Get ONMINE

Stay Ahead, Stay ONMINE

What’s next for AI in 2025

MIT Technology Review’s What’s Next series looks across industries, trends, and technologies to give you a first look at the future. You can read the rest of them here. For the last couple of years we’ve had a go at predicting what’s coming next in AI. A fool’s game given how fast this industry moves. But we’re on a roll, and we’re doing it again. How did we score last time round? Our four hot trends to watch out for in 2024 included what we called customized chatbots—interactive helper apps powered by multimodal large language models (check: we didn’t know it yet, but we were talking about what everyone now calls agents, the hottest thing in AI right now); generative video (check: few technologies have improved so fast in the last 12 months, with OpenAI and Google DeepMind releasing their flagship video generation models, Sora and Veo, within a week of each other this December); and more general-purpose robots that can do a wider range of tasks (check: the payoffs from large language models continue to trickle down to other parts of the tech industry, and robotics is top of the list).  We also said that AI-generated election disinformation would be everywhere, but here—happily—we got it wrong. There were many things to wring our hands over this year, but political deepfakes were thin on the ground.  So what’s coming in 2025? We’re going to ignore the obvious here: You can bet that agents and smaller, more efficient, language models will continue to shape the industry. Instead, here are five alternative picks from our AI team. 1. Generative virtual playgrounds  If 2023 was the year of generative images and 2024 was the year of generative video—what comes next? If you guessed generative virtual worlds (a.k.a. video games), high fives all round. We got a tiny glimpse of this technology in February, when Google DeepMind revealed a generative model called Genie that could take a still image and turn it into a side-scrolling 2D platform game that players could interact with. In December, the firm revealed Genie 2, a model that can spin a starter image into an entire virtual world. Other companies are building similar tech. In October, the AI startups Decart and Etched revealed an unofficial Minecraft hack in which every frame of the game gets generated on the fly as you play. And World Labs, a startup cofounded by Fei-Fei Li—creator of ImageNet, the vast data set of photos that kick-started the deep-learning boom—is building what it calls large world models, or LWMs. One obvious application is video games. There’s a playful tone to these early experiments, and generative 3D simulations could be used to explore design concepts for new games, turning a sketch into a playable environment on the fly. This could lead to entirely new types of games.  But they could also be used to train robots. World Labs wants to develop so-called spatial intelligence—the ability for machines to interpret and interact with the everyday world. But robotics researchers lack good data about real-world scenarios with which to train such technology. Spinning up countless virtual worlds and dropping virtual robots into them to learn by trial and error could help make up for that.    —Will Douglas Heaven 2. Large language models that “reason” The buzz was justified. When OpenAI revealed o1 in September, it introduced a new paradigm in how large language models work. Two months later, the firm pushed that paradigm forward in almost every way with o3—a model that just might reshape this technology for good. Most models, including OpenAI’s flagship GPT-4, spit out the first response they come up with. Sometimes it’s correct; sometimes it’s not. But the firm’s new models are trained to work through their answers step by step, breaking down tricky problems into a series of simpler ones. When one approach isn’t working, they try another. This technique, known as “reasoning” (yes—we know exactly how loaded that term is), can make this technology more accurate, especially for math, physics, and logic problems. It’s also crucial for agents. In December, Google DeepMind revealed an experimental new web-browsing agent called Mariner. In the middle of a preview demo that the company gave to MIT Technology Review, Mariner seemed to get stuck. Megha Goel, a product manager at the company, had asked the agent to find her a recipe for Christmas cookies that looked like the ones in a photo she’d given it. Mariner found a recipe on the web and started adding the ingredients to Goel’s online grocery basket. Then it stalled; it couldn’t figure out what type of flour to pick. Goel watched as Mariner explained its steps in a chat window: “It says, ‘I will use the browser’s Back button to return to the recipe.’” It was a remarkable moment. Instead of hitting a wall, the agent had broken the task down into separate actions and picked one that might resolve the problem. Figuring out you need to click the Back button may sound basic, but for a mindless bot it’s akin to rocket science. And it worked: Mariner went back to the recipe, confirmed the type of flour, and carried on filling Goel’s basket. Google DeepMind is also building an experimental version of Gemini 2.0, its latest large language model, that uses this step-by-step approach to problem solving, called Gemini 2.0 Flash Thinking. But OpenAI and Google are just the tip of the iceberg. Many companies are building large language models that use similar techniques, making them better at a whole range of tasks, from cooking to coding. Expect a lot more buzz about reasoning (we know, we know) this year. —Will Douglas Heaven 3. It’s boom time for AI in science  One of the most exciting uses for AI is speeding up discovery in the natural sciences. Perhaps the greatest vindication of AI’s potential on this front came last October, when the Royal Swedish Academy of Sciences awarded the Nobel Prize for chemistry to Demis Hassabis and John M. Jumper from Google DeepMind for building the AlphaFold tool, which can solve protein folding, and to David Baker for building tools to help design new proteins. Expect this trend to continue next year, and to see more data sets and models that are aimed specifically at scientific discovery. Proteins were the perfect target for AI, because the field had excellent existing data sets that AI models could be trained on.  The hunt is on to find the next big thing. One potential area is materials science. Meta has released massive data sets and models that could help scientists use AI to discover new materials much faster, and in December, Hugging Face, together with the startup Entalpic, launched LeMaterial, an open-source project that aims to simplify and accelerate materials research. Their first project is a data set that unifies, cleans, and standardizes the most prominent material data sets.  AI model makers are also keen to pitch their generative products as research tools for scientists. OpenAI let scientists test its latest o1 model and see how it might support them in research. The results were encouraging.  Having an AI tool that can operate in a similar way to a scientist is one of the fantasies of the tech sector. In a manifesto published in October last year, Anthropic founder Dario Amodei highlighted science, especially biology, as one of the key areas where powerful AI could help. Amodei speculates that in the future, AI could be not only a method of data analysis but a “virtual biologist who performs all the tasks biologists do.” We’re still a long way away from this scenario. But next year, we might see important steps toward it.  —Melissa Heikkilä 4. AI companies get cozier with national security There is a lot of money to be made by AI companies willing to lend their tools to border surveillance, intelligence gathering, and other national security tasks.  The US military has launched a number of initiatives that show it’s eager to adopt AI, from the Replicator program—which, inspired by the war in Ukraine, promises to spend $1 billion on small drones—to the Artificial Intelligence Rapid Capabilities Cell, a unit bringing AI into everything from battlefield decision-making to logistics. European militaries are under pressure to up their tech investment, triggered by concerns that Donald Trump’s administration will cut spending to Ukraine. Rising tensions between Taiwan and China weigh heavily on the minds of military planners, too.  In 2025, these trends will continue to be a boon for defense-tech companies like Palantir, Anduril, and others, which are now capitalizing on classified military data to train AI models.  The defense industry’s deep pockets will tempt mainstream AI companies into the fold too. OpenAI in December announced it is partnering with Anduril on a program to take down drones, completing a year-long pivot away from its policy of not working with the military. It joins the ranks of Microsoft, Amazon, and Google, which have worked with the Pentagon for years.  Other AI competitors, which are spending billions to train and develop new models, will face more pressure in 2025 to think seriously about revenue. It’s possible that they’ll find enough non-defense customers who will pay handsomely for AI agents that can handle complex tasks, or creative industries willing to spend on image and video generators.  But they’ll also be increasingly tempted to throw their hats in the ring for lucrative Pentagon contracts. Expect to see companies wrestle with whether working on defense projects will be seen as a contradiction to their values. OpenAI’s rationale for changing its stance was that “democracies should continue to take the lead in AI development,” the company wrote, reasoning that lending its models to the military would advance that goal. In 2025, we’ll be watching others follow its lead.  —James O’Donnell 5. Nvidia sees legitimate competition For much of the current AI boom, if you were a tech startup looking to try your hand at making an AI model, Jensen Huang was your man. As CEO of Nvidia, the world’s most valuable corporation, Huang helped the company become the undisputed leader of chips used both to train AI models and to ping a model when anyone uses it, called “inferencing.” A number of forces could change that in 2025. For one, behemoth competitors like Amazon, Broadcom, AMD, and others have been investing heavily in new chips, and there are early indications that these could compete closely with Nvidia’s—particularly for inference, where Nvidia’s lead is less solid.  A growing number of startups are also attacking Nvidia from a different angle. Rather than trying to marginally improve on Nvidia’s designs, startups like Groq are making riskier bets on entirely new chip architectures that, with enough time, promise to provide more efficient or effective training. In 2025 these experiments will still be in their early stages, but it’s possible that a standout competitor will change the assumption that top AI models rely exclusively on Nvidia chips. Underpinning this competition, the geopolitical chip war will continue. That war thus far has relied on two strategies. On one hand, the West seeks to limit exports to China of top chips and the technologies to make them. On the other, efforts like the US CHIPS Act aim to boost domestic production of semiconductors. Donald Trump may escalate those export controls and has promised massive tariffs on any goods imported from China. In 2025, such tariffs would put Taiwan—on which the US relies heavily because of the chip manufacturer TSMC—at the center of the trade wars. That’s because Taiwan has said it will help Chinese firms relocate to the island to help them avoid the proposed tariffs. That could draw further criticism from Trump, who has expressed frustration with US spending to defend Taiwan from China.  It’s unclear how these forces will play out, but it will only further incentivize chipmakers to reduce reliance on Taiwan, which is the entire purpose of the CHIPS Act. As spending from the bill begins to circulate, next year could bring the first evidence of whether it’s materially boosting domestic chip production.  —James O’Donnell

