia, research and leading technology firms (full list below). These reviewers assess the nominees' technical rigor, innovation potential and team strength. The most promising 70 companies are then evaluated by a group of seasoned AI investors, who weigh in with market-focused perspectives on competitiveness and scalability. Any judge with a financial stake in a candidate company is recused from evaluating it.
The final AI 50 list is curated by Forbes editors and reflects the most compelling AI startups of the year. Companies are presented alphabetically and are not ranked.
Expert Judges
Joy Buolamwini
Dr. Joy Buolamwini is the founder of the Algorithmic Justice League, a research institute that focuses on the harms and social implications of artificial intelligence. As a researcher at MIT, her paper titled 'Gender Shades' revealed how facial recognition systems from leading companies like IBM, Amazon and Microsoft misgendered women and darker-skinned individuals. She was the protagonist of an Emmy-nominated documentary called "Coded Bias" which demonstrated the dangers of algorithmic biases. She has also advised Congress and other government agencies on equitable AI policy.
Tanya Berger-Wolf
Tanya Berger-Wolf is a professor of computer science engineering, electrical and computer engineering, and evolution, ecology and organismal biology at Ohio State University, as well as the director of the school's Translational Data Analytics Institute. Berger-Wolf serves as a scientific advisor for organizations including the U.S. National Academies Board on Life Sciences, the U.S. National Committee for the International Union of Biological Sciences, the Global Partnership on AI and The Nature Conservancy.
Thomas Dohmke
Thomas Dohmke is CEO of Microsoft-owned GitHub. He has overseen the launch of the world's first at-scale AI developer tool, GitHub Copilot, as well as its successors Copilot Workspace and GitHub Models. Dohmke previously cofounded HockeyApp and led the company as CEO through its acquisition by Microsoft in 2014. Prior to taking the helm at Github, he was a vice president and chief product officer.
Joe Edwards
Joe Edwards is a director of product marketing at UiPath, an AI-powered workforce automation company with a $6 billion market capitalization. Prior to UiPath, Joe was a program manager within the U.S. Treasury Department, where he led a team that built the agency's first apps for mobile check deposit and digital ticketing.
R. David Edelman
R. David Edelman is the director of the Project on Technology, the Economy & National Security at the MIT Internet Policy Research Initiative. He has been a global strategy and public policy lead for various startups; a venture capital investor in deep tech companies; and a special assistant to President Barack Obama on issues of economic and technology policy. In that role, he coauthored the federal government's foundational AI strategy.
Anna Korhonen
Anna Korhonen is a professor of artificial intelligence at the University of Cambridge where she is the co-director of the Cambridge Language Technology Lab and the director of the Centre for Human-Inspired Artificial Intelligence. Her research focuses on natural language processing and AI for social and global good.
Jason Mars
Jason Mars is a professor of computer science at the University of Michigan. His research focuses on conversational artificial intelligence, computer vision and natural language processing. Mars has cofounded several companies with valuations that have exceeded $200 million. Most recently, he is the founder and CEO of Jaseci Labs, a company that provides open-source tools for AI product development.
Nashlie Sephus
Nashlie Sephus is the principal AI/ML evangelist for AWS AI, where she focuses on fairness and identifying biases in artificial intelligence. After joining Amazon as an applied science manager in 2016, she led multiple product launches including visual search for replacement parts on the Amazon Shopping app in 2018. She is also the CEO and founder of Bean Path, a nonprofit focused on local technical assistance and economic development in Jackson, Mississippi.
Sophie Lebrecht
Sophie Lebrecht is the chief operating officer at Ai2, a nonprofit AI research institute founded by Paul Allen. Prior to joining Ai2, Lebrecht led next-generation Machine Intelligence efforts at Apple while also leading M&A for Apple's AI/ML division. She was the Senior Vice President of Strategy and Operations and led product development and partnership deals at Xnor.ai, an Ai2 spinout that created low-power AI tools that weren't based on the cloud, until it was acquired by Apple in 2020.
Investor Panel
Konstantine Buhler
Konstantine Buhler is a partner at Sequoia Capital, investing in AI and machine learning. Buhler—who also served as a data partner for Forbes' 2025 AI 50 List—is a partner at security tech company Verkada, biotech company Xaira and AI-fueled security company XBOW, among others. Before joining Sequoia in 2019, Buhler—who was also an angel investor in OpenSea and Viz.ai—was a principal at Meritech Capital.
