AI video games are on the way, hints DeepMind CEO

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 ‏‏‎ ‎

• Why Google is betting on fusion power

• The EU’s new plan for quantum computing and the challenges involved

• Samsung’s move to delay its $44 billion chip factory

• … and much more

Let's jump in!

1. Google bets on fusion power, as its carbon emissions grow 50%

We start with Google, who are placing a bold bet on the future of clean energy. The company has recently agreed to purchase 200MW of electricity from an upcoming nuclear fusion plant.

This marks the search giant's first power purchase agreement for fusion energy, although they’ve backed the company behind this project for years.

The timing of this is interesting. Google's latest sustainability report reveals its carbon emissions have surged over 50% since 2019, largely thanks to AI's huge demand for computing power.

Commonwealth Fusion Systems (CFS) will be in charge of the nuclear fusion plant and they believe that it will be connected to Virginia's grid by the early 2030s.

That's remarkably optimistic for a technology that only achieved net energy gain in 2022, so time will tell if they’re right.

It’s worth noting that Google is not the only tech company doing this. Microsoft struck a similar deal with Helion Energy two years ago, with both parties aiming for a 2028 delivery.

Private investors have also invested heavily in the technology. In the last few years, they have poured around $7 billion into fusion startups.

To put Google’s energy demands in context, the company has a 22,000MW clean energy portfolio - which leans heavily on wind and solar.

This 200MW represents a tiny fraction of that, but if fusion actually works at scale then it could transform how we power everything - from data centres to entire cities.

For now, that’s a big “if” and the technology is in a very early stage, but it’s great to see this investment from Google.

2. EU releases its plan for quantum computing, but the same challenges remain

Europe finds itself at a familiar crossroads yet again, as quantum computing emerges as the next big technology and global powers invest heavily in the sector.

Despite leading the world in quantum research publications, the EU captures a mere 5% of global private quantum funding. That’s a stark contrast to America's 50% and China's 40%.

The EU and member states have committed $13 billion to quantum development, but industry leaders worry about the bloc's tendency to spread funding too thin across multiple technologies.

If we compare that to America’s DARPA, which is a government department that encourages defence innovation, they have selected 18 companies for very focused challenges.

There are strong incentives too. Companies that meet DARPA’s milestones could earn up to $300 million.

Quite frankly, Europe has two big problems. Firstly, its venture capital landscape is fragmented and weak compared to the US. This encourages these founders to move to the US and build their startups there instead.

Secondly, European leaders are less likely to take risks. While that’s a very big generalisation, and it might be a little unfair, it’s something I see a lot. Sometimes, that’s a good thing as you can ensure that technology benefits your citizens - not endangers them.

But if you’re too averse to risk, you can allow other people to develop the technology instead and they may not share your values.

It’s time for the EU to create a wider technology strategy that supports founders, lures away top researchers, and invests significantly to close these gaps. If they don’t, the US and China will always remain ahead.

3. Websites can now charge AI companies for scraping their content

Cloudflare’s new marketplace could reshape how websites monetise their content in the age of AI, as traffic to their websites dwindle over time.

The cloud infrastructure giant, which powers 20% of the web, has launched "Pay per Crawl". This will allow publishers to charge AI companies when their website is scraped.

While website scraping has been going for decades, it has accelerated in recent years - with companies rushing to capture all the data they can.

Google scrapes website content and adds it to their search engine, which is fine since publishers are sent lots of online traffic in return.

However, AI companies are destroying that model and the numbers are quite stark. Cloudflare found that while Google's crawler scrapes sites 14 times for every traffic referral, OpenAI scrapes 1,700 times per referral, and Anthropic a staggering 73,000 times.

That's a lot of taking without much giving back and this simply isn’t sustainable for many online business models.

To start charging these AI companies, publishers can simply set their rates and AI companies will decide whether to pay up - with Cloudflare sitting in the middle to handle the transactions.

This could become a very lucrative market for Cloudflare, as they can sit back and handle payments for millions of websites. With a scramble to block AI companies, it should also lead to a jump in their overall market share too.

4. Tesla reports a 14% drop in Q2 vehicle deliveries

Tesla's fortunes continue their downward spiral, with the electric vehicle maker posting a 14% year-on-year decline in Q2 deliveries.

Production held relatively steady at 410,244 units, but that's cold comfort when customers aren't buying.

Whilst Tesla managed to beat some market forecasts, the company’s future looks bleak for the short-to-medium term.

European sales have slumped for five consecutive months, while Chinese competitors - like BYD - are eating into Tesla's market share.

There are several reasons for this. Firstly, Elon Musk has alienated his customer base over the last few years - who previously bought the vehicles to reduce their impact on the environment.

Secondly, China has perfected its supply chain and is now able to build millions of electric cars at low-cost. There are allegations that the state is heavily subsidising these costs, but that only concerns politicians in other countries - not consumers that are facing a rise in the cost of living.

