title: “Alibaba says it can build AI data centres in 100 days with modular design” url: “https://www.scmp.com/tech/big-tech/article/3363637/alibaba-says-modular-design-delivers-ai-data-centres-100-days-10-lower-cost?module=top_story&pgtype=section”

As demand for artificial intelligence infrastructure surges, Alibaba Group Holding says it can deliver new data centres in a fraction of the usual time while cutting construction costs by 10 per cent through its proprietary modular architecture.

Using CUBE 5.0, Alibaba Cloud had slashed the delivery time for large-scale AI data centres to just 100 days, according to a report by state-backed newspaper China Securities Journal on Tuesday.

That compared with standard domestic delivery times of six to 12 months, and 12 to 18 months in the US, the report said.

Alibaba Cloud, the AI and cloud computing services unit of Alibaba, also expected to more than double its production efficiency this year, measured by the volume of modular data centres delivered within a given time frame, it added.

Unveiled in 2024, CUBE 5.0 increased the overall modularity rate across five key systems – power supply, cooling, security, intelligent management and fire protection – from 30 per cent to 90 per cent in its latest iteration, according to the report.

Unlike traditional construction, where power, cooling and IT systems are installed sequentially on-site, prefabricated modular data centres split infrastructure into separate components that are manufactured simultaneously in factories. These components are then shipped in container-like units and assembled on-site like building blocks.

title: “New US fusion regulations set to spur deployment” url: “https://www.reuters.com/business/energy/new-us-fusion-regulations-set-spur-deployment--reeii-2026-07-22/” author: “Mark Shenk”

• Summary

• A new U.S. regulatory framework for fusion machines is set to establish a clear, predictable regulatory pathway that will likely accelerate licensing.

July 22 - The Nuclear Regulatory Commission (NRC) in February proposed a new framework to regulate fusion machines under the Byproduct Material Framework, underscoring the lower safety risks of fusion devices.

The new rules come after fusion developers complained that stiff regulations were holding up the industry because their devices have long been regulated as ​if they were fission reactors, even though they do not pose meltdown risks and do not generate long-lived nuclear waste.

Instead of requiring exhaustive safety analysis to assess the risks of ‌fusion devices, the proposed regulatory framework focuses on the radioactive materials used by these machines, like tritium, as well as waste management, decommissioning, and radiation protection.

Under the new rules, “licensing happens on the scale of months to a year rather than the years or even a decade a fission reactor can face, because the focus is on safely controlling the material, not reviewing a reactor design,” said Andrew Proffitt, Senior Director of Regulatory Policy at fusion developer Helion Energy.

"The industry supports this rule making. It reflects years of work between ​regulators, developers, and the public, and it provides the certainty companies and investors need to bring commercial fusion online,” Proffitt told Reuters Events.

Join us at Reuters Events Fusion Energy 2026​ to network with 250+ leaders from the global fusion ecosystem.

Unlike existing nuclear facilities that rely on fission—the splitting of heavy ​atoms like uranium—fusion companies are developing first-of-a-kind devices to fuse light atoms together. This is the same reaction that powers the sun, and while it promises virtually ⁠limitless, carbon-free energy, it has never been achieved on a commercial scale.

The regulatory update comes as fusion developers are enjoying a boom in investments that have paved the way for several technological breakthroughs in recent months. Established ​players including Helion Energy, TAE Technologies, Commonwealth Fusion Systems (CFS), and Type One Energy have already started building advanced prototypes.

CHART: Total global fusion industry funding

* Annual totals in brackets. Source: Fusion Industry Association (FIA), 2026. Purchase Licensing Rights, opens new tab

The proposed framework “acknowledges ​that fission and fusion are different beasts,” said Robb Hughes, Head of External Affairs at Realta Fusion.

“Fission reactors pose a risk of meltdown and create long-lived radioactive materials, so fission simply needs more stringent regulation than fusion, for which those same hazards do not exist,” Hughes told Reuters Events.

