Google DeepMind’s new AI uses large language models to crack real-world problems

Stay Ahead, Stay ONMINE

Google DeepMind’s new AI uses large language models to crack real-world problems

Google DeepMind has once again used large language models to discover new solutions to long-standing problems in math and computer science. This time the firm has shown that its approach can not only tackle unsolved theoretical puzzles, but improve a range of important real-world processes as well. Google DeepMind’s new tool, called AlphaEvolve, uses the Gemini 2.0 family of large language models (LLMs) to produce code for a wide range of different tasks. LLMs are known to be hit and miss at coding. The twist here is that AlphaEvolve scores each of Gemini’s suggestions, throwing out the bad and tweaking the good, in an iterative process, until it has produced the best algorithm it can. In many cases, the results are more efficient or more accurate than the best existing (human-written) solutions. “You can see it as a sort of super coding agent,” says Pushmeet Kohli, a vice president at Google DeepMind who leads its AI for Science teams. “It doesn’t just propose a piece of code or an edit, it actually produces a result that maybe nobody was aware of.” In particular, AlphaEvolve came up with a way to improve the software Google uses to allocate jobs to its many millions of servers around the world. Google DeepMind claims the company has been using this new software across all of its data centers for more than a year, freeing up 0.7% of Google’s total computing resources. That might not sound like much, but at Google’s scale it’s huge. Jakob Moosbauer, a mathematician at the University of Warwick in the UK, is impressed. He says the way AlphaEvolve searches for algorithms that produce specific solutions—rather than searching for the solutions themselves—makes it especially powerful. “It makes the approach applicable to such a wide range of problems,” he says. “AI is becoming a tool that will be essential in mathematics and computer science.” AlphaEvolve continues a line of work that Google DeepMind has been pursuing for years. Its vision is that AI can help to advance human knowledge across math and science. In 2022, it developed AlphaTensor, a model that found a faster way to solve matrix multiplications—a fundamental problem in computer science—beating a record that had stood for more than 50 years. In 2023, it revealed AlphaDev, which discovered faster ways to perform a number of basic calculations performed by computers trillions of times a day. AlphaTensor and AlphaDev both turn math problems into a kind of game, then search for a winning series of moves. FunSearch, which arrived in late 2023, swapped out game-playing AI and replaced it with LLMs that can generate code. Because LLMs can carry out a range of tasks, FunSearch can take on a wider variety of problems than its predecessors, which were trained to play just one type of game. The tool was used to crack a famous unsolved problem in pure mathematics. AlphaEvolve is the next generation of FunSearch. Instead of coming up with short snippets of code to solve a specific problem, as FunSearch did, it can produce programs that are hundreds of lines long. This makes it applicable to a much wider variety of problems. In theory, AlphaEvolve could be applied to any problem that can be described in code and that has solutions that can be evaluated by a computer. “Algorithms run the world around us, so the impact of that is huge,” says Matej Balog, a researcher at Google DeepMind who leads the algorithm discovery team. Survival of the fittest Here’s how it works: AlphaEvolve can be prompted like any LLM. Give it a description of the problem and any extra hints you want, such as previous solutions, and AlphaEvolve will get Gemini 2.0 Flash (the smallest, fastest version of Google DeepMind’s flagship LLM) to generate multiple blocks of code to solve the problem. It then takes these candidate solutions, runs them to see how accurate or efficient they are, and scores them according to a range of relevant metrics. Does this code produce the correct result? Does it run faster than previous solutions? And so on. AlphaEvolve then takes the best of the current batch of solutions and asks Gemini to improve them. Sometimes AlphaEvolve will throw a previous solution back into the mix to prevent Gemini from hitting a dead end. When it gets stuck, AlphaEvolve can also call on Gemini 2.0 Pro, the most powerful of Google DeepMind’s LLMs. The idea is to generate many solutions with the faster Flash but add solutions from the slower Pro when needed. These rounds of generation, scoring, and regeneration continue until Gemini fails to come up with anything better than what it already has. Number games The team tested AlphaEvolve on a range of different problems. For example, they looked at matrix multiplication again to see how a general-purpose tool like AlphaEvolve compared to the specialized AlphaTensor. Matrices are grids of numbers. Matrix multiplication is a basic computation that underpins many applications, from AI to computer graphics, yet nobody knows the fastest way to do it. “It’s kind of unbelievable that it’s still an open question,” says Balog. The team gave AlphaEvolve a description of the problem and an example of a standard algorithm for solving it. The tool not only produced new algorithms that could calculate 14 different sizes of matrix faster than any existing approach, it also improved on AlphaTensor’s record-beating result for multipying two four-by-four matrices. AlphaEvolve scored 16,000 candidates suggested by Gemini to find the winning solution, but that’s still more efficient than AlphaTensor, says Balog. AlphaTensor’s solution also only worked when a matrix was filled with 0s and 1s. AlphaEvolve solves the problem with other numbers too. “The result on matrix multiplication is very impressive,” says Moosbauer. “This new algorithm has the potential to speed up computations in practice.” Manuel Kauers, a mathematician at Johannes Kepler University in Linz, Austria, agrees: “The improvement for matrices is likely to have practical relevance.” By coincidence, Kauers and a colleague have just used a different computational technique to find some of the speedups AlphaEvolve came up with. The pair posted a paper online reporting their results last week. “It is great to see that we are moving forward with the understanding of matrix multiplication,” says Kauers. “Every technique that helps is a welcome contribution to this effort.” Real-world problems Matrix multiplication was just one breakthrough. In total, Google DeepMind tested AlphaEvolve on more than 50 different types of well-known math puzzles, including problems in Fourier analysis (the math behind data compression, essential to applications such as video streaming), the minimum overlap problem (an open problem in number theory proposed by mathematician Paul Erdős in 1955), and kissing numbers (a problem introduced by Isaac Newton that has applications in materials science, chemistry, and cryptography). AlphaEvolve matched the best existing solutions in 75% of cases and found better solutions in 20% of cases. Google DeepMind then applied AlphaEvolve to a handful of real-world problems. As well as coming up with a more efficient algorithm for managing computational resources across data centers, the tool found a way to reduce the power consumption of Google’s specialized tensor processing unit chips. AlphaEvolve even found a way to speed up the training of Gemini itself, by producing a more efficient algorithm for managing a certain type of computation used in the training process. Google DeepMind plans to continue exploring potential applications of its tool. One limitation is that AlphaEvolve can’t be used for problems with solutions that need to be scored by a person, such as lab experiments that are subject to interpretation. Moosbauer also points out that while AlphaEvolve may produce impressive new results across a wide range of problems, it gives little theoretical insight into how it arrived at those solutions. That’s a drawback when it comes to advancing human understanding. Even so, tools like AlphaEvolve are set to change the way researchers work. “I don’t think we are finished,” says Kohli. “There is much further that we can go in terms of how powerful this type of approach is.”

