You’re receiving this email because you subscribed to the newsletter. If you’d like to unsubscribe, you can manage your preferences here with one click (or use the button at the bottom of the email).

Welcome to this edition of Loop!

To kick off your week, I’ve rounded-up the most important technology and AI updates that you should know about.

‏‏‎ ‎ HIGHLIGHTS ‏‏‎ ‎

Let's jump in!

1. Nvidia continues to print money in the AI gold rush

We start this week with Nvidia, who show no signs of slowing down. The company has just released its third quarter results, smashing Wall Street's expectations with revenues of $57 billion - up 62% from last year.

To put that in perspective, Nvidia's revenues were just $7.1 billion back in Q3 2021. The explosion since then has been fuelled by OpenAI, Anthropic, Microsoft, Meta, and AWS racing to develop new AI models.

But the real story is Nvidia's data centre business, which pulled in a record $51.2 billion this quarter and has grown by $10 billion. That's being driven by the need for more computing power, increasingly sophisticated AI models, and the rise of agentic apps.

Their new Blackwell GPU chips are in incredibly high demand - cloud GPU capacity is completely sold out, with plans to expand even further. To keep up with the growing demand, Nvidia has announced that it will build infrastructure for another 5 million GPUs.

Of course, we do have an AI bubble and lots of tech companies are overvalued. There's a real sense of FOMO that’s driving investors, with some companies seeing their valuations doubling in just a few months.

Some startups will completely fail, despite billions being invested. But Nvidia is one of the few that are actually profitable and delivering real returns.

We've been in an AI gold rush for three years now, and Nvidia's selling the shovels.

2. Archer is developing eVTOLs for the military

Earlier in the year, I wrote that eVTOLs are the perfect aircraft for military operations - they’re silent, have minimal heat signature, and can travel over a good range.

Archer Aviation, one of the leading companies in this sector, has agreed to work with Anduril and will develop autonomous aircraft for the military.

Both companies are working alongside the UAE to test and develop the technology, who have already committed to buying 50 aircraft.

It's the first time Archer has made its core technology available to other companies, which will open up an important revenue stream for the company.

While the technology has been impressive so far, a lot more testing is needed and there are more regulatory hurdles to overcome. That’s incredibly expensive for startups to do, especially when there’s no revenue coming in the door.

With this partnership with Anduril, it will give Archer the financial runway and real-world testing that they desperately need. They can also prove that their tech works in defence scenarios, which will accelerate their own learnings and could make it easier for regulators to approve the aircraft.

It's a smart move on both sides. Archer gets the funding it needs, while Anduril gets access to some of the best eVTOL technology out there - without having to build it from scratch.

3. Europe scales back its privacy and AI laws

Europe is backing down from its tough stance on the technology sector, which saw the EU set standards for the entire globe.

Following pressure from industry and the US government, Brussels will weaken key protections in its GDPR legislation and relax AI rules to boost economic growth.

This new proposal will make it easier for companies to share datasets and allow AI firms to legally train models on personal data, provided they meet the other GDPR requirements.

The changes will also water down the AI Act, which came into force in 2024. High-risk AI systems - which pose serious threats to the health or rights of citizens - were supposed to face stricter rules next summer, but that's now been pushed back indefinitely until the necessary support tools "are available".

On the bright side, you should see fewer cookie pop-ups as they won't be needed for "non-risk" use cases.

Smaller technology firms are also given a boost, with simpler documentation needed for AI systems and a central hub for reporting cybersecurity incidents.

The shift reflects a growing anxiety in Brussels that its regulatory approach has left it on the sidelines of the global AI race.

Europe has virtually no credible players competing with American and Chinese giants like OpenAI, Google, and DeepSeek. That’s partly down to regulation, but it’s largely because startups find it easier to raise huge sums in the US.

Until that changes, Europe will only fall further behind the US and China.

4. Music labels see the writing on the wall, agree to develop AI music tools

The three major music labels are notorious for the lengths they'll go to protect their position. Over the years, we’ve seen them pull their catalogues from MTV, YouTube, Spotify, and TikTok when negotiations don't go their way.