MIT Technology Review’s What’s Next series looks across industries, trends, and technologies to give you a first look at the future. You can read the rest of them here.

For the last couple of years we’ve had a go at predicting what’s coming next in AI. A fool’s game given how fast this industry moves. But we’re on a roll, and we’re doing it again.

How did we score last time round? Our four hot trends to watch out for in 2024 included what we called customized chatbots—interactive helper apps powered by multimodal large language models (check: we didn’t know it yet, but we were talking about what everyone now calls agents, the hottest thing in AI right now); generative video (check: few technologies have improved so fast in the last 12 months, with OpenAI and Google DeepMind releasing their flagship video generation models, Sora and Veo, within a week of each other this December); and more general-purpose robots that can do a wider range of tasks (check: the payoffs from large language models continue to trickle down to other parts of the tech industry, and robotics is top of the list). 

We also said that AI-generated election disinformation would be everywhere, but here—happily—we got it wrong. There were many things to wring our hands over this year, but political deepfakes were thin on the ground. 

So what’s coming in 2025? We’re going to ignore the obvious here: You can bet that agents and smaller, more efficient, language models will continue to shape the industry. Instead, here are five alternative picks from our AI team.

1. Generative virtual playgrounds 

If 2023 was the year of generative images and 2024 was the year of generative video—what comes next? If you guessed generative virtual worlds (a.k.a. video games), high fives all round.

We got a tiny glimpse of this technology in February, when Google DeepMind revealed a generative model called Genie that could take a still image and turn it into a side-scrolling 2D platform game that players could interact with. In December, the firm revealed Genie 2, a model that can spin a starter image into an entire virtual world.

Other companies are building similar tech. In October, the AI startups Decart and Etched revealed an unofficial Minecraft hack in which every frame of the game gets generated on the fly as you play. And World Labs, a startup cofounded by Fei-Fei Li—creator of ImageNet, the vast data set of photos that kick-started the deep-learning boom—is building what it calls large world models, or LWMs.