Google DeepMind’s new tool, called AlphaEvolve, uses the Gemini 2.0 family of large language models (LLMs) to produce code for a wide range of different tasks. LLMs are known to be hit and miss at coding. The twist here is that AlphaEvolve scores each of Gemini’s suggestions, throwing out the bad and tweaking the good, in an iterative process, until it has produced the best algorithm it can. In many cases, the results are more efficient or more accurate than the best existing (human-written) solutions.

“You can see it as a sort of super coding agent,” says Pushmeet Kohli, a vice president at Google DeepMind who leads its AI for Science teams. “It doesn’t just propose a piece of code or an edit, it actually produces a result that maybe nobody was aware of.”

In particular, AlphaEvolve came up with a way to improve the software Google uses to allocate jobs to its many millions of servers around the world. Google DeepMind claims the company has been using this new software across all of its data centers for more than a year, freeing up 0.7% of Google’s total computing resources. That might not sound like much, but at Google’s scale it’s huge.

Jakob Moosbauer, a mathematician at the University of Warwick in the UK, is impressed. He says the way AlphaEvolve searches for algorithms that produce specific solutions—rather than searching for the solutions themselves—makes it especially powerful. “It makes the approach applicable to such a wide range of problems,” he says. “AI is becoming a tool that will be essential in mathematics and computer science.”