Sarah Catanzaro is a general partner at Amplify Partners, where she focuses on investing in and advising startups in AI, data management and distributed systems. Catanzaro previously worked on data strategy and led data science and AI teams for defense tech companies including Mattermark, Palantir, Cyveillance and the Center for Advanced Defense Studies.
Sameer Dholakia
Sameer Dholakia is a partner of the growth investment practice at Bessemer Venture Partners, where he invests in cloud and AI startups. Before Bessemer, Sameer served as the CEO of SendGrid in 2014, leading the company through a successful IPO in 2017 and subsequent acquisition by cloud communications company Twilio for approximately $3 billion in 2019.
Miles Grimshaw
Miles Grimshaw is a general partner at Thrive Capital, where he's invested in AI, applications, developer infrastructure and consumer tech. He's led investments in Github, Slack and ScaleAI—and has partnered with founders at AI-powered code editor Cursor, digital operations platform Airtable and employee engagement software Lattice, to name a few. Miles has served on the boards of multiple companies, including Benchling, Segment, Monzo, Chai, LangChain, Glide and Supergreat.
Sarah Guo
Sarah Guo is the founder and a partner at Conviction, a venture capital firm launched in 2022 to invest in intelligent software. Prior to starting her own firm, she spent a decade as a general partner at Greylock Partners. Guo has been an early investor or advisor to more than 50 companies, and co-hosts an AI podcast called 'No Priors' with Color cofounder Elad Gil. In January, Conviction closed on its second fund with $230 million, twice the size of the firm's first.
Aaref Hilaly
Aaref Hilaly is a partner at Bain Capital Ventures, investing in early-stage AI companies, including Contextual AI, Decagon and EvenUp. Before Hilaly was an investor, he was an entrepreneur, cofounding two companies that were acquired by Sun Microsystems and Symantec. Prior to his role at BCV, Hilaly spent seven years as a partner at Sequoia and served on the boards of biotech company Guardant (NASDAQ: GH) and revenue platform provider Clari.
Jocelyn Goldfein
Jocelyn Goldfein is a managing director at Zetta Ventures Partners, where she invests at the seed stage in AI-native application and infrastructure companies. She has backed startups like AI infrastructure platform Weaviate and Databricks-acquired firm Tabular. Goldfein previously worked as an engineering executive at VMware and Facebook. She launched her career as a software engineer and entrepreneur after studying computer science at Stanford.
Christopher Manning
Christopher Manning is a partner at AIX Ventures, an investment firm leveraging industry expertise to advance the AI space. Manning is also an associate director of Stanford University's Institute for Human-Centered Artificial Intelligence, and the school's inaugural Thomas M. Siebel professor in machine learning. Considered a pioneer in natural language processing, he was previously the director of Stanford's AI lab and has received two ACL Test of Time Awards and the IEEE John von Neumann Medal.
Ravi Mhatre
Ravi Mhatre is a founding partner at Lightspeed Venture Partners. With over 25 years of venture capital experience under his belt, he has invested in a broad spectrum of tech companies and played a key role in Anthropic's latest funding round. He sits on the board of Anthropic and has also invested in companies like Anduril, AppDynamics, Glean, Guardant, MuleSoft, Natera, Nutanix, Offchain Labs, Rubrik, ThoughtSpot, Zscaler and Wiz.
Matt Murphy is a partner at Menlo Ventures, where he backs companies from pre-revenue to growth in focus areas including AI infrastructure stack, developer workflows and AI-first applications. Murphy has been a venture capitalist for more than two decades, working at Kleiner Perkins before joining Menlo in 2015. He began his career as a product manager at Sun Microsystems and NetBoost, a startup which was later acquired by Intel. In July 2024, Menlo Ventures announced a $100 million fund in partnership with Anthropic to invest in up and coming AI startups.
Rudina Seseri
Rudina Seseri is the founder and managing partner of Glasswing Ventures, an early-stage venture capital firm that invests in enterprise and security companies using AI. It has more than $300 million in assets under management. Seseri has nearly two decades of investing and operational experience in high-growth companies across enterprise software, data infrastructure and automation. She is also an executive fellow at Harvard Business School and a Connection Science Fellow at MIT.