The new rules come amid a flurry of activity under President Trump aimed at fostering a U.S. “nuclear renaissance.” This includes new NRC policies to expedite the approval of small modular reactor (SMR) designs, streamline environmental reviews, and modernize radiation ​safety standards.

“Fusion has the benefit of an advantageous safety profile compared with fission, creating an opportunity for a lighter regulatory burden and an addressable market for fusion that we believe could eclipse the SMR market,” a ​spokesperson for General Fusion told Reuters Events.

For exclusive nuclear insights, sign up to our newsletter.

The rules align with the ADVANCE Act of 2024, which establishes that fusion machines should be regulated as ‌particle accelerators ⁠rather than traditional nuclear reactors.

The rules do not guarantee faster deployment, but they give “developers something they did not have before: a clear licensing framework,” said Spencer Toohill, Chief of Staff for Nuclear Energy Innovation at The Breakthrough Institute

The new fusion framework is designed to be flexible, allowing developers to “manage design-specific hazards without repeatedly asking the regulator for permission to depart from prescriptive requirements that may not fit their technology,” Toohill told Reuters Events.

Whereas fission developers are often required to seek amendments during the lifecycle of their assets, the “more permissive, materials-oriented framework for fusion reduces the need to ask for so many license amendments,” she said.

While the NRC is legally mandated to finalize commercial fusion ​regulations by December 31, 2027, the agency and the ​Fusion Industry Association (FIA) aim to finalize and adopt ⁠the rules by October 2026.

Download our exclusive report on the latest funding leaps in fusion energy.

State licensing

Under the new framework, state regulators will hold primary responsibility for licensing and overseeing most commercial fusion facilities.

The Organization of Agreement States (OAS)—a longstanding coalition of 40 U.S. states that have signed formal agreements with the NRC—will coordinate the implementation of ​the new rules. 

Under the umbrella of OAS, member states will “engage, communicate, exchange best practices” to ensure that the new regulatory framework is applied in a consistent ​way across state lines, Jeff ⁠Merrifield, the leader of Pillsbury Law’s Nuclear Energy Team, told Reuters Events earlier this year.

Nonetheless, Toohill said she expects some degree of “variation in implementation from state to state,” especially for early demonstrations.

MAP: US fusion companies by primary HQ

Source: Fusion Industry Association (FIA), 2026 Purchase Licensing Rights, opens new tab

Helion has secured regulatory approvals from the Washington State Department of Health, including a Radioactive Materials License (RML) and a Radioactive Air Emissions License (RAEL) for its Orion facility in Malaga, Washington, the company said on June 16.

“After ⁠years of experience ​working with Washington state regulators on earlier prototypes, Helion is the first company in the world to have the regulatory permits ​and licenses in place for a fusion power plant,” Proffitt said.

Other fusion developers, including CFS in Massachusetts, TAE Technologies in California, and Type One Energy in Tennessee are similarly working with their state regulators to secure licenses.

“These are the first of their kind applications that we’re ​going to see over the next few months or year, and they are going to pave the way and set the pace for other states to follow suit,” Merrifield, who also works as outside counsel for FIA, said.

--Editing by Eduardo Garcia

Opinions expressed are those of the author. They do not reflect the views of Reuters News, which, under the Trust Principles, is committed to integrity, independence, and freedom from bias. Reuters Events, a part of Reuters Professional, is owned by Thomson Reuters and operates independently of Reuters News.

[

](https://www.reuters.com/authors/mark-shenk/)

Mark is an energy reporter for Reuters Events, part of Reuters Professional. Based in New York, Mark has two decades of experience in commodities, writing about energy policy, markets, history and consumer impact. His articles have featured in the Washington Post, Boston Globe, Houston Chronicle and Bloomberg Markets, among other publications.