One obvious application is video games. There’s a playful tone to these early experiments, and generative 3D simulations could be used to explore design concepts for new games, turning a sketch into a playable environment on the fly. This could lead to entirely new types of games. 

But they could also be used to train robots. World Labs wants to develop so-called spatial intelligence—the ability for machines to interpret and interact with the everyday world. But robotics researchers lack good data about real-world scenarios with which to train such technology. Spinning up countless virtual worlds and dropping virtual robots into them to learn by trial and error could help make up for that.   

—Will Douglas Heaven

2. Large language models that “reason”

The buzz was justified. When OpenAI revealed o1 in September, it introduced a new paradigm in how large language models work. Two months later, the firm pushed that paradigm forward in almost every way with o3—a model that just might reshape this technology for good.

Most models, including OpenAI’s flagship GPT-4, spit out the first response they come up with. Sometimes it’s correct; sometimes it’s not. But the firm’s new models are trained to work through their answers step by step, breaking down tricky problems into a series of simpler ones. When one approach isn’t working, they try another. This technique, known as “reasoning” (yes—we know exactly how loaded that term is), can make this technology more accurate, especially for math, physics, and logic problems.

It’s also crucial for agents.

In December, Google DeepMind revealed an experimental new web-browsing agent called Mariner. In the middle of a preview demo that the company gave to MIT Technology Review, Mariner seemed to get stuck. Megha Goel, a product manager at the company, had asked the agent to find her a recipe for Christmas cookies that looked like the ones in a photo she’d given it. Mariner found a recipe on the web and started adding the ingredients to Goel’s online grocery basket.

Then it stalled; it couldn’t figure out what type of flour to pick. Goel watched as Mariner explained its steps in a chat window: “It says, ‘I will use the browser’s Back button to return to the recipe.’”

It was a remarkable moment. Instead of hitting a wall, the agent had broken the task down into separate actions and picked one that might resolve the problem. Figuring out you need to click the Back button may sound basic, but for a mindless bot it’s akin to rocket science. And it worked: Mariner went back to the recipe, confirmed the type of flour, and carried on filling Goel’s basket.

Google DeepMind is also building an experimental version of Gemini 2.0, its latest large language model, that uses this step-by-step approach to problem solving, called Gemini 2.0 Flash Thinking.

But OpenAI and Google are just the tip of the iceberg. Many companies are building large language models that use similar techniques, making them better at a whole range of tasks, from cooking to coding. Expect a lot more buzz about reasoning (we know, we know) this year.

—Will Douglas Heaven

3. It’s boom time for AI in science 

One of the most exciting uses for AI is speeding up discovery in the natural sciences. Perhaps the greatest vindication of AI’s potential on this front came last October, when the Royal Swedish Academy of Sciences awarded the Nobel Prize for chemistry to Demis Hassabis and John M. Jumper from Google DeepMind for building the AlphaFold tool, which can solve protein folding, and to David Baker for building tools to help design new proteins.

Expect this trend to continue next year, and to see more data sets and models that are aimed specifically at scientific discovery. Proteins were the perfect target for AI, because the field had excellent existing data sets that AI models could be trained on. 

The hunt is on to find the next big thing. One potential area is materials science. Meta has released massive data sets and models that could help scientists use AI to discover new materials much faster, and in December, Hugging Face, together with the startup Entalpic, launched LeMaterial, an open-source project that aims to simplify and accelerate materials research. Their first project is a data set that unifies, cleans, and standardizes the most prominent material data sets. 

AI model makers are also keen to pitch their generative products as research tools for scientists. OpenAI let scientists test its latest o1 model and see how it might support them in research. The results were encouraging. 

Having an AI tool that can operate in a similar way to a scientist is one of the fantasies of the tech sector. In a manifesto published in October last year, Anthropic founder Dario Amodei highlighted science, especially biology, as one of the key areas where powerful AI could help. Amodei speculates that in the future, AI could be not only a method of data analysis but a “virtual biologist who performs all the tasks biologists do.” We’re still a long way away from this scenario. But next year, we might see important steps toward it. 

—Melissa Heikkilä

4. AI companies get cozier with national security

There is a lot of money to be made by AI companies willing to lend their tools to border surveillance, intelligence gathering, and other national security tasks. 

The US military has launched a number of initiatives that show it’s eager to adopt AI, from the Replicator program—which, inspired by the war in Ukraine, promises to spend $1 billion on small drones—to the Artificial Intelligence Rapid Capabilities Cell, a unit bringing AI into everything from battlefield decision-making to logistics. European militaries are under pressure to up their tech investment, triggered by concerns that Donald Trump’s administration will cut spending to Ukraine. Rising tensions between Taiwan and China weigh heavily on the minds of military planners, too. 

In 2025, these trends will continue to be a boon for defense-tech companies like Palantir, Anduril, and others, which are now capitalizing on classified military data to train AI models. 

The defense industry’s deep pockets will tempt mainstream AI companies into the fold too. OpenAI in December announced it is partnering with Anduril on a program to take down drones, completing a year-long pivot away from its policy of not working with the military. It joins the ranks of Microsoft, Amazon, and Google, which have worked with the Pentagon for years. 

Other AI competitors, which are spending billions to train and develop new models, will face more pressure in 2025 to think seriously about revenue. It’s possible that they’ll find enough non-defense customers who will pay handsomely for AI agents that can handle complex tasks, or creative industries willing to spend on image and video generators. 

But they’ll also be increasingly tempted to throw their hats in the ring for lucrative Pentagon contracts. Expect to see companies wrestle with whether working on defense projects will be seen as a contradiction to their values. OpenAI’s rationale for changing its stance was that “democracies should continue to take the lead in AI development,” the company wrote, reasoning that lending its models to the military would advance that goal. In 2025, we’ll be watching others follow its lead. 