AlphaEvolve continues a line of work that Google DeepMind has been pursuing for years. Its vision is that AI can help to advance human knowledge across math and science. In 2022, it developed AlphaTensor, a model that found a faster way to solve matrix multiplications—a fundamental problem in computer science—beating a record that had stood for more than 50 years. In 2023, it revealed AlphaDev, which discovered faster ways to perform a number of basic calculations performed by computers trillions of times a day. AlphaTensor and AlphaDev both turn math problems into a kind of game, then search for a winning series of moves.

FunSearch, which arrived in late 2023, swapped out game-playing AI and replaced it with LLMs that can generate code. Because LLMs can carry out a range of tasks, FunSearch can take on a wider variety of problems than its predecessors, which were trained to play just one type of game. The tool was used to crack a famous unsolved problem in pure mathematics.

AlphaEvolve is the next generation of FunSearch. Instead of coming up with short snippets of code to solve a specific problem, as FunSearch did, it can produce programs that are hundreds of lines long. This makes it applicable to a much wider variety of problems.

In theory, AlphaEvolve could be applied to any problem that can be described in code and that has solutions that can be evaluated by a computer. “Algorithms run the world around us, so the impact of that is huge,” says Matej Balog, a researcher at Google DeepMind who leads the algorithm discovery team.

Survival of the fittest

Here’s how it works: AlphaEvolve can be prompted like any LLM. Give it a description of the problem and any extra hints you want, such as previous solutions, and AlphaEvolve will get Gemini 2.0 Flash (the smallest, fastest version of Google DeepMind’s flagship LLM) to generate multiple blocks of code to solve the problem.

It then takes these candidate solutions, runs them to see how accurate or efficient they are, and scores them according to a range of relevant metrics. Does this code produce the correct result? Does it run faster than previous solutions? And so on.

AlphaEvolve then takes the best of the current batch of solutions and asks Gemini to improve them. Sometimes AlphaEvolve will throw a previous solution back into the mix to prevent Gemini from hitting a dead end.

When it gets stuck, AlphaEvolve can also call on Gemini 2.0 Pro, the most powerful of Google DeepMind’s LLMs. The idea is to generate many solutions with the faster Flash but add solutions from the slower Pro when needed.

These rounds of generation, scoring, and regeneration continue until Gemini fails to come up with anything better than what it already has.

Number games

The team tested AlphaEvolve on a range of different problems. For example, they looked at matrix multiplication again to see how a general-purpose tool like AlphaEvolve compared to the specialized AlphaTensor. Matrices are grids of numbers. Matrix multiplication is a basic computation that underpins many applications, from AI to computer graphics, yet nobody knows the fastest way to do it. “It’s kind of unbelievable that it’s still an open question,” says Balog.

The team gave AlphaEvolve a description of the problem and an example of a standard algorithm for solving it. The tool not only produced new algorithms that could calculate 14 different sizes of matrix faster than any existing approach, it also improved on AlphaTensor’s record-beating result for multipying two four-by-four matrices.

AlphaEvolve scored 16,000 candidates suggested by Gemini to find the winning solution, but that’s still more efficient than AlphaTensor, says Balog. AlphaTensor’s solution also only worked when a matrix was filled with 0s and 1s. AlphaEvolve solves the problem with other numbers too.

“The result on matrix multiplication is very impressive,” says Moosbauer. “This new algorithm has the potential to speed up computations in practice.”

Manuel Kauers, a mathematician at Johannes Kepler University in Linz, Austria, agrees: “The improvement for matrices is likely to have practical relevance.”

By coincidence, Kauers and a colleague have just used a different computational technique to find some of the speedups AlphaEvolve came up with. The pair posted a paper online reporting their results last week.

“It is great to see that we are moving forward with the understanding of matrix multiplication,” says Kauers. “Every technique that helps is a welcome contribution to this effort.”

Real-world problems

Matrix multiplication was just one breakthrough. In total, Google DeepMind tested AlphaEvolve on more than 50 different types of well-known math puzzles, including problems in Fourier analysis (the math behind data compression, essential to applications such as video streaming), the minimum overlap problem (an open problem in number theory proposed by mathematician Paul Erdős in 1955), and kissing numbers (a problem introduced by Isaac Newton that has applications in materials science, chemistry, and cryptography). AlphaEvolve matched the best existing solutions in 75% of cases and found better solutions in 20% of cases.