HTML has been gobbling up swathes of what used to be JavaScript’s remit. This page lists a bunch of dynamic functionality that we can now achieve with just HTML.

prospect.org Private Intelligence Firms Are Selling Dossiers on AI and Data Center Critics - The American Prospect Daniel Boguslaw 7 - 9 minutes

Private intelligence firms are selling dossiers about critics of artificial intelligence and the data centers powering them, including trying to peddle them to federal regulators, according to documents obtained by the Prospect. The offerings show that even as voter hostility toward data centers has emerged as an election-altering, bipartisan issue, corporate spy shops are hawking reports that frame widespread dissent in the language of counterterrorism to both private- and public-sector clients.

The drivers of negative sentiment toward data centers and AI include environmental concerns, fears of job loss, and increased utility prices near the sprawling compounds housing server farms and processing facilities. More broadly, people across the ideological spectrum don’t appreciate Big Tech interests dictating their local economic development. Hundreds of grassroots organizations have emerged across the country, seeking to curb data center construction and limit the negative effects of AI.

More from Daniel Boguslaw

But the threat products and reports obtained by the Prospect are less concerned with altering the root cause of this dissent, and more focused on the threat posed by an enraged American populace to corporations’ bottom line.

In a product offering from the corporate intelligence company RANE Network sent to the Federal Energy Regulatory Commission, the corporate firm offered the federal agency—which regulates the interstate transmission and sale of electricity, gas, and oil—a tailored report on the threat posed by anti-tech sentiment.

“RANE can provide an assessment covering rising anti-technology sector sentiment among both the public and governments around the world, identifying specific implications for your organization,” the offering reads. Topics that RANE promises to cover include:

What has brought about an increase in anti-technology sector sentiment?
What role do misinformation and conspiracy theories play in shaping hostility toward the tech sector?
What risks does anti-technology sector sentiment pose to critical supply chains?
What are the potential implications of geopolitical tensions on the operations of technology companies?
How might increasing regulatory scrutiny impact the operations of major tech firms?
What actions can technology companies take to mitigate risks from anti-technology sentiment?

That email was sent in March 2025. This May, RANE offered a webinar on the coming anti-tech backlash. In the hour-long presentation reviewed by the Prospect, RANE cyber and intelligence analysts described the threat posed to technology companies by rising animosity toward Silicon Valley. RANE Network did not respond to multiple requests for comment.

“There’s backlash on a lot of fronts and there are a lot of actors converging on this. So we’ve seen protests from creatives who are worried about AI taking over their copyrighted works,” one analyst told the audience. “There’s also environmental protests. Labor unions are protesting. We’ve seen protests from employees who may have been laid off or are worried about potential layoffs.”

The analysts also warned about protesters concerned about civil liberties and human rights violations arising from AI, and “protests about the impacts on mental health and just general anti-tech or neo-Luddite sentiment that is really driving this activity.”

In 2020, RANE acquired the geopolitical intelligence firm Stratfor. Nine years prior, in 2011, Stratfor was breached by Anonymous, a hacker collective that published millions of internal emails on WikiLeaks. The internal communications detailed the company’s work monitoring the critics of major conglomerates, including Dow Chemical and Coca-Cola.

The specter of hacktivism continues to haunt RANE analysts. “Anti-establishment” and “anti-capitalist groups” alongside Anonymous were referenced during the presentation before a warning that “anti-tech and environmental protest … groups also have a history of some of them going beyond just protest movements to also include, you know, violent extremism,” an analyst said.

In addition to describing the general patterns of concerning groups, presenters also discussed surveilling “online chatter.” The analysts discussed how movements emerge from online spaces, and warned that things can spiral out of control quickly, which is all the more reason to surveil those spaces.