—James O’Donnell

5. Nvidia sees legitimate competition

For much of the current AI boom, if you were a tech startup looking to try your hand at making an AI model, Jensen Huang was your man. As CEO of Nvidia, the world’s most valuable corporation, Huang helped the company become the undisputed leader of chips used both to train AI models and to ping a model when anyone uses it, called “inferencing.”

A number of forces could change that in 2025. For one, behemoth competitors like Amazon, Broadcom, AMD, and others have been investing heavily in new chips, and there are early indications that these could compete closely with Nvidia’s—particularly for inference, where Nvidia’s lead is less solid. 

A growing number of startups are also attacking Nvidia from a different angle. Rather than trying to marginally improve on Nvidia’s designs, startups like Groq are making riskier bets on entirely new chip architectures that, with enough time, promise to provide more efficient or effective training. In 2025 these experiments will still be in their early stages, but it’s possible that a standout competitor will change the assumption that top AI models rely exclusively on Nvidia chips.

Underpinning this competition, the geopolitical chip war will continue. That war thus far has relied on two strategies. On one hand, the West seeks to limit exports to China of top chips and the technologies to make them. On the other, efforts like the US CHIPS Act aim to boost domestic production of semiconductors.

Donald Trump may escalate those export controls and has promised massive tariffs on any goods imported from China. In 2025, such tariffs would put Taiwan—on which the US relies heavily because of the chip manufacturer TSMC—at the center of the trade wars. That’s because Taiwan has said it will help Chinese firms relocate to the island to help them avoid the proposed tariffs. That could draw further criticism from Trump, who has expressed frustration with US spending to defend Taiwan from China. 

It’s unclear how these forces will play out, but it will only further incentivize chipmakers to reduce reliance on Taiwan, which is the entire purpose of the CHIPS Act. As spending from the bill begins to circulate, next year could bring the first evidence of whether it’s materially boosting domestic chip production. 

—James O’Donnell

Shape
Shape
Stay Ahead

Explore More Insights

More Articles

Stay ahead with more perspectives on cutting-edge power, infrastructure, energy,  bitcoin and AI solutions. Explore these articles to uncover strategies and insights shaping the future of industries.

Shape

Ransomware gangs find a new hostage: Your AWS S3 buckets

To succeed, attackers typically look for S3 buckets that have: versioning disabled ( so old versions can’t be restored), object-lock disabled ( so files can be overwritten or deleted), wide write permissions (via mis-configured IAM policies or leaked credentials), and hold high-value data (backup files, production config dumps). Once inside,

Read More »

Cisco initiative targets device security

Cisco is announcing a security initiative that will push customers to update or replace aging infrastructure components, such as routers, switches and firewalls, as well as discourage them from using any insecure features. Called Resilient Infrastructure, the plan calls for Cisco to strengthen network security by increasing default protections, removing

Read More »

NetOps teams struggle with AI readiness

Some 87% of respondents indicated that internet and cloud environments are creating network blind spots in many areas. Half of organizations reported a lack of adequate insight into public clouds, 44% of respondents indicated transit and peering networks created blind spots, and 43% said remote work environments lack visibility. Other

Read More »

ExxonMobil to Acquire 40 Percent of Bahia NGL Pipeline

Enterprise Products Partners LP will farm out 40 percent of the Bahia natural gas liquids (NGLs) pipeline to Exxon Mobil Corp, in a deal expected to be completed “early 2026” subject to regulatory approvals, Enterprise said Thursday. “The 550-mile Bahia pipeline, which has begun commissioning activities and will begin commercial operations immediately thereafter, will have an initial capacity to transport 600,000 barrels per day (bpd) of NGLs from the Midland and Delaware basins of West Texas to Enterprise’s Mont Belvieu fractionation complex”, the Houston, Texas-based oil and gas midstream company said in a press release. A.J. Teague, co-chief executive of Enterprise’s general partner Enterprise Products Holdings LLC, earlier said in Enterprise’s quarterly report the pipeline was on track to start operations this month. “Upon closing of the transaction, Enterprise and ExxonMobil plan to increase Bahia’s capacity to one million bpd by adding incremental pumping capacity and constructing a 92-mile extension of Bahia to ExxonMobil’s Cowboy natural gas processing plant in Eddy County, New Mexico”, Enterprise added. “The extension will also connect to multiple Enterprise-owned processing facilities in the Delaware Basin. “The expansion and extension are expected to be completed in the fourth quarter of 2027, with ExxonMobil’s interest referred to as the ‘Cowboy Connector’. Enterprise will serve as operator of the combined system”. Teague said, “As the ratio of natural gas and NGL production to crude oil production continues to increase in the Permian, the Bahia pipeline will be an essential artery to deliver mixed NGLs to the fractionation complex in Mont Belvieu. From 2024 to 2030, NGL production in the Permian Basin is expected to increase by over 30 percent”. In the third quarter Enterprise logged NGL pipeline volumes of 4.7 million bpd, up 391,000 bpd from the same three-month period last year. NGL marine terminal volumes averaged 908,000 bpd in July-September 2025, up 21,000 bpd against the third quarter of 2024. NGL fractionation volumes totaled 1.6

Read More »