Google DeepMind then applied AlphaEvolve to a handful of real-world problems. As well as coming up with a more efficient algorithm for managing computational resources across data centers, the tool found a way to reduce the power consumption of Google’s specialized tensor processing unit chips.

AlphaEvolve even found a way to speed up the training of Gemini itself, by producing a more efficient algorithm for managing a certain type of computation used in the training process.

Google DeepMind plans to continue exploring potential applications of its tool. One limitation is that AlphaEvolve can’t be used for problems with solutions that need to be scored by a person, such as lab experiments that are subject to interpretation.

Moosbauer also points out that while AlphaEvolve may produce impressive new results across a wide range of problems, it gives little theoretical insight into how it arrived at those solutions. That’s a drawback when it comes to advancing human understanding.

Even so, tools like AlphaEvolve are set to change the way researchers work. “I don’t think we are finished,” says Kohli. “There is much further that we can go in terms of how powerful this type of approach is.”

Stay Ahead

Explore More Insights

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.

NOAA stops tracking cost of extreme weather and climate disasters

Dive Brief: The National Oceanic and Atmospheric Administration announced Thursday its National Centers for Environmental Information would stop tracking the cost of extreme weather and climate disasters in its Billion Dollar Weather and Climate Disaster database. The product is being retired “in alignment with evolving priorities, statutory mandates, and staffing

Network data hygiene: The critical first step to effective AI agents

Many network teams manage some 15 to 30 different dashboards to track data across all the components in an environment, struggling to cobble together relevant information across domains and spending hours troubleshooting a single incident. In short, they are drowning in data. Artificial intelligence tools—and specifically AI agents—promise to ease

Key takeaways from IBM Think partner event

The first week of May means flowers from April showers and that it’s time for IBM Think in Boston. The first day of the event has historically been the Partner Plus day, which is devoted to content for IBM partners, which include ISVs, technology partners and resellers. The 2025 keynote

LandBridge Posts Higher Revenue

LandBridge Company LLC has reported $44 million in revenue for the first quarter of 2025, up from $36.5 million for the fourth quarter of 2024 and $19 million for the corresponding quarter a year prior. The company attributed the sequential increase to increases in surface use royalties of $6.8 million,

Crude Slips After Inventory Surge

Oil dropped after a government report showed US crude inventories rose the most in two months, overshadowing the lingering relief from the trade truce between the world’s two largest economies. West Texas Intermediate dipped 0.8% to settle near $63 a barrel, snapping a four-session winning streak. Global benchmark Brent closed around $66. US crude stockpiles rose 3.45 million barrels, the biggest gain since March, the Energy Information Administration said Wednesday. “Crude futures are in a corrective phase,” said Dennis Kissler, senior vice president for trading at BOK Financial Securities. Crude’s 50-day moving average of about $63.90 has so far provided a ceiling for gains, he said. The bearish government data stalled a rally spurred by the US-China trade détente and President Donald Trump’s hostile rhetoric on Iranian supply. Trump reiterated Wednesday that Tehran can’t have a nuclear weapon — while Iran’s foreign minister urged the US to take a “more realistic” approach to talks. The friction comes a day after Trump said during a visit to Saudi Arabia that he’ll drive Iran’s oil exports to zero if a nuclear agreement can’t be reached. Washington also announced new sanctions on a shipping network it said was moving billions of dollars’ worth of the Persian Gulf state’s oil. Oil futures have rallied from the lowest closing price in four years at the beginning of last week. They’re still down more than 10% this year, underperforming major equity markets. Trump is visiting the Middle East this week, hoping to strike deals with countries including Saudi Arabia. The kingdom — the de-facto leader of the Organization of the Petroleum Exporting Countries and its allies — has pushed the cartel to increase output to punish noncompliant members. A further boost expected at a June 1 meeting would add to concerns about a surplus. The group’s