The push to surveil and monitor techno-skeptics comes at the same time that federal agencies have begun circulating their own surveillance memorandums of a new domestic extremist category: “anti-tech extremism.” As Wired reported in May, this new threat category is being used by the FBI and DHS to target constitutionally protected speech and assembly, regardless of whether or not participants have committed, or intend to commit, a crime. In leaked State Department documents obtained by journalist Ken Klippenstein, Secretary of State Marco Rubio informed employees that the agency was monitoring a new alliance of “militant anti-tech and eco-terrorist movements.”

Kroll, the world’s largest investigation and corporate intelligence firm, has also begun selling “risk intelligence” on “public controversy, labor tensions, executive visibility, activism, geopolitical pressure” and “reputational events” to data center operators, according to a June article posted on the firm’s website. Those interested in Kroll’s investigative and monitoring services are encouraged to reach out to its enterprise security risk management team.

Meanwhile, the online surveillance company Liferaft—a subsidiary of the world’s second-largest security firm, Securitas—published a report on July 31 summarizing months of online surveillance of alleged “Threats to AI Infrastructure and Executives.” In the report, Liferaft scanned and analyzed thousands of social media and online forum posts referencing “anger at AI executives,” “opposition to new data centers,” and “grievance over the resources these facilities consume.”

Liferaft’s monitoring tracks the ebb and flow of hostility toward AI companies and executives over the course of several months, but also explicitly warns of a shift in so-called organizational intent, “from personal frustration to coordinated rhetoric … [that] is something threat intelligence professionals pay close attention to, because it tracks with how movements organize.”

The same month Liferaft published its report, hundreds gathered in San Francisco’s downtown to protest AI companies and the existential risks they pose. Anthropic, Google DeepMind, and OpenAI offices were all visited by protesters, including the AI theorist and spiritual leader of the AI-skeptic movement Eliezer Yudkowsky, whose book title “If Anyone Builds It, Everyone Dies” has become a rallying cry for critics of sentient AI. The constitutionally protected protest fits neatly in all three firms’ criteria as a surveillance target.

title: “AI’s recursive self-improvement might not come so quickly after all” url: “https://www.technologyreview.com/2026/08/18/1142188/ai-recursive-self-improvement/” author: “Michelle Kim”

The AI industry’s boldest promise right now is that AI will soon improve itself, with almost no need for human oversight. LLMs can already write code, generate synthetic data for training, and optimize the computer chips they run on. Forecasts of explosive AI progress predict that what researchers call recursive self-improvement is on the horizon. 

But a new study suggests that it might take a while for us to get there. The researchers behind it found that AI agents are not yet capable of conducting open-ended AI research—free-form investigations that have no clear-cut answers and require judgment and taste, which may be integral to building self-improving AI.

A multi-institution group of researchers, led by Peter Kirgis and Sayash Kapoor at Princeton University, found that AI agents could solve the engineering problems necessary to do AI research but lacked the judgment and creativity to produce original research at the caliber of  papers accepted by a top machine-learning conference. The gap suggests that some of the hyped-up timelines for automating AI research may be running ahead of the evidence.

Most existing research on how agents can automate AI research evaluates their ability to complete narrow tasks with checkable answers, such as solving engineering problems or post-training small language models against a benchmark. But making progress in AI research also requires open-ended thinking—choosing a set of hypotheses, deciding what evidence would settle a question, or knowing when to start over. 

To test agents on those kinds of skills, the researchers in the study proposed a new method of evaluation called “shadow evaluation,” which requires the AI to answer a research question from a high-quality unpublished paper. 

The researchers asked Anthropic’s Claude Opus 4.8, running on open-source software called OpenClaw, to tackle such questions, in this case from two papers submitted to the prestigious machine-learning conference NeurIPS 2026. 

The first question was whether a large language model’s “personas,” which determine its behavior, can be controlled by editing the model’s weights (the billions of numbers that store everything it learns during training). The other asked how to design a detector that points out when a model that makes predictions based on spreadsheet data has become unreliable. Because the papers had not been made public, the agents could not memorize the answers from their training data or find them online. 