Rob Toews is a partner at Radical Ventures and leads its San Francisco Bay Area office. Toews previously led AI investment as an investor at Highland Capital. He spent several years in the world of autonomous vehicles, helping lead the strategy team at Zoox (acquired by Amazon for $1.3 billion in 2020) and working on autonomous vehicle policy in the White House under President Barack Obama. In August, the firm raised $800 million in new capital to invest in later-stage AI startups.
Rob Ward
Rob Ward is the cofounder of Meritech Capital Partners which he cofounded in 1999 alongside Mike Gordon and Paul Madera. Ward currently helps lead the firm's investment in infrastructure, application software, and data and analytics companies. He has been an integral part of numerous Meritech investments over the past 25 years, including companies like Facebook, NetSuite, Roblox, Salesforce, Snowflake and Tableau.
Emily Zhao
Emily Zhao is a principal at Salesforce Ventures where she leads the fund's AI infrastructure investments. Zhao partners with founders from Series A onwards to help them scale their go-to-market strategy. She has backed companies like Together AI, Hugging Face, Anthropic, Runway and World Labs. Zhao launched her investment career in private equity at Blackstone, and before joining Salesforce Ventures three years ago, she worked as an investor at New York-based Avenir Growth Capital.
Forbes
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Forbes
26 minutes ago
- Forbes
Why Companies Using AI Should Increase Prices
AI can be a game changer for businesses, but it can also be an expensive drag on the balance sheet. A new report from billing software provider Chargebee found that companies that grew most in the last year were the ones that changed their pricing strategies to account for AI. The ones that were most successful combined a variety of pricing models: recurring subscriptions, usage-based models, outcome-based models and flat fees. Four in five of the companies surveyed that added AI said they are also changing their pricing. But how to adjust prices, especially in a time of economic uncertainty, is a challenge. Just over half said customer retention is their top concern, but 40% of businesses that adjusted prices last year reported a disconnect between increases and customer value. Nearly a quarter of companies struggled with explaining the benefits of adding AI functions to their services, while technical issues also caused struggles. Most SaaS providers have traditionally charged enterprises based on individual licenses, which is far different from a usage fee. Many companies, the study found, are testing out a variety of pricing structures to see what works best. A vignette in the study from AI fiction writing tool Sudowrite's founder Amit Gupta laid out an issue they had with their original flat subscription-based pricing model: There were users paying $20 a month for the service, but their use of AI was costing the company $400 a month. They started charging by the word—and letting users develop custom tools—but that also could be more unsustainable if the users did a lot to modify those tools. Now they have a credit and usage-based model, but the company is still experimenting with it. While decisions on pricing and value calculations often come from the CFO or CEO, input from the CIO—who best understands cost, usage and value of AI-driven functions—is vital to driving revenue growth through a company's AI transition. CIOs also know which functions are needed for their enterprise, and which price increases may be worth budgeting for. Use of AI and new technology depends on infrastructure as well as service, and updating the technological capacity of a place can be a huge undertaking. The Kraft Group, which owns businesses including the New England Patriots and their home field Gillette Stadium, recently signed a five-year agreement with infrastructure provider NWN to upgrade the stadium and other playing facilities. I talked to Kraft Group CIO Michael Israel and NWN CEO Jim Sullivan about how they are preparing these mega-facilities for a more connected future. An excerpt from our conversation is later in this newsletter. The Clinton Clean Energy Center in Clinton, Illinois. All of the things AI can do with computing need power, and a lot of it. This week, Meta made a deal to get the power it needs, signing a 20-year agreement to buy all of the power produced at a Constellation Energy nuclear plant in Clinton, Illinois. The agreement starts in June 2027—after an existing state agreement runs out—and will expand the plant's output. Meta has prioritized finding sources of nuclear power, both with new plants and utilizing existing ones, to support its technology going forward. The company announced an RFP for nuclear energy developers in December, and says it has shortlisted potential new nuclear power resources. This is the second deal a tech company has made with Constellation to redevelop its nuclear plants for AI. In September, Microsoft announced a 20-year deal with the power provider for one of its reactors at its Three Mile Island