POLICY

EU launches AI plan to catch US and China

The EU plans to start the bidding process for AI gigafactory builders in early 2026, it said in a press release.

The move brings Europe’s governing body a step closer to mobilizing the 20 billion euros it has committed to building five AI gigafactories across the bloc. Each gigafactory would house about 100,000 chips, around four times as many as Europe’s current generation of AI factories. They won’t be European chips, however: NVIDIA and other non-EU-manufactured chips will fill the data centers, an EU official told the Wall Street Journal.

Europe currently lags both the US and China in model development and AI adoption among businesses. Given the economic gains from AI, the EU is investing its own funds to jump-start the sector’s growth. 

As part of a memorandum of understanding signed by the European Commission and the European Investment Bank, the EU’s lending arm said it would help provide possible financing for the AI gigafactories. It also said it believes the loans will stimulate additional private investment in European AI and help build infrastructure for European AI startups, research, and industry development.

AI firms are shelling out hundreds of billions of dollars annually for access to data centers filled with advanced computer chips. Data compiled by the Brookings Institute shows that the United States currently leads the data center race by a wide margin, with roughly 10 times more than the next-closest country, Germany.

Europe increasingly feels like an afterthought in an AI race clearly dominated by the US and China. Europe’s most notable model, Mistral Large, is simply not as powerful as the best American or Chinese models. If the scaling wars end in a winner-take-most scenario in which only the best models prove commercially viable, this poses a serious problem for Europe’s AI ambitions. Making matters worse, Europe does not manufacture sufficiently advanced computer chips for AI. So in the long run, these gigafactories may wind up being large boxes where non-European chips are used to train non-European AI. Still, public-sector investment is an encouraging sign for anyone betting that plucky Europeans can stage an AI comeback.

TOGETHER WITH IBM

AI data errors could cost you

How good is your AI? To get the highest business value from your AI, your data must be integrated throughout your organization so it can be seamlessly accessed and used where and when your AI needs it.

Get the most value with AI that leverages:

• Secured data access

• Flexible deployment

• Seamless data integration

Avoid the data-integration mistakes that can cost your organization and discover a right-sized, incremental approach to implementing AI built on an AI-ready data foundation.

CULTURE

Can climate tech startups use AI to fix AI?

Climate tech companies are running against the clock, but AI might be helping them fight back. 

Environmentally focused startups are using AI for everything from saving water to weather forecasting to cutting HVAC emissions — and the tech is giving these firms a massive speed boost in developing systems that push back against the ever-present, multifaceted problems of climate change. 

However, these organizations must balance the speed and capabilities they’re unlocking with the environmental impact AI itself poses: Power demand is projected to skyrocket amid the massive buildout of AI infrastructure, reaching 106 gigawatts by 2035. Without enough renewable energy to go around, the impact of these data centers could be dire. 

But the utility and power of AI models in these use cases may open the door to faster breakthroughs sorely needed, Lisbeth Kaufman, head of climate tech startups at AWS, told The Deep View.

“AI can be very good at making things go faster, and with the climate crisis, speed is so critical,” said Kaufman. “We need to fix this problem as soon as possible.” 

And in some cases, AI may be integral to solving its own problem, said Kaufman. For example, many AI industry leaders are betting on fusion energy as a means of meeting the intense energy demand. It’s something that AWS has targeted in its own startup program, the Compute for Climate Fellowship, which includes three fusion energy startups, Kaufman noted. One of these startups, Thea Energy, uses AI to simplify the complexities of stellarators, or the core hardware involved in fusion, to make it cheaper and easier to scale over time. 

“AI – and the world broadly – needs more renewable and clean energy,” said Kaufman. “This … will be a really powerful way to generate clean energy.”

AI is forcing the tech industry to reckon with the energy crisis. But necessity is the mother of invention. While AI datacenters are spiking energy demand, the need for renewable energy was already present before AI began its rapid ascent. The question is whether AI models can be used to clean up the mess faster than they’re making it. AWS is betting AI technology can be a catalyst for the future of renewable energy.

TOGETHER WITH CODACY

Codacy rolls out new tools to de-risk AI-accelerated coding

Codacy launched two new features to help engineering, security and compliance teams de-risk codebases exposed to GenAI code contributions:

• The AI Risk Hub introduces a new protection layer for insecure AI code through a curated set of enforceable “AI Policies”. It's designed to help teams govern AI-related risks such as unapproved model calls, secrets exposure and dependency vulnerabilities.

