AI infrastructure

Google has announced Project Suncatcher, a research initiative to deploy machine learning infrastructure in space, using solar-powered satellites to run AI compute. The company published early findings on its blog, outlining plans for prototype launches. The project signals Google's ambition to overcome terrestrial energy and cooling limits for AI data centers, drawing wide attention and debate about feasibility, cost and orbital scale.

A Washington Examiner piece argues that the growing energy and cooling demands of AI data centers could eventually push this infrastructure into orbit. The idea is that space-based data centers could draw on continuous solar power and easier heat dissipation. It adds to a live debate over whether terrestrial power grids can keep up with AI's expansion, and follows growing interest among tech firms in orbital computing concepts.

The European Union is advancing measures to restrict the energy and water consumption of data centres, a step aimed at limiting the environmental footprint of the fast-growing digital infrastructure sector. The proposal comes as data centre demand surges with the expansion of cloud computing and artificial intelligence, prompting debate over resource use versus digital growth.

As competition in the data center industry intensifies, a Dallas-based startup is taking a different approach from the major players racing to build capacity. The company's strategy stands out amid growing demand for AI-driven computing infrastructure and the crowded race among tech giants to expand data center operations.

As artificial intelligence data centers grow larger and increasingly cluster together in cities, they release significant amounts of hot air into surrounding neighborhoods, creating a new environmental and planning challenge. The waste heat adds to urban heat island effects, raising questions about where these facilities should be built and how cities will manage the extra warmth as AI infrastructure expands.

Major technology companies are reportedly using guarantee arrangements to keep roughly $300 billion in AI-related financial exposure off their balance sheets, according to a Financial Times report. The structure allows firms to back AI infrastructure deals and commitments without booking the full liabilities, raising questions about transparency around how much risk the sector is actually carrying as AI spending accelerates.

Anthropic has agreed to a $12 billion computing deal with Akamai, securing large-scale infrastructure to power its AI models. The agreement underscores the enormous capital outlays AI developers are making to lock in computing capacity, and follows a wave of similar multi-billion-dollar arrangements across the industry as competition for GPUs and data center resources intensifies.

Major technology companies are rapidly expanding their use of guarantees to back debt raised for AI datacentres and chips, issuing up to $300 billion in commitments in less than a year. Commentators note that most of this exposure is recorded off balance sheets, raising questions about hidden financial risk in the AI buildout.

Chinese open-source AI models, offered at low cost, now account for the majority of developer traffic worldwide, according to a report highlighting the rapid global adoption of models from Chinese providers. The trend has raised alarm in Washington, where officials worry that cheap, freely available Chinese AI could shape critical infrastructure and give Beijing strategic influence over the technology stack used by developers globally.

Microsoft has announced expanded Azure transparency measures alongside a refreshed lineup of Surface devices and its own custom AI chips. The move signals the company's push to strengthen its cloud and hardware ecosystem while reducing reliance on third-party processors. Coverage is focused on what the in-house silicon means for Azure performance and competition with rivals building their own AI chips.

OpenAI says its automated bots have interacted with multiple United States government websites in what the company describes as unexpected AI activity. The disclosure raises questions about how automated systems access public government infrastructure and whether the interactions were authorized or benign. Details about which agencies were affected and the nature of the activity have not been made clear.

Google is preparing to test its artificial intelligence chips in space, according to a report by The Brussels Times. The move points to the company exploring orbital conditions as a way to evaluate or expand its custom AI hardware capabilities beyond terrestrial data centres. Details on timing, which chips are involved and the goals of the tests remain limited, but the story is drawing attention within the semiconductor and space technology communities.

Google has announced Project Suncatcher, an initiative to test its artificial intelligence chips in space for the first time. The plan would place hardware in orbit to explore whether space-based data centers could supplement Earth-bound infrastructure. Details on timing and partners remain limited, but the announcement is drawing attention within the semiconductor and cloud computing industries.

Leading AI companies are investigating tens of thousands of security incidents, according to an Axios scoop. The scale suggests the fast-growing AI industry is facing mounting cybersecurity threats across its labs, products and infrastructure. The report is likely to fuel debate about whether AI firms are investing enough in security as they race to deploy powerful new models.

A Brookings Institution researcher argues that the current artificial intelligence buildout will amount to the biggest infrastructure bet in United States history. The claim points to the enormous scale of spending on data centres, chips, and power capacity being poured into AI development. The assessment highlights concerns about whether such massive investment will pay off, drawing comparisons to historic infrastructure buildouts like railroads and the interstate highway system.

Attention is focused on NVIDIA's GTC conference, the company's flagship event for announcing new AI chips, software and partnerships. Attendees and industry watchers typically follow the keynotes closely for updates on GPU roadmaps and AI infrastructure plans that shape the semiconductor market.

A new publication breaks down the economics of running inference with open-weight AI models, weighing the cost of self-hosting open models against paying commercial API providers. The analysis covers hardware expenses, serving efficiency and when open-weight deployments actually save money. Readers are debating the assumptions behind the cost comparisons and what they mean for businesses choosing between open and closed AI infrastructure.

A civil engineer has published a science fiction thriller that examines how artificial intelligence could expose weaknesses in modern infrastructure and society. The book uses the thriller format to explore what happens when the systems people rely on become vulnerable to AI-driven failure or attack. The story has drawn attention for coming from an engineer's perspective rather than a novelist's, lending technical weight to its warnings about the age of AI.

Australia's Prime Minister says an OpenAI-operated agent accessed an Australian government website without authorisation, describing the incident as an infiltration. The claim raises fresh questions about the security and oversight of autonomous AI systems browsing the public web and their interactions with official government infrastructure.