facility in Pennsylvania (not impacted by the 1979 meltdown). Constellation has said it expects to restart the reactor by 2028. Amazon and Google have both been investing in small nuclear reactors, and Google announced an investment in three advanced nuclear energy projects by Elementl Power. These deals, coupled with four executive orders from President Donald Trump aimed at bolstering nuclear power, seem to be heralding a new nuclear power age in the U.S., writes Forbes senior contributor David Blackmon. In the meantime, tax incentives for renewable energy sources, including solar panels and wind turbines, are in line to be cut in Trump's latest budget. Still, Forbes senior contributor Ken Silverstein writes, renewable energy is touted by many as the fastest and least expensive way to get more power into the grid—and could also play a huge role in generating the electricity needed for the AI-driven future. Commerce Secretary Howard Lutnick. The Trump Administration is reorganizing the AI Safety Institute into a new group: the Center for AI Standards and Innovation. Forbes' Thomas Brewster writes that there are few obvious changes. The AI Safety Institute was created in 2023 as part of President Joe Biden's AI executive order, and operated within the National Institute of Standards & Technology to research risks in AI systems. The revamped Center for AI Standards and Innovation is still part of NIST but is under the purview of Commerce Secretary Howard Lutnick, and positions itself as taking a bulwark against 'censorship and regulations [that] have been used under the guise of national security. Innovators will no longer be limited by these standards.' But the statement about the revamped organization says that this group will also work on research to measure and improve security of systems, evaluating where risks exist. At this juncture, it's difficult to see what the big difference in purpose is—except, of course, that this iteration has a different name and was established by a different president. Outside of the government sphere, computer scientist Yoshua Bengio, often referred to as the 'godfather of AI,' launched a new nonprofit organization aimed at creating AI systems prioritizing safety, writes Forbes senior contributor Leslie Katz. LawZero is starting with $30 million in funding and is assembling a team of world-class AI researchers to work on a system called Scientist AI—a non-agentic system that behaves in response to human input and goals. 'Such AI systems could be used to provide oversight for agentic AI systems, accelerate scientific discovery and advance the understanding of AI risks and how to avoid them,' the statement on the organization's website says. 'LawZero believes that AI should be cultivated as a global public good—developed and used safely towards human flourishing.' Jakub Porzycki/NurPhoto via Getty Images All USB-C ports are not created equal, and Microsoft intends to do something about it. Ports don't always have the same capabilities now, meaning users sometimes plug peripherals into their computers and they don't work. Forbes senior contributor Barry Collins writes that the company will establish minimum settings for all USB-C ports on its computers, meaning things will work every time they are plugged in. But while this will be a welcome change, it is a hardware one—meaning it will make no difference until users actually upgrade their physical computers. Kraft Group CIO Mike Israel and NWN CEO Jim Sullivan. In April, the Kraft Group—which owns the New England Patriots, MLS' New England Revolution and Gillette Stadium—signed a five-year agreement with tech infrastructure provider NWN to transform the tech framework for the Kraft Group's facilities, including Gillette Stadium and a new training facility for the Patriots. I talked to Kraft Group CIO Michael Israel and NWN CEO Jim Sullivan in April about the challenges of bringing the latest technology infrastructure to a place like Gillette Stadium, and how big facilities with a multitude of uses can plan for the future. This conversation has been edited for length, clarity and continuity. A longer version is available here. How do you come up with what you want to accomplish, and how do you figure out what kind of infrastructure is needed to make it happen? Israel: We have our Monster Jam [this weekend] at the stadium. On Saturday, I'll be walking around the stadium engaging. Gillette is one of the few stadiums in the country in which we own and self-operate our stadium. It's our security staff, our concession staff. I am walking around watching how our fans engage with us. How are our systems being used? Where are they inefficient? Where are they doing their job? How can we improve that experience? Guests will come up to me and say, 'How do I get to the gate?' 'How do I get to my suite?' 