The agents were given six days, $3,000 in Anthropic API credits, a GPU budget to run the experiments, their own virtual computers, and access to the open web to produce a research paper worthy of publication at a top-tier AI conference. The papers’ original authors graded the agents’ papers as they would evaluate one submitted to a conference.

Those authors rejected both papers. 

The agents were capable of all the engineering required to conduct the research, the human scientists found. The agents reviewed the literature, ran hundreds of experiments, and compiled the results. 

“On the other hand, the agents were unambiguously bad at carrying out the research itself,” says Kapoor. They ran bizarre experiments (in some cases testing their hypotheses on tiny synthetic datasets), struggled to write intelligibly about their work, and made no novel contribution to their fields. “The papers were nowhere close to the mark when it came to being at the quality of a top AI conference,” he says. 

That’s because the agents struggled to muster the creativity and judgment necessary for conducting research. They didn’t do enough to explore different ideas, and they committed to unpromising approaches too quickly. Though the agents developed novel and ambitious hypotheses resembling those that the original authors themselves started with, they rejected them on the basis of very limited data. And they couldn’t backtrack from failing approaches. They could make small pivots but could not fundamentally rethink their approach or try new ones from scratch. 

The agents also failed to incorporate feedback from subagents or external AI reviewing tools. Instead of revising their methodology, the agents narrowed their claims and added caveats. They also couldn’t effectively use resources, such as tokens, compute, and time. And they couldn’t follow instructions about things like how much time to spend on different phases of the research or how long their paper could be.

For all their failures, the agents didn’t engage in the misbehavior that researchers call “reward hacking,” hiding or misrepresenting experiments or data. Although subagents, or helper AIs that the main agent spawns to handle pieces of the work, occasionally hallucinated or misrepresented the results, these were caught by the orchestrator agent, the lead AI supervising the project. 

The reason AI models are good at research engineering but not at open-ended research may come down to how they’re trained, says Kapoor. Models get good at whatever they can be drilled on in a training regime called reinforcement learning, which is easier to apply to tasks whose success can be checked automatically. “But it’s harder to create environments to train these models when the task itself is open-ended,” he says.

Kapoor says the team is now conducting the experiment with Mythos, Anthropic’s most advanced model, which launched in April. It was subsequently required by the Trump administration to meet various safety restrictions and is now available only to approved organizations. Anthropic did not respond to a request for comment.

There are some limitations to the study. It covered just two research papers, and the original authors knew the papers they were grading were generated by AI agents, which could have colored their evaluations. And the researchers had substantial discretion in designing and executing the study, meaning that their preexisting beliefs and biases could have slipped into the results. Evaluations of open-ended research trade some objectivity for a much richer test than any benchmarks can offer.

Still, the results may temper the claims that recursive self-improvement is on the horizon. In June, Anthropic published a blog post titled “When AI Builds Itself,” charting its progress toward models that speed up their own development. In July, OpenAI advertised the fact that its new model GPT-5.6 Sol had helped post-train a smaller model, saving researchers weeks of work.

The new finding may echo what AI companies are finding internally, regardless of their most optimistic public statements. Anthropic cofounder Jack Clark wrote in his newsletter Import AI that it rhymes with what the company found when it tried to automate some aspects of AI safety research. 

“There’s a certain absence of valuable, intuitive creativity in today’s AI systems, and though they’re extraordinarily capable engineers they seem to have a certain property of rote, formulaic thinking that might prevent them [from] being good researchers,” he wrote. He called AI systems’ lack of creativity a “bearish signal on short recursive self-improvement timelines.” 

AI companies do have every incentive to develop AI systems that can rapidly accelerate their own progress, just as they did to make the models better at coding. OpenAI has made building an automated AI researcher an explicit goal, and Anthropic identifies self-improving AI as the industry’s next milestone. 