Nigeria Says Exxon Commits $1.5B to Deepwater Development

Exxon Mobil Corp. plans to invest $1.5 billion in a deepwater oil field in Nigeria as Africa’s top producer looks to boost output, according to the nation’s regulator. The investment, expected within the next two years, will focus on reviving production in the Usan field and is in addition to funding earmarked for planned developments in Owowo and Erha, the Nigerian Upstream Petroleum Regulatory Commission said in a statement Tuesday, citing a visit by Shane Harris, Exxon’s managing director in Nigeria. The producer aims to reach oil output of 2.4 million barrels a day after dropping to less than half of that level in 2022. Nigeria has taken measures to reduce vandalism and improve regulations, as oil majors have divested from onshore and shallow water fields due to security concerns. Local independent companies are expected to raise output as assets are transferred from recent sales. While Exxon’s investment plans signal its continued interest in the country, the company initially considered spending as much as $10 billion on Nigeria’s offshore oil developments, the government said in September. “This move counters speculation about Exxon Mobil’s potential withdrawal from Nigeria, instead underscoring a strategic expansion and strengthening of its operational footprint in the country,” the NUPRC said in Tuesday’s statement. Exxon could reach a final investment decision on Usan in the third quarter, subject to approvals for its field development plan and by partners, according to NUPRC. WHAT DO YOU THINK? Generated by readers, the comments included herein do not reflect the views and opinions of Rigzone. All comments are subject to editorial review. Off-topic, inappropriate or insulting comments will be removed. MORE FROM THIS AUTHOR Bloomberg

Carney Says He’s Open to Changing Green Laws to Boost Energy Projects

Prime Minister Mark Carney said he’ll consider altering environmental regulations passed by his predecessor, Justin Trudeau, in order to facilitate investment in major projects in Canada, including oil and gas pipelines. “We will change things at the federal level that need to be changed in order for projects to move forward,” Carney told CTV News in an interview that aired Tuesday, shortly after his new cabinet was sworn into office. Asked if those changes might apply to C-69 — legislation that created new rules for environmental assessments — or to a proposed cap on oil and gas industry emissions, he replied: “Absolutely it could include both. But I’m not going to do it conceptually. I’m going to do it on specifics, do it for moving forward.” The two laws have been regularly cited by Pierre Poilievre, leader of the opposition Conservative Party, as impediments to major conventional energy developments. Before entering politics, Carney wrote and spoke at length about the importance of transitioning away from hydrocarbon-based energy to fight climate change. His Liberal government faces a tense relationship with the premier of the oil-rich province of Alberta, where some are even agitating to secede from Canada. The former central banker won Canada’s April 28 election, with his Liberal Party winning the most seats but falling just short of a House of Commons majority. The new energy minister is Tim Hodgson, a former director of Ontario electricity utility Hydro One Ltd. and Canadian oil producer MEG Energy Corp., and previously the CEO of Goldman Sachs Group Inc. in Canada. Carney said his plan to make Canada a superpower “in both clean and conventional energies” involves using domestic oil and gas production to displace imports — particularly from the US — as well as increasing exports. But it also means Canada should play a major role

Amid record demand rise, efficiency and demand response can lower Texas power bills: Aurora

Expanded use of energy efficiency and demand response could help lower Texas power bills while also boosting grid reliability, according to a May 6 paper from Aurora Energy Research. Electrification, data centers and industrial growth are driving energy demand higher across the U.S., and the trend is particularly evident in Texas. In April, the Electric Reliability Council of Texas published a load forecast that predicted nearly 60 GW of new peak summer demand by 2031. Electricity supplied by the grid is expected to rise from 486 TWh this year to 1,038 TWh in 2031. Retrieved from Electric Reliability Council of Texas. “Texas’s economic momentum is driving a sharp rise in electricity demand, creating both a challenge and an opportunity,” Olivier Beaufils, head of USA Central for Aurora, said in a statement. “For the average Texas household, converting to a heat pump instead of resistance heat can decrease the yearly cost of electricity by $424, approximately 45%,” the Aurora report found. “Increased demand response participation can save industrial consumers $1.9M/yr and residential consumers $225 per year.” If 7.7 GW of demand side response capacity, including both residential and industrial consumers, is added to the ERCOT system by 2040, it would reduce the need for new gas-fired generation by 1 GW, the report said. Texas has been trying to build more gas plants through a state backed loan program, but the Texas Energy Fund has seen 35% of its proposed dispatchable capacity canceled or withdrawn amid escalating project costs and supply chain challenges. “Energy efficiency and demand response programs provide critical additional levers for Texas to keep pace with the demand while cutting costs for consumers and strengthening the reliability of the grid,” Beaufils said.