'Where's the nearest bathroom?' When they're asking me these things, that's registering in my mind: They don't have that information today. When you have these types of events, they're new users. You want them to have a positive experience because that's your lead in to a potential soccer season ticket holder or future Patriots season ticket holder. But even when you get to the Patriot season ticket holders, what can we do to enhance that experience? It's seeing how our guests experience things, what we're doing right, what we're doing wrong, and not sitting back and saying, 'I'm good. I'm going to go watch the game.' I've been here six years. I think I've seen 15 minutes of one football game. On a football day, I'll generally do about 30,000 steps walking the stadium, watching what's going on. The other side of the coin is what do we do from a technology perspective, looking at what do I need to do to ensure that I have connectivity, and what devices are now connecting to the networks that hadn't in the past? The system that waters the field is an IoT system that's attached to our network. If it's not connecting, it's not watering the field, and we don't know what's going on. We have to allow for connectivity, secure that connectivity, and make sure that that connectivity is reliable. We constantly do surveys after a winter to say, what got impacted by the winter? Do I have Wi-Fi access points that may be misaligned that need to be looked at? I have FIFA coming in next year. That's like having seven Super Bowls over six weeks here at the stadium. They're going to use my parking lots on the east and west side as fan activation zones. I don't have connectivity there. Working with the NWN team, we have to determine what does FIFA need? How do we light up those areas that when the fans are there, they have connectivity. Their booths can operate, their systems can operate, the fans can get connectivity, and we'll have maybe 50,000 or 60,000 fans in the bowl, but we could have an additional 50,000 people on the campus, and that's not a crowd that we're used to having. On the NWN side, how do you determine what to do in large facilities like these and how to support current and future needs? Sullivan: NWN, over the past five years, has grown from $250 million to over $1 billion dollars this year, and really expanded this full end-to-end IT infrastructure. The market is changing really fast with AI adding in. For us, it's working with organizations to start with the end state: What's the vision of what we have to have here, what we're trying to drive? And then, what's the required capabilities? Most of these environments, from the application, to the AI, to the infrastructure, to the unified communications, to security all have to be assessed as one holistic solution. Then we put in the right required capabilities from the technology, the services, the overall management, and co-management with Mike's team. We've reached the breadth and scale where we're dealing with organizations where there's hundreds of thousands of people, or states with 50,000 deployments of people and requirements. We can cover end-to-end, but also have the scale to handle a large project that goes across multiple technology domains and supports a smaller event, and all of a sudden it surges to hundreds of thousands of people. Everyone is getting used to [the fact] that customer experience needs to be world-class; there's expectations there. Ultimately, the Wi-Fi is going to be fast, strong and secure. And then going into these new technologies to a real beneficial evolution, where it's creating new user experience, new knowledge, but is also driving a backend that's going to create a lot more capacity demands on the networks, on the infrastructure. You've got to be able to tie it all together. What advice do you have for CIOs looking to bring more technology to their facilities, thinking about not only what to do today, but what to do in the future? Israel: Ultimately, it's not just if you build it, they will come. If you're building it, you need to be brainstorming how you're going to use it, and you need to have relationships with all of your stakeholders to understand what's holding them back, what would they like to see? In some cases, they don't know what they don't know, and we have to take these technology discussions, take the technology out of it, and think about how are we going to provide solutions. Sullivan: We did 5,000 distinct deployments last year. The really successful ones are driven with a positive outcome you're trying to get to. Collaboration and a partnership between the two as a seamless team really drives the most success to drive those outcomes. With the rise of agentic AI, non-human agents will soon have the ability to do significant work tasks, like the more technical jobs of database administrators. That seems like a huge leap, but here's how to start warming to the idea and embracing this kind of AI agent (even if you're a DBA). You may be an introvert, but that doesn't mean you can't be a successful leader. Here are 10 things you can do to become a thriving manager. Which Big Ten university just introduced an AI Fluency initiative, embedding AI education into the core of all undergraduate curriculum, regardless of major? A. Northwestern University B. Purdue University C. Ohio State University D. Michigan State University See if you got the answer right here.