Oil Prices Have Fallen Sharply

Crude oil prices have fallen sharply, Nadir Belbarka, an analyst at XMArabia, said in a statement sent to Rigzone on Friday, highlighting that Brent was at $62.67 per barrel and WTI was at $58.29 per barrel. “The decline reflects rising expectations of oversupply, fading geopolitical supply risks, and growing coverage of reported progress toward a U.S.-endorsed Russia-Ukraine peace agreement,” Belbarka said in the statement. “Upcoming data – including today’s flash PMIs across the U.S., UK, Germany, and France, along with remarks from ECB President Lagarde – will direct near-term sentiment,” Belbarka added. “Weak readings could heighten recession fears and deepen demand destruction before triggering a technical rebound. Positive surprises could strengthen the dollar and reinforce downward pressure on crude,” the XMArabia analyst continued. Belbarka went on to state that, “in the absence of major inventory drawdowns or a significant supply shock, crude is likely to remain constrained within its new trading range through year-end, awaiting meaningful geopolitical or macroeconomic catalysts”. “Close attention to inventories, IEA [International Energy Agency] and OPEC forecasts, and dollar performance remains essential,” Belbarka warned. In a separate market comment sent to Rigzone on Friday, Eric Chia, Financial Markets Strategist at Exness, noted that crude oil prices “were under pressure today, extending this week’s downside bias as the market digested the potential for geopolitical de-escalation and structural oversupply”. “WTI prices were trading below $58 per barrel, down roughly two percent intraday and set for weekly losses of more than three percent,” Chia added. “The emergence of a Russia-Ukraine peace framework could weigh on the oil market as the prospect of future normalization of Russian crude exports tempered the impact of new U.S. sanctions on Rosneft and Lukoil,” Chia said. “Higher Russian oil exports could also add to the current oversupply narrative. However, a failed deal could help lift

Read More »

USA Crude Oil Stocks Drop by 3.4 Million Barrels WoW

U.S. commercial crude oil inventories, excluding those in the Strategic Petroleum Reserve (SPR) decreased by 3.4 million barrels from the week ending November 7 to the week ending November 14, the U.S. Energy Information Administration (EIA) highlighted in its latest weekly petroleum status report. This EIA report, which was released on November 19 and included data for the week ending November 14, showed that crude oil stocks, not including the SPR, stood at 424.2 million barrels on November 14, 427.6 million barrels on November 7, and 430.3 million barrels on November 15, 2024. Crude oil in the SPR stood at 410.9 million barrels on November 14, 410.4 million barrels on November 7, and 389.2 million barrels on November 15, 2024, the report highlighted. Total petroleum stocks – including crude oil, total motor gasoline, fuel ethanol, kerosene type jet fuel, distillate fuel oil, residual fuel oil, propane/propylene, and other oils – stood at 1.680 billion barrels on November 14, the report revealed. Total petroleum stocks were down 2.2 million barrels week on week and up 47.1 million barrels year on year, the report showed. “At 424.2 million barrels, U.S. crude oil inventories are about five percent below the five year average for this time of year,” the EIA said in its latest weekly petroleum status report. “Total motor gasoline inventories increased by 2.3 million barrels from last week and are about three percent below the five year average for this time of year. Finished gasoline inventories decreased, while blending components inventories increased last week,” the EIA added. “Distillate fuel inventories increased by 0.2 million barrels last week and are about seven percent below the five year average for this time of year. Propane/propylene inventories remained unchanged from last week and are about 16 percent above the five year average for this

Read More »

Norway Gas Output Hits Six-Month High

Norway produced 336.76 million standard cubic meters a day (MMscmd) of natural gas in October, its highest over the last six months, according to preliminary monthly production figures from the country’s upstream regulator. However, last month’s gas output fell 1.7 percent compared to October 2024, though it beat the Norwegian Offshore Directorate’s (NOD) projection by 2.1 percent. Norway sold 10.4 billion scm of gas last month, up 1.9 billion scm from September, the NOD reported on its website. The Nordic country’s oil production in October averaged 1.82 million barrels per day (MMbpd), down 3.6 percent from September but up 2.1 percent from October 2024. The figure exceeded the NOD forecast by 0.4 percent. Total liquids production was 2.02 MMbpd, down 2.8 percent month-on-month but up 0.8 percent year-on-year. “Preliminary production figures for October 2025 show an average daily production of 2,017,000 barrels of oil, NGL and condensate”, the NOD said. “The total petroleum production so far in 2025 is about 197.1 million Sm3 oil equivalents (MSm3 o.e.), broken down as follows: about 87.8 MSm3 o.e. of oil, about 9.7 MSm3 o.e. of NGL and condensate and about 99.5 MSm3 o.e. of gas for sale”, it said. “The total volume is 4.1 MSm3 o.e. less than 2024”. For the third quarter majority state-owned Equinor ASA reported Norwegian equity liquid and gas production of 1.42 million barrels of oil equivalent a day (MMboed), up from 1.36 MMboed in Q2 and 1.31 MMboed in Q3 2024. “In the third quarter of 2025, new fields coming onstream (Johan Castberg and Halten East) drove an increase in production compared to the same quarter last year”, Equinor said of its Norwegian production in its quarterly report October 29. “High production efficiency from Johan Sverdrup, new wells and a lower impact from turnarounds and maintenance more than

Read More »

CEO Denies Alleged TotalEnergies Link to Mozambique Crimes

TotalEnergies SE Chief Executive Officer Patrick Pouyanne rejected accusations the French energy firm has responsibilities in alleged killing of civilians four years ago at its liquefied natural gas project site in Mozambique. The company “is accused of having directly financed and materially supported” a group of armed forces, who “allegedly detained, tortured and killed dozens of civilians” at the LNG project in the north of the country, the European Center for Constitutional and Human Rights said in a statement Tuesday. It filed a criminal complaint over the allegations with the French National Anti-Terrorism Prosecutor this week. “We will defend ourselves and we will explain that all this has nothing to do with TotalEnergies,” Pouyanne said Wednesday on LCI television station. “We’ve done inquiries. We never managed to find evidence” of the allegations.  The complaint comes as Total is on the verge of restarting construction of the project for the first time since the site was shut in 2021 due to an Islamist insurgency. Other global corporations operating in conflict areas have had cases brought against them including Holcim Ltd.’s Lafarge, on trial in France over operations in Syria, and a US ruling against BNP Paribas related to Sudan. The ECCHR complaint, citing an account by Politico, accuses Total of “complicity in war crimes” through a financial link to a Mozambican army unit that allegedly held civilians in shipping containers where dozens of them were tortured and killed at the project between July and September 2021. The company had evacuated the site earlier that year after an attack by insurgents and declared a force majeure. In 2023, Jean-Christophe Rufin, a former French ambassador hired by Total to review the security and humanitarian situation around the project, warned that the developers should stop paying bonuses to Mozambique’s security forces protecting the site.  Total asked government authorities to