Lawmakers spar over IRA credits in budget hearings

Dive Brief: As Republicans work to advance budget legislation that would make deep cuts to the Inflation Reduction Act, members of the House are debating the issue in committee markup hearings, with Democrats arguing that early phaseouts of IRA credits would harm developers and raise energy bills. The IRA is anticipated to cost taxpayers between $780 billion and $2 trillion over its first ten years, making it an appealing target for Republicans looking to reduce federal spending. However, a large portion of IRA spending has so far benefited Republican districts, and some Republicans have advocated for careful cuts to the legislation – including Rep. Claudia Tenney, R-N.Y., who warned against a wind down of the 45U nuclear production tax credit during the House Ways and Means Committee markup Tuesday. Dive Insight: “Nuclear power represents nearly a quarter of New York’s electrical supply, and my district in Western and Central New York is home to all four of New York’s reactors,” Tenney said. “Those reactors are critical to the baseload power of our region and our state and should be considered separately from our conversations on renewable energy.” In the House Energy and Commerce Committee markup on Tuesday, Rep. Kathy Castor, D-Fla., proposed an amendment to the committee’s budget that would “delay any provisions in this bill from taking effect until the Energy Information Administration studies the impact of the bill and the actions taken so far by the Trump administration.” “People need to know about this,” she said. “You can’t rush a bill like this, with such a large impact, through without people understanding.” Rep. Bob Latta, R-Ohio, opposed the amendment, saying the “best way to lower energy prices is to expand use of America’s abundant energy resources — gas, oil, coal, hydropower — that provide affordable reliable power and fuels, and

IFMA adds its name to Energy Star fight

The International Facility Management Association last week called on Congress to step in and stop the Trump administration from ending Energy Star, the federal government’s program for encouraging the manufacture of energy efficient appliances and systems. “Consumers and businesses alike rely on Energy Star ratings when making informed decisions about appliances and equipment,” Dean Stanberry, immediate past chair of IFMA, said in a statement May 8. “Eliminating this publicly funded resource without consulting the industries and individuals who depend on it is shortsighted and irresponsible.” The administration hasn’t said it wants to eliminate the program, but in its fiscal year 2026 budget request for the Environmental Protection Agency, it proposes to eliminate all funding for the Atmospheric Protection Program, which administers Energy Star. “The Atmospheric Protection Program is an overreach of Government authority that imposes unnecessary and radical climate change regulations on businesses and stifles economic growth,” the administration says in its budget proposal. “By prioritizing climate change over job creation and energy independence, the program has burdened American industries with costly mandates, ultimately hurting consumers and taxpayers. This program is eliminated in the 2026 Budget.” In response to news organizations’ request for clarification on the plan, EPA was noncommittal. “The EPA is delivering organizational improvements to the personnel structure that will directly benefit the American people and better advance the agency’s core mission, while Powering the Great American Comeback,” the agency told CBS News in an email. The Competitive Enterprise Institute, an anti-regulation organization, says the issue isn’t the program’s effectiveness but whether its energy-saving mission is within the proper scope of government. “Green purchasing programs assume the federal government needs to meddle in the marketplace by providing its seal of approval on what it deems to be environmentally satisfactory products,” the organization says in a white paper. “If

HPE ‘morphs’ private cloud portfolio with improved virtualization, storage and data protection