And finally, Tesla’s vehicles no longer have the appeal that they used to. There was a minor design refresh recently, but the cars essentially look the same as they did 10 years ago.

To keep the share price high, Musk constantly promises that his robotaxi investments will pay off. Maybe they will, but I can’t see this happening in the short term - as the technology still faces many hurdles and cannot be safely deployed for every environment.

5. Samsung delays its $44 billion chip factory

Unfortunately, the company can't find enough customers to justify completing the Texas facility.

The South Korean tech giant had originally planned to start semiconductor production in 2024, but that timeline has already slipped to 2026.

According to sources that are familiar with the situation, Samsung is essentially sitting on a factory that’s 91.8% ready - but they don’t have any customers to sell the chips to.

This delay highlights Samsung's broader struggles in the chipmaking space, where it holds just 7.7% market share compared to TSMC's dominant 67.6%.

While TSMC faced its own construction headaches in Arizona, it ultimately secured major AI clients including Nvidia, AMD, Amazon and Google.

Samsung maintains the facility will open in 2026 as planned, but they will need to land significant customers if they want to produce meaningful volumes of chips.

For now, it's an expensive waiting game in the heart of Texas.

DeepMind hints that AI video games are on the way

Demis Hassabis, the CEO of DeepMind, has fuelled speculation that their video model could be used for gaming.

Both Hassabis and Logan Kilpatrick, who leads the product team for Google's AI Studio, have cryptically replied to users asking for the feature and suggested that work is ongoing.

Of course, this isn’t an announcement of any sort - but it does point to a wider trend that I highlighted several months ago.

We are seeing emerging capabilities with AI video models, as they become better at simulating virtual environments and can “remember” what previous frames looked like.

Google's already laying groundwork in this space. Genie 2 is a model that can generate simple games and allow the character to move around.

A few weeks ago, I also covered Odyssey’s early research work into this area and where it could progress further.

Given that Google’s Veo 3 model can create impressive clips and synchronise the audio, I’ve no doubt that they can progress into playable worlds that respond to our commands.

However, the challenge is making those world’s consistent and repeatable. If a gaming studio can’t reproduce that virtual world in a consistent way, then it has much less value for professionals.

US and China are in race to power the future

Both superpowers are locked in a high-stakes energy race, but have very different views on the future. This isn’t something that typically gets a lot of attention, but I think it’s worth exploring as these countries are going in completely different directions.

President Trump’s America has decided to double down on fossil fuels, with GM also scrapping their electric motor plans and will now build V8 engines instead.

The US has also encouraged their allies, like Japan and South Korea, to invest heavily in natural gas projects.

Whereas, China has installed more wind turbines and solar panels in the last year than the rest of the world combined. It’s a huge gulf between the two nations.

Since China doesn’t have significant reserves of fossil fuels, it has poured hundreds of billions into clean tech subsidies. This has led to a manufacturing juggernaut that now controls over 90% of global polysilicon production.

The country is now using its clean energy dominance as soft power. Chinese companies are currently building nuclear plants in Pakistan, wind farms in Kenya, and EV factories in Brazil.

Which strategy will win out? It’s hard to say. There has been growing backlash against environmental targets in the West, as they’re seen as too expensive for the average citizen.

Over the last few years, the UK has been a noticeable uptick in the number of electric cars sold. In 2024, EVs accounted for over 21% of new car sales. But for that to increase further and see more adoption, we need cheaper EVs.

If China ramps up their vehicle production even further and can flood global markets, they could be the nation that wins out. It seems that the US will need to invest significantly more in their own manufacturing base.

🍎 Google releases new AI tools for the classroom

🧠 Meta restructures its AI unit under "Superintelligence Labs"

🔊 Apple considers whether Siri should be powered by Anthropic and OpenAI

🤖 Amazon deploys its 1 millionth robot, releases a GenAI model for robots

💼 Microsoft will lay off 9,000 employees

💕 Lovable on track to raise $150 million at a $2 billion valuation

🚀 Castelion raises $350 million to scale their hypersonic missile business

🛡️ Ilya Sutskever will lead Safe Superintelligence, following CEO's exit

✈️ Joby delivers its first aircraft in Dubai, air taxi service nears launch

💬 Google releases Gemini chatbots for Word Docs, Sheets, and Gmail

🛰️ Planet signs a huge $283 million satellite imagery deal with German government

📱 UK releases a new app for citizens to use government services

Genesis AI

This startup has burst onto the robotics scene with a hefty $105 million seed round, backed by Eclipse and Khosla Ventures.

The startup was founded last year by a robotics researcher at Carnegie Mellon and a former scientist at Mistral AI. With this huge amount in seed funding, they aim to build a universal brain for robots.

Ultimately, this new foundation model would be used to power the robot and allow it to work in different settings - such as research labs.

Most companies are rushing to collect real-world data, as this is needed to train new robotics models. However, the team at Genesis believe that they