Just last year, they sued two AI music startups - Udio and Suno - for copyright infringement, alleging that the companies had stolen their entire music catalogues to train these AI models.

Fast forward to today and the music labels have completely changed their tune. Universal, Sony, and Warner Music have changed course and are now partnering with these AI startups. So, why the sudden change?

Frankly, these companies have seen the writing on the wall and decided that they need to partner with the technology companies to survive.

The damage has already been done, as their back catalog has already been used to train these advanced AI models, and this is an opportunity for them to make money from AI music in the future.

They can either get onboard with this huge shift in the music industry, or get left behind and allow the tech companies to do it without them. Of course, this is quite disheartening for small artists who hoped that there was another option: sue the AI companies for everything they have.

But that hasn’t worked in other court cases and judges have repeatedly sided with the big tech companies, as they’ve been able to successfully argue that the AI results are different from the training data.

If the music labels continue down that path, there's a real chance they'll lose in court and walk away with nothing. As a result, they’ve decided to get onboard and partner with the leading AI companies instead.

While this move will allow artists to be compensated for some of their music, any existing music has likely been stolen and used to train AI models.

On the same day that Warner’s partnership was signed, Suno announced that it had raised $250 million at a $2.45 billion valuation. Unfortunately for small artists, the direction of travel is pretty clear here and the music labels are now jumping onboard.

5. Google launches Gemini 3 and overtakes OpenAI

The company has come a long way since the rocky launch of their ChatGPT competitor in 2023. When they first announced Bard, Google's share price dropped 8% - wiping out $100 billion in market value.

Things have dramatically changed since then. It’s widely viewed that Anthropic and Google have the best AI models - with Gemini 2.5 Pro being particularly good at coding and general intelligence.

Gemini 3 seems to be a solid improvement over the previous model and shows impressive performance scores, but it’s worth waiting a bit longer as those figures aren’t always relevant for the real world.

The model's native multimodality is perhaps its most significant leap forward. Unlike previous systems that process text, images, and audio separately, Gemini 3 Pro can handle them all simultaneously.

This opens up new use cases, such as converting recipe photos into cookbooks, or analysing a video lecture and using it to create flashcards.

Google is also testing what it calls "generative interfaces". This is something that I flagged earlier in the year, which is often referred to as generative UI.

When you ask ChatGPT to give the weather forecast, it will often respond with long paragraphs of text. Google wants to get rid of this approach entirely.

Instead, you'd see interactive cards that update dynamically – showing current conditions, hourly breakdowns, and forecasts in a format that's actually quick to scan.

This is something that all the top tech companies are working on, as it’s a much more interactive way to see the AI model’s results, but it’s difficult to make them look consistent.

Helsing adds AI capabilities to the Eurofighter jet

Helsing is one of Europe’s most exciting startups and one that I regularly flag in this newsletter. The company has just secured a contract with Saab and will add AI capabilities to Germany’s Eurofighter jet.

The deal is worth “hundreds of millions of Euros” and centres on Helsing's Cirra software, which will be integrated into Saab's sensors over the next three years.

This is important for modern combat, as enemy radars constantly change their patterns and frequencies to avoid being detected or jammed. This makes them extremely difficult for pilots to deal with, but Cirra is designed to solve this problem.

Using deep learning, Cirra can identify unknown enemy radar systems and work out what they're trying to do. This happens in-real time and is done onboard the aircraft, which allows pilots to instantly understand the threats they’re facing - even if they haven’t encountered it before.

The goal is to give European pilots an advantage over Russian forces and make the Eurofighter jets more capable. Eventually, Helsing plans to share this intelligence with other Luftwaffe aircraft and develop swarm systems.

I’ve covered some of Helsing’s other advances, especially on robotics and advanced drones, which you can read here.

Meta can now generate 3D models from any video

Meta has released SAM 3 and SAM 3D, which allow us to do a lot more with computer vision.