What do you get when combining Morpheus with Aruba? As part of the extensible platform message that HPE is promoting with Morpheus, it’s also working in some capabilities from the broader HPE portfolio. One integration is with HPE Aruba for networking microsegmentation. Bhardwaj noted that a lot of HPE Morpheus users are looking for microsegmentation in order to make sure that the traffic between two virtual machines on a server is secure. “The traditional approach of doing that is on the hypervisor, but that costs cycles on the hypervisor,” Bhardwaj said. “Frankly, the way that’s being delivered today, customers have to pay extra cost on the server.” With the HPE Aruba plugin that now works with HPE Morpheus, the microsegmentation capability can be enabled at the switch level. Bhardwaj said that by doing the microsegmentation in the switch and not the hypervisor, costs can be lowered and performance can be increased. The integration brings additional capabilities, including the ability to support VPN and network address translation (NAT) in an integrated way between the switch and the hypervisor. VMware isn’t the only hypervisor supported by HPE The HPE Morpheus VM Essentials Hypervisor is another new element in the HPE cloud portfolio. The hypervisor is now being integrated into HPE’s private cloud offerings for both data center as well as edge deployments.

AMD targets hosting providers with affordable EPYC 4005 processors

According to Pinkesh Kotecha, chairman and MD of Ishan Technologies, AMD’s 4th Gen EPYC processors stood out because they offer the right combination of high performance, energy efficiency, and security. “Their high core density and ability to optimize performance per watt made them ideal for managing data-intensive operations like real-time analytics and high-frequency transactions. Additionally, AMD’s strong AI roadmap and growing portfolio of AI-optimised solutions position them as a forward-looking partner, ready to support our customers’ evolving AI and data needs. This alignment made AMD a clear choice over alternatives,” Kotecha said. By integrating AMD EPYC processors, Ishan Technologies’ Ishan Cloud plans to empower enterprises across BFSI, ITeS, and manufacturing industries, as well as global capability centers and government organizations, to meet India’s data localization requirements and drive AI-led digital transformation. “The AMD EPYC 4005 series’ price-to-performance ratio makes it an attractive option for cloud hosting and web services, where cost-efficient, always-on performance is essential,” said Manish Rawat, analyst, TechInsights. Prabhu Ram, VP for the industry research group at CMR, said EPYC 4005 processors deliver a compelling mix of performance-per-watt, higher core counts, and modern I/O support, positioning it as a strong alternative to Intel’s Xeon E-2400 and 6300P, particularly for edge deployments. Shah of Counterpoint added, “While ARM-based Ampere Altra promises higher power efficiencies and is ideally adopted in more cloud and hyperscale data centers, though performance is something where x86-based Zen 5 architecture excels and nicely balances the efficiencies with lower TDPs, better software compatibilities supported by a more mature ecosystem.”

Shell’s immersive cooling liquids the first to receive official certification from Intel

Along with the certification, Intel is offering a Xeon processor single-phase immersion warranty rider. This indicates Intel’s confidence in the durability and effectiveness of Shell’s fluids. Yates explained that the rider augments Intel’s standard warranty terms and is available to data center operators deploying 4th and 5th generation Xeon processors in Shell immersion fluids. The rider is intended to provide data center operators confidence that their investment is guaranteed when deployed correctly. Shell’s fluids are available globally and can be employed in retrofitted existing infrastructure or used in new builds. Cuts resource use, increases performance Data centers consume anywhere from 10 to 50 times more energy per square foot than traditional office buildings, and they are projected to drive more than 20% of the growth in electricity demand between now and 2030. Largely due to the explosion of AI, data center energy consumption is expected to double from 415 terawatt-hours in 2024 to around 945 TWh by 2030. There are several other technologies used for data center cooling, including air cooling, cold plate (direct-to-chip), and precision cooling (targeted to specific areas), but the use of immersion cooling has been growing, and is expected to account for 36% of data center thermal management revenue by 2028. With this method, servers and networking equipment are placed in cooling fluids that absorb and dissipate heat generated by the electronic equipment. These specialized fluids are thermally conductive but not electrically conductive (dielectric) thus making them safe for submerging electrical equipment.