Read More »

Powering the grid: embracing EPC for extra-high-voltage growth

Across the country, the demand for power is soaring. Hyperscale facilities, rising industrial load, extreme weather impacts and the loss of firm power capacity are pushing the grid harder than ever. Integration of renewable and distributed generation sources — often far from load centers — has been constrained as infrastructure build-out has lagged soaring demand. The response from the energy sector has been a boom in capital investment, significant new construction and rebuilds of aging infrastructure, aiming to dramatically increase capacity on the grid. The complexity and sheer scale of these projects pose serious risks. A streamlined approach to project delivery, utilizing the engineer-procure-construct (EPC) model, will be key to delivering at the rate the market demands. Accelerating the front end of projects, from concept to mobilization, offers opportunities to optimize through integrated delivery and collaborative contracting. Three important takeaways: Extra-high-voltage (EHV) projects, such as 765-kV transmission lines, are an important part of the sector’s response to modern challenges. Given limited practical experience with such projects, partnerships can better leverage that pool of experience. The portfolio-based approach required to scale extra-high-voltage infrastructure needs EPC delivery for maximum efficiency. The importance of collaboration and coordination is magnified for region-spanning efforts. Bridging Experience Gaps Solving capacity challenges means significant capital investment is essential, particularly in EHV transmission infrastructure. However, most of the limited 765-kV infrastructure in the U.S. was built decades ago. The number of people in today’s workforce who have hands-on experience with design, construction or commissioning at that scale is small and shrinking. The supply of experienced workers — especially field personnel, skilled linemen and engineering leadership — for high-voltage projects is a major constraint in an already-tight labor market. The risk created by that lack of bench strength requires trust among all stakeholders for the projects. Intentional knowledge transfer

Read More »

Nvidia chips sold out? Cut back on AI plans, or look elsewhere

He added, “geopolitical events would be the most likely origin of any type of medium to long term disruption, think China-Taiwan, expansion of the Russia-Ukraine conflict, or escalation in the US-China trade war.” For lower impact events, he said, “[Nvidia] does a nice job of setting conservative shipment goals and targets for Wall Street, which they almost invariably beat quarter after quarter. This provides some cushion for them to absorb a labor, process, or geopolitical hiccup and still meet their stated goals. Shipment volumes may not exceed targets, but shipments would continue to flow; the spice must flow after all.” In a worst-case scenario where shipments are materially impacted, there is little recourse for enterprises that are not large-scale cloud consumers with clout with the limited providers in the space, Bickley added. Enterprises joining a ‘very long queue’ According to Sanchit Vir Gogia, the chief analyst at Greyhound Research, the Nvidia earnings call “confirms that the bottleneck in enterprise AI is no longer imagination or budget. It is capacity. Nvidia reported $57 billion in quarterly revenue, with more than $51 billion from data center customers alone, yet still described itself as supply-constrained at record levels.” Blackwell and Blackwell Ultra, he said, have become the default currency of AI infrastructure, yet even at a build rate of roughly 1,000 GPU racks per week, the company cannot meet demand. Long-term supply and capacity commitments, said Gogia, “now stand at around $50.3 billion, and multi-year cloud service agreements have jumped to $26 billion, implying that much of the next wave of capacity has already been pre-booked by hyperscalers and frontier labs. Enterprises are not stepping into an open market. They are joining the back of a very long queue.”

Read More »

Nvidia is flying high: Is there anything left to say?

Supply chain risks, he said, “are numerous in nature; however, it is clear that Nvidia is customer Number One with all of their suppliers, which drives an inordinate allocation of resources to ensure that production flows. Any disruption would likely be materials-based as opposed to a process or labor issue from their vendor base.” He added, “geopolitical events would be the most likely origin of any type of medium to long term disruption, think China-Taiwan, expansion of the Russia-Ukraine conflict, or escalation in the US-China trade war.” For lower impact events, he said, “[Nvidia] does a nice job of setting conservative shipment goals and targets for Wall Street, which they almost invariably beat quarter after quarter. This provides some cushion for them to absorb a labor, process, or geopolitical hiccup and still meet their stated goals. Shipment volumes may not exceed targets, but shipments would continue to flow; the spice must flow after all.” In a worst-case scenario where shipments are materially impacted, there is little recourse for enterprises that are not large-scale cloud consumers with clout with the limited providers in the space, Bickley added. Enterprises joining a ‘very long queue’ According to Sanchit Vir Gogia, the chief analyst at Greyhound Research, the Nvidia earnings call “confirms that the bottleneck in enterprise AI is no longer imagination or budget. It is capacity. Nvidia reported $57 billion in quarterly revenue, with more than $51 billion from data center customers alone, yet still described itself as supply-constrained at record levels.” Blackwell and Blackwell Ultra, he said, have become the default currency of AI infrastructure, yet even at a build rate of roughly 1,000 GPU racks per week, the company cannot meet demand.