• The AI Reviewer pairs deterministic analysis with context-aware reviews to catch real issues legacy scanners miss. It’s designed for an environment where alert fatigue, automation bias, and overloaded review pipelines are causing devs to bypass security and quality checks.

Try Codacy

Use the code DEEPVIEW to get your first month off your Codacy Team subscription. Just start your free trial and apply the promo code at checkout.

STARTUPS

LeCun's 'world models' startup is Meta-free

The AI world has fixed its attention on world models. 

As industry leaders start to question the notion that large language models are the magic key to generalized intelligence, many are turning their attention to world models, or those capable of understanding the world around us the way humans do. The latest to join this movement is AI pioneer Yann LeCun, Meta AI's chief scientist, who is leaving the company to start a firm dedicated to world models. 

Contrary to early reports about LeCun's move, Meta does not appear to be getting a piece of the pie. On Thursday, LeCun told Bloomberg that the social media giant would not invest in his new startup, after having previously described Meta as a “partner.” 

“What's happened in the last few months is that we both realized that the spectrum of applications of this was kind of beyond what Meta was interested in,” LeCun said. 

Despite Meta’s disinterest, AI that can perceive and act on the world around us is increasingly grabbing the attention of tech industry power players, with AI pioneers like Dr. Fei-Fei Li releasing a commercial “spatial intelligence” model and robotics startups nabbing hundreds of millions in funding. 

Physical AI and visual intelligence were also major points of focus at AWS re:Invent in Las Vegas this week, with companies such as RLWRLD, TwelveLabs, Bedrock Robotics and others taking center stage. 

“The currently existing LLMs, or even the multimodal models, are just a brain in a can,” Jung-hee Ryu, CEO of RLWRLD, told The Deep View. “If we decide to build artificial general intelligence or artificial super intelligence, we should give it a body, an embodiment, to feel real-world situations and sensory information.” 

With several AI leaders calling world models and spatial intelligence the next frontier of the market, Meta might be missing the boat by not scoring an investment in LeCun’s startup. The benefits could have been two-fold: For one, Meta’s experience in building the metaverse and creating world models could have played to its strengths. And given that Meta is struggling to keep up in the world of LLMs with the likes of Google, Anthropic, and OpenAI, betting early on a startup led by one of the godfathers of AI could have been a golden ticket. On the other hand, LeCun not being limited to Meta's models is likely a win for a new startup. A Bloomberg report that Meta is planning to cut costs by up to 30% in its metaverse division could also be a factor.

LINKS

• Legal AI firm Harvey raises $160 million at $8 billion valuation

• OpenAI is testing a way to make LLMs fess up to mistakes 

• Wikipedia eyes AI licensing deals to monetize models’ reliance on its content

• Anthropic, AWS partner to bring Claude to Dartmouth University

• AI cloud firm Fluidstack in talks to raise $700 million

• AI search engine extension Phia raises $30 million at $180 million valuation

• Gemini 3 Deep Think: Google’s most advanced reasoning model, capable of exploring multiple hypotheses at a time to give you the best answer.

• Perplexity BrowseSafe: An open-source detection model to prevent malicious prompt injection in real-time. 

• Anything Max: A vibe-coding tool with full-stack control for building production-ready apps. 

• Google Workplace Studio: A new automation tool to design, manage and share AI agents, powered by Gemini 3.

• React-1: A video model by Sync Labs that lets you edit performance, timing and expression in any scene

The Forward Deployed Engineer isn’t another hire — it’s a new way to build.

Forward Deployed Engineers embed where decisions happen, translating business priorities into shipped product faster than traditional engineering teams.

While others debate timelines and chase alignment, smart companies deploy engineers who unblock work, accelerate delivery and reduce iteration cycles.

At Athyna, we provide senior Forward Deployed Engineers who integrate directly into your team:

• Modern tech stack expertise — engineers experienced in Python, TypeScript, React, Node.js, AWS, Kubernetes, Terraform, and more.

• AI-powered matching finds engineers aligned to your stack and use case.

• 4 days from brief to interviews — no hiring drag, no bottlenecks.

• Save up to $80K per hire vs. U.S. talent without compromising seniority or impact.

👉 Meet Your Engineer

(sponsored)

GAMES

POLL RESULTS

Who would you trust most as a partner for an enterprise AI deployment?

• OpenAI (14%)

• Anthropic (23%)

• Google (22%)

• Amazon (12%)

• Microsoft (15%)

• Other (share your thoughts) (14%)