With these new foundation models, we can use natural language to identify people or objects within a video. For example, you can ask the model to highlight any “car” it sees within the video.

While this has been possible for years, we could only use computer vision models to place boxes around those objects. We also had to collect our own data and train the model ourselves, which was quite time consuming and costly to do.

The first SAM model was a real breakthrough, as it was able to identify almost any object and companies no longer needed to train their own models - Meta’s model was able to work immediately for most use cases.

Importantly, SAM was able to segment objects properly and draw a precise outline around a car - rather than placing a huge box over it. This opens up a lot more use cases, as you can perfectly edit the image and remove the car.

The second version allowed us to edit videos, which is a huge step forward and version three has made a similar leap.

With SAM 3, we can give the model more detailed prompts than before and ask it to identify a specific object - rather than everything that matches the category. For example, the model can spot “the man wearing a red jumper” rather than everyone who matches the phrase “man”.

It can also generate 3D models of anything in your video - whether it’s a building, plant pot, or even a person skateboarding. It’s remarkable to see this working and there are plenty of use cases.

This could be valuable for Meta’s advertising platform and allow companies to better market their products. Your local furniture shop won’t be able to create 3D scans of their products, like IKEA or Herman Miller can.

But this model could level the playing field and allow them to do that in seconds. If Meta allows advertisers to show a “View in my room” button, customers can better visualise the product and will be more likely to buy it.

💬 OpenAI is launching group chats in ChatGPT

🎨 Google releases a pro version of its "Nano Banana" AI image editor, now free to try

⚡ Porsche announces an electric Cayenne with 1,156 horsepower

🚕 Tesla gets the green light to launch robotaxis in Arizona. Zoox will also begin offering robotaxi rides to public in San Francisco

🛰️ NASA and the European Space Agency have launched a satellite to monitor sea levels

✈️ Google's AI Mode can now help you visualise your travel plans

⚛️ US Energy department shifts priorities and favours fusion power, rather than renewables

🚗 Waymo is expanding to Minneapolis, New Orleans, and Tampa

🎵 TikTok will let you limit the amount of AI-generated content you're shown

💰 Ramp hits $32 billion valuation, just 3 months after hitting $22.5 billion

🚀 Jeff Bezos becomes co-CEO of a new AI startup, called Project Prometheus

😞 Microsoft disappoints with its Windows Copilot (paywall)

This Finnish startup has secured €100 million to develop AI systems for defence, robotics, and other real-world applications.

The funding marks a significant European push into physical AI, which is increasingly dominated by American and Chinese players.

The startup has partnered with Nokia to develop defence-focused AI products, reflecting Europe's growing emphasis on technological sovereignty amid the ongoing Ukraine-Russia war.

NestAI will also support the Finnish Defence Forces in adopting the latest AI advances, which is particularly relevant as Finland shares a long border with Russia.

While it’s still early days for the startup, they have been backed by Peter Sarlin - a Finnish entrepreneur who sold his previous AI startup to AMD for $665 million last year and was an early backer of Lovable.

They’ve also got a lot of experienced people onboard. Much of the founding team previously worked at Intel, while others have joined from defence companies like Kongsberg, Palantir, and Saab.

It's great to see NestAI secure this funding, which reflects a changing mood in Europe and a growing desire to build these defence capabilities within the EU - rather than relying on America to build it instead.

The raise puts NestAI alongside other well-funded defence startups, like Anduril and Helsing, which have raised billions in recent years.

This Week’s Art

We’ve covered quite a bit this week, including:

If you found something interesting in this week’s edition, please feel free to share this newsletter with your colleagues.

Or if you’re interested in chatting with me about the above, simply reply to this email and I’ll get back to you.

Have a good week!

Liam

Share with Others

If you found something interesting in this week’s edition, feel free to share this newsletter with your colleagues.

Liam McCormick is a Senior AI Engineer and works within Kainos' Innovation team. He identifies business value in emerging technologies, implements them, and then shares these insights with others.