Read More »

Server memory prices could double by 2026 as AI demand strains supply

Limited options for enterprise buyers As supply tightens, most enterprises face limited leverage in selecting suppliers. “Enterprise will have less control over what memory supplier they can choose unless you are a hyperscaler or tier-2 AI datacenter scale enterprise,” Neil Shah, VP for research and partner at Counterpoint Research, told NetworkWorld. “For most enterprises investing in AI infrastructure, they will rely on vendors such as Dell, Lenovo, HPE, Supermicro, and others on their judgment to select the best memory supplier.” Shah advised enterprises with control over their bill of materials to negotiate and lock in supply and costs in advance. “In most cases for long-tail enterprises, smaller buyers without volume leverage, they will have little control as demand outstrips supply, so the prudent thing would be to spread out the rollout over time to average out the cost spikes,” he said. Legacy shortage opens door for Chinese suppliers The current pricing pressure has its roots in production decisions made months ago. According to Counterpoint, the supply crunch originated at the low end of the market as Samsung, SK Hynix, and Micron redirected production toward high-bandwidth memory for AI accelerators, which commands higher margins but consumes three times the wafer capacity of standard DRAM. That shift created an unusual price inversion: DDR4 used in budget devices now trades at approximately $2.10 per gigabit, while server-grade DDR5 sells for around $1.50 per gigabit, according to the firm. This tightness is creating an opportunity for China’s CXMT, noted Shah. “DDR4 is being used in low- to mid-tier smart devices and considering bigger vendors such as Samsung and SK Hynix planned to ramp down DDR4 capacity, CXMT could gain advantage and balance the supply versus demand dynamics moving into the second half of next year,” Shah said.

Read More »

Cobalt 200: Microsoft’s next-gen Arm CPU targets lower TCO for cloud workloads

These architectural improvements underpin Cobalt 200’s claimed increase in performance, which, according to Stephen Sopko, analyst at HyperFRAME Research, will lead to a reduction in total cost of ownership (TCO) compared to its predecessor. As a result, enterprise customers can benefit from consolidating workloads onto fewer machines. “For example, a 1k-instance cluster can see up to 30-40% TCO gains,” Sopko said, adding that this also helps enterprises free up resources to allocate to other workloads or projects. Moor Strategy and Insights principal analyst Matt Kimball noted that the claimed improvements in throughput-per-watt could be beneficial for compute-intensive workloads such as AI inferencing, microservices, and large-scale data processing. Some of Microsoft’s customers are already using Cobalt 100 virtual machines (VMs) for large-scale data processing workloads, and the chips are deployed across 32 Azure data centers, the company said. With Cobalt 200, the company will directly compete with AWS’s Graviton series and Google’s recently announced Axion processors, both of which leverage Arm architecture to deliver better price-performance for cloud workloads. Microsoft and other hyperscalers have been forced to design their own chips for data centers due to the skyrocketing costs for AI and cloud infrastructure, supply constraints around GPUs, and the need for energy-efficient yet customizable architectures to optimize workloads.

Read More »

AWS boosts its long-distance cloud connections with custom DWDM transponder

By controlling the entire hardware stack, AWS can implement comprehensive security measures that would be challenging with third-party solutions, Rehder stated. “This initial long-haul deployment represents just the first implementation of the in-house technology across our extensive long-haul network. We have already extended deployment to Europe, with plans to use the AWS DWDM transponder for all new long-haul connections throughout our global infrastructure,” Rehder wrote. Cloud vendors are some of the largest optical users in the world, though not all develop their own DWDM or other optical systems, according to a variety of papers on the subject. Google develops its own DWDM, for example, but others like Microsoft Azure develop only parts and buy optical gear from third parties. Others such as IBM, Oracle and Alibaba have optical backbones but also utilize third-party equipment. “We are anticipating that the time has come to interconnect all those new AI data centers being built,” wrote Jimmy Yu, vice president at Dell’Oro Group, in a recent optical report. “We are forecasting data center interconnect to grow at twice the rate of the overall market, driven by increased spending from cloud providers. The direct purchases of equipment for DCI will encompass ZR/ZR+ optics for IPoDWDM, optical line systems for transport, and DWDM systems for high-performance, long-distance terrestrial and subsea transmission.”

Read More »

Nvidia’s first exascale system is the 4th fastest supercomputer in the world

The world’s fourth exascale supercomputer has arrived, pitting Nvidia’s proprietary chip technologies against the x86 systems that have dominated supercomputing for decades. For the 66th edition of the TOP500, El Capitan holds steady at No. 1 while JUPITER Booster becomes the fourth exascale system on the list. The JUPITER Booster supercomputer, installed in Germany, uses Nvidia CPUs and GPUs and delivers a peak performance of exactly 1 exaflop, according to the November TOP500 list of supercomputers, released on Monday. The exaflop measurement is considered a major milestone in pushing computing performance to the limits. Today’s computers are typically measured in gigaflops and teraflops—and an exaflop translates to 1 billion gigaflops. Nvidia’s GPUs dominate AI servers installed in data centers as computing shifts to AI. As part of this shift, AI servers with Nvidia’s ARM-based Grace CPUs are emerging as a high-performance alternative to x86 chips. JUPITER is the fourth-fastest supercomputer in the world, behind three systems with x86 chips from AMD and Intel, according to TOP500. The top three supercomputers on the TOP500 list are in the U.S. and owned by the U.S. Department of Energy. The top two supercomputers—the 1.8-exaflop El Capitan at Lawrence Livermore National Laboratory and the 1.35-exaflop Frontier at Oak Ridge National Laboratory—use AMD CPUs and GPUs. The third-ranked 1.01-exaflop Aurora at Argonne National Laboratory uses Intel CPUs and GPUs. Intel scrapped its GPU roadmap after the release of Aurora and is now restructuring operations. The JUPITER Booster, which was assembled by France-based Eviden, has Nvidia’s GH200 superchip, which links two Nvidia Hopper GPUs with CPUs based on ARM designs. The CPU and GPU are connected via Nvidia’s proprietary NVLink interconnect, which is based on InfiniBand and provides bandwidth of up to 900 gigabytes per second. JUPITER first entered the Top500 list at 793 petaflops, but