Cisco joins AI infrastructure alliance

“The addition of Cisco reinforces AIP’s commitment to an open-architecture platform and fostering a broad ecosystem that supports a diverse range of partners on a non-exclusive basis, all working together to build a new kind of AI infrastructure,” the group said in a statement. Separately, Cisco announced AI initiatives centered in the Middle East region. Last week, Cisco CEO Chuck Robbins visited Saudi Arabia, UAE, Qatar, and Bahrain. This week, Jeetu Patel, executive vice president and chief product officer, is in Saudi Arabia, where he is participating in President Trump’s state visit to the region, according to Cisco. Related new projects include: An initiative with HUMAIN, Saudi Arabia’s new AI enterprise to help build an open, scalable, resilient and cost-efficient AI infrastructure: “This landmark collaboration will set a new standard for how AI infrastructure is designed, secured and delivered – combining Cisco’s global expertise with the Kingdom’s bold AI ambitions. The multi-year initiative aims to position the country as a global leader in digital innovation,” Cisco stated. A collaboration with the UAE-basedG42 to co-develop a secure AI portfolio and AI-native services: Cisco and G42 will work together to assess the potential to co-develop and jointly deploy AI-powered cybersecurity packages, as well as a reference architecture that integrates Cisco’s networking, security, and infrastructure solutions specifically designed for high-performance computing. This collaboration aims to help customers build and secure AI-ready data centers and develop AI workloads effectively, according to the companies. Interest in Qatar’s digital transformation: Qatar’s Ministry of Interior and Cisco signed a letter of intent to collaborate on Qatar’s digital transformation, AI, infrastructure development and cybersecurity.

Cato Networks introduces AI-powered policy analysis engine

Cato Networks this week announced a new policy analysis engine for its cloud-based secure access service edge platform that the company says will optimize and improve SASE policies, reduce risk, simplify compliance, and reduce manual maintenance efforts. Cato Autonomous Policies is built into the Cato SASE Cloud Platform and can provide enterprises with AI-driven recommendations to eliminate security exposure, tighten access controls, and improve network performance. The first use case of the policy engine is designed for firewall as a service (FWaaS) environments in which “firewall rule bloat” is present, Cato explained in a statement. The bloat comes from organizations accumulating thousands of rules that were designed to protect the environment, but after becoming outdated or misconfigured, actually lead to increased risk. “Most enterprises rely on a mix of firewalls deployed in data centers, branch offices, and cloud environments. Over time, rule sets grow, become inconsistent, and are filled with redundant, outdated, or conflicting entries,” wrote Demetris Booth, product marketing director at Cato Networks, in a blog post on the product news. “As a result, security policies become hard to manage, even harder to audit, and often misaligned with zero-trust principles. AI-driven firewall policy management is necessary for modern enterprises to streamline and optimize security operations.”

Riverbed bolsters network acceleration for AI’s performance bottlenecks

“Enterprises are worried about bad actors capturing encrypted traffic and saving copies for when quantum computing advances can break the encryption, providing the bad actors with free access to data. It’s a real concern,” Frey explains. “Post-quantum cryptography is a way to get ahead of that now.” Riverbed also introduced the SteelHead 90 series of network acceleration appliances, which the company says will provide resilient network performance to customers. The series includes: SteelHead 8090, which delivers up to 60 Gbps of data movement over a WAN. It supports multiple 100 Gigabyte network interfaces to pull data from the LAN. SteelHead 6090, which delivers up to 20 Gbps of data movement over a WAN, targeted for mid-scale data centers. SteelHead 4090 and 2090, which support mid-sized data center and edge use cases, with 500 Mbps and 200 Mbps of accelerated traffic, as well as up to 10 Gbps of total traffic processing for quality of service (QoS) and application classification use cases. Riverbed SteelHead Virtual, is a software-only version designed for virtualization environments and private cloud deployments, which is compatible with VMWare ESXI, KVM, and Microsoft Hyper-V. “For customers that are familiar with Riverbed, this is a big change in performance. We’ve gone from moving one appliance at 30 Gbps to 60 Gbps. We want to make sure that whether it’s new AI projects or existing data projects, we have ubiquitous availability across clouds,” says Chalan Aras, senior vice president and general manager of Acceleration at Riverbed. “We’re making it less expensive to move data—we are about half the price of traditional data movement methods.” With this announcement, Riverbed also unveiled its Flex licensing subscription offering. According to Riverbed, Flex makes it possible for enterprises to transfer licenses from hardware to virtual to cloud devices at no cost. Enterprises can reassign

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.

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

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

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

Stay Ahead, Stay ONMINE