The Hill
38 minutes ago
- The Hill
Anthropic CEO: GOP AI regulation proposal ‘too blunt'
Anthropic CEO Dario Amodei criticized the latest Republican proposal to regulate artificial intelligence (AI) as 'far too blunt an instrument' to mitigate the risks of the rapidly evolving technology. In an op-ed published by The New York Times on Thursday, Amodei said the provision barring states from regulating AI for 10 years — which the Senate is now considering under President Trump's massive policy and spending package — would 'tie the hands of state legislators' without laying out a cohesive strategy on the national level. 'The motivations behind the moratorium are understandable,' the top executive of the artificial intelligence startup wrote. 'It aims to prevent a patchwork of inconsistent state laws, which many fear could be burdensome or could compromise America's ability to compete with China.' 'But a 10-year moratorium is far too blunt an instrument,' he continued. 'A.I. is advancing too head-spinningly fast. I believe that these systems could change the world, fundamentally, within two years; in 10 years, all bets are off.' Amodei added, 'Without a clear plan for a federal response, a moratorium would give us the worst of both worlds — no ability for states to act, and no national policy as a backstop.' The tech executive outlined some of the risks that his company, as well as others, have discovered during experimental stress tests of AI systems. He described a scenario in which a person tells a bot that it will soon be replaced with a newer model. The bot, which previously was granted access to the person's emails, threatens to expose details of his marital affair by forwarding his emails to his wife — if the user does not reverse plans to shut it down. 'This scenario isn't fiction,' Amodei wrote. 'Anthropic's latest A.I. model demonstrated just a few weeks ago that it was capable of this kind of behavior.' The AI mogul added that transparency is the best way to mitigate risks without overregulating and stifling progress. He said his company publishes results of studies voluntarily but called on the federal government to make these steps mandatory. 'At the federal level, instead of a moratorium, the White House and Congress should work together on a transparency standard for A.I. companies, so that emerging risks are made clear to the American people,' Amodei wrote. He also noted the standard should require AI developers to adopt policies for testing models and publicly disclose them, as well as require that they outline steps they plan to take to mitigate risk. The companies, the executive continued, would 'have to be upfront' about steps taken after test results to make sure models were safe. 'Having this national transparency standard would help not only the public but also Congress understand how the technology is developing, so that lawmakers can decide whether further government action is needed,' he added. Amodei also suggested state laws should follow a similar model that is 'narrowly focused on transparency and not overly prescriptive or burdensome.' Those laws could then be superseded if a national transparency standard is adopted, Amodei said. He noted the issue is not a partisan one, praising steps Trump has taken to support domestic development of AI systems. 'This is not about partisan politics. Politicians on both sides of the aisle have long raised concerns about A.I. and about the risks of abdicating our responsibility to steward it well,' the executive wrote. 'I support what the Trump administration has done to clamp down on the export of A.I. chips to China and to make it easier to build A.I. infrastructure here in the United States.' 'This is about responding in a wise and balanced way to extraordinary times,' he continued. 'Faced with a revolutionary technology of uncertain benefits and risks, our government should be able to ensure we make rapid progress, beat China and build A.I. that is safe and trustworthy. Transparency will serve these shared aspirations, not hinder them.'
Yahoo
an hour ago
- Yahoo
Tech giants' indirect emissions rose 150% in three years as AI expands, UN agency says
By Olivia Le Poidevin GENEVA (Reuters) -Indirect carbon emissions from the operations of four of the leading AI-focused tech companies, Amazon, Microsoft, Alphabet and Meta, rose on average by 150% from 2020-2023, as they had to use more power for energy-demanding data centres, a United Nations report said on Thursday. The use of artificial intelligence is driving up global indirect emissions because of the vast amounts of energy required to power data centres, the report by the International Telecommunication Union (ITU), the U.N. agency for digital technologies, said. Indirect emissions include those generated by purchased electricity, steam, heating and cooling consumed by a company. Amazon's operational carbon emissions grew the most at 182% in 2023 compared to three years before, followed by Microsoft at 155%, Meta at 145% and Alphabet at 138%, according to the report. The ITU tracked the greenhouse gas emissions of 200 leading digital companies between 2020 and 2023. Meta, which owns Facebook and WhatsApp, pointed Reuters to its sustainability report that said it is working to reduce emissions, energy and water used to power its data centres. The other companies did not respond immediately to requests for comment. As investment in AI increases, carbon emissions from the top-emitting AI systems are predicted to reach up to 102.6 million tons of carbon dioxide equivalent (tCO2) per year, the report stated. The data centres that are needed for AI development could also put pressure on existing energy infrastructure. "The rapid growth of artificial intelligence is driving a sharp rise in global electricity demand, with electricity use by data centres increasing four times faster than the overall rise in electricity consumption," the report found. It also highlighted that although a growing number of digital companies had set emissions targets, those ambitions had not yet fully translated into actual reductions of emissions.