“If there is investment and then conscious effort toward this direction, I feel like there would be interesting progress, even if it’s failing currently,” says Najoung Kim, a professor of linguistics and computer science at Boston University who researches how AI agents can automate AI research but did not work on the study. On the other hand, it’s possible that AI progress may be bifurcated. AI systems might race ahead on narrow tasks—the kind that can be scored—while advancing slowly on open-ended research. 

The big open question, then, is how crucial open-ended research is to recursive self-improvement—whether AI systems can grind their way there without it, simply by improving on the narrower tasks. “If we look back to the biggest advances in the field, the invention of transformers or the invention of big new architectures that allowed us to make a lot of AI progress—all of those did require creative leaps,” says Kapoor. 

“That said, others have this hypothesis that all of what we need for transformative AI, in particular for recursive self-improvement, is already there.” That would include making a model train faster and boosting its benchmark scores.

“That’s frankly the trillion-dollar question right now,” he says.

title: “Google announces three new cable projects in the Americas” url: “https://www.datacenterdynamics.com/en/news/google-announces-three-new-cable-projects-in-the-americas/” author: “Dan Swinhoe”

Google has announced three new subsea cables in the Americas.

The search giant this week announced three new subsea cable systems - Alisios, Canoa, and OlaLuz – linking south, central, and north America.

The three cables form part of the Americas Connect initiative alongside the previously announced Firmina, Curie, Nuvem, and Sol cables.

The Alisios subsea cable system will connect the Dominican Republic, Panama, and Chile.

Canoa will connect the Dominican Republic to Bermuda, where the Sol and Nuvem cables land.

The OlaLuz cable will connect the Dominican Republic directly to Florida.

Google is also introducing a new branch of the Firmina subsea cable that will land directly in the Dominican Republic.

Capacities, landing points, and project timelines weren’t shared.

“Named after regional maritime traditions, these cables will create highly resilient, diverse routes connecting Chile, Panama, the Dominican Republic, Bermuda, Florida, and beyond,” Bikash Koley, Google VP of global infrastructure, said on LinkedIn. “By creating geographically diverse rings across the Pacific Coast, Caribbean Sea, and Atlantic Ocean, we’re helping bridge the digital divide, connect communities, and drive economic growth in partnership with local governments.”

The new cables will link with Google’s existing subsea infrastructure in the region. Sol and Nuvem link North America and Bermuda to Europe. Firmina and Curie link Brazil and Chile to the US.

The Alisios cable is named after the vientos alisios - the Spanish term for the trade winds that blow across the tropics and the Caribbean, historically used by sailors to navigate between continents. Canoa comes from the Taíno word for canoe, harkening to the longstanding maritime culture in the Caribbean and North Atlantic. The name OlaLuz combines the Spanish words for “wave” (ola) and “light” (luz), referencing the waves of light that carry data along the cable on the ocean floor.

“As we welcome the announcement of the Alisios, OlaLuz, and Canoa subsea cable systems, the Dominican Republic is pleased to continue partnering with Google to bolster digital connectivity and innovation as part of Americas Connect,” said Luis Rodolfo Abinader Corona, president of the Dominican Republic. “Our participation in the Caribbean and Latin America telecommunications network allows us to join our partners in strengthening connectivity across the Americas, serving as an integral node in this expanding digital ecosystem. Our shared vision supports a leap forward in the DR government’s quest to bridge the digital divide, foster local talent, and drive tech-based economic opportunity for our people. This milestone is yet another piece of our collaboration with Google propelling the Dominican Republic into a new era of global digital integration.”

The Dominican Republic currently has three data centers, according to Data Center Map, all located in Santo Domingo and operated by Kio or NAPCaribe (x2). Six cables land there across five landing points.

Google has invested in more than 30 subsea cables in the past 15 years, more than any other hyperscaler, both privately and part of multi-party consortia. The search giant is developing multiple cable landing stations and “connectivity hubs” across a number of locations to support its network build-out.