Read More »

Microsoft will invest $80B in AI data centers in fiscal 2025

And Microsoft isn’t the only one that is ramping up its investments into AI-enabled data centers. Rival cloud service providers are all investing in either upgrading or opening new data centers to capture a larger chunk of business from developers and users of large language models (LLMs).  In a report published in October 2024, Bloomberg Intelligence estimated that demand for generative AI would push Microsoft, AWS, Google, Oracle, Meta, and Apple would between them devote $200 billion to capex in 2025, up from $110 billion in 2023. Microsoft is one of the biggest spenders, followed closely by Google and AWS, Bloomberg Intelligence said. Its estimate of Microsoft’s capital spending on AI, at $62.4 billion for calendar 2025, is lower than Smith’s claim that the company will invest $80 billion in the fiscal year to June 30, 2025. Both figures, though, are way higher than Microsoft’s 2020 capital expenditure of “just” $17.6 billion. The majority of the increased spending is tied to cloud services and the expansion of AI infrastructure needed to provide compute capacity for OpenAI workloads. Separately, last October Amazon CEO Andy Jassy said his company planned total capex spend of $75 billion in 2024 and even more in 2025, with much of it going to AWS, its cloud computing division.

Read More »

John Deere unveils more autonomous farm machines to address skill labor shortage

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More Self-driving tractors might be the path to self-driving cars. John Deere has revealed a new line of autonomous machines and tech across agriculture, construction and commercial landscaping. The Moline, Illinois-based John Deere has been in business for 187 years, yet it’s been a regular as a non-tech company showing off technology at the big tech trade show in Las Vegas and is back at CES 2025 with more autonomous tractors and other vehicles. This is not something we usually cover, but John Deere has a lot of data that is interesting in the big picture of tech. The message from the company is that there aren’t enough skilled farm laborers to do the work that its customers need. It’s been a challenge for most of the last two decades, said Jahmy Hindman, CTO at John Deere, in a briefing. Much of the tech will come this fall and after that. He noted that the average farmer in the U.S. is over 58 and works 12 to 18 hours a day to grow food for us. And he said the American Farm Bureau Federation estimates there are roughly 2.4 million farm jobs that need to be filled annually; and the agricultural work force continues to shrink. (This is my hint to the anti-immigration crowd). John Deere’s autonomous 9RX Tractor. Farmers can oversee it using an app. While each of these industries experiences their own set of challenges, a commonality across all is skilled labor availability. In construction, about 80% percent of contractors struggle to find skilled labor. And in commercial landscaping, 86% of landscaping business owners can’t find labor to fill open positions, he said. “They have to figure out how to do

Read More »

2025 playbook for enterprise AI success, from agents to evals

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More 2025 is poised to be a pivotal year for enterprise AI. The past year has seen rapid innovation, and this year will see the same. This has made it more critical than ever to revisit your AI strategy to stay competitive and create value for your customers. From scaling AI agents to optimizing costs, here are the five critical areas enterprises should prioritize for their AI strategy this year. 1. Agents: the next generation of automation AI agents are no longer theoretical. In 2025, they’re indispensable tools for enterprises looking to streamline operations and enhance customer interactions. Unlike traditional software, agents powered by large language models (LLMs) can make nuanced decisions, navigate complex multi-step tasks, and integrate seamlessly with tools and APIs. At the start of 2024, agents were not ready for prime time, making frustrating mistakes like hallucinating URLs. They started getting better as frontier large language models themselves improved. “Let me put it this way,” said Sam Witteveen, cofounder of Red Dragon, a company that develops agents for companies, and that recently reviewed the 48 agents it built last year. “Interestingly, the ones that we built at the start of the year, a lot of those worked way better at the end of the year just because the models got better.” Witteveen shared this in the video podcast we filmed to discuss these five big trends in detail. Models are getting better and hallucinating less, and they’re also being trained to do agentic tasks. Another feature that the model providers are researching is a way to use the LLM as a judge, and as models get cheaper (something we’ll cover below), companies can use three or more models to

Read More »

OpenAI’s red teaming innovations define new essentials for security leaders in the AI era

Join our daily and weekly newsletters for the latest updates and exclusive content on industry-leading AI coverage. Learn More OpenAI has taken a more aggressive approach to red teaming than its AI competitors, demonstrating its security teams’ advanced capabilities in two areas: multi-step reinforcement and external red teaming. OpenAI recently released two papers that set a new competitive standard for improving the quality, reliability and safety of AI models in these two techniques and more. The first paper, “OpenAI’s Approach to External Red Teaming for AI Models and Systems,” reports that specialized teams outside the company have proven effective in uncovering vulnerabilities that might otherwise have made it into a released model because in-house testing techniques may have missed them. In the second paper, “Diverse and Effective Red Teaming with Auto-Generated Rewards and Multi-Step Reinforcement Learning,” OpenAI introduces an automated framework that relies on iterative reinforcement learning to generate a broad spectrum of novel, wide-ranging attacks. Going all-in on red teaming pays practical, competitive dividends It’s encouraging to see competitive intensity in red teaming growing among AI companies. When Anthropic released its AI red team guidelines in June of last year, it joined AI providers including Google, Microsoft, Nvidia, OpenAI, and even the U.S.’s National Institute of Standards and Technology (NIST), which all had released red teaming frameworks. Investing heavily in red teaming yields tangible benefits for security leaders in any organization. OpenAI’s paper on external red teaming provides a detailed analysis of how the company strives to create specialized external teams that include cybersecurity and subject matter experts. The goal is to see if knowledgeable external teams can defeat models’ security perimeters and find gaps in their security, biases and controls that prompt-based testing couldn’t find. What makes OpenAI’s recent papers noteworthy is how well they define using human-in-the-middle

Read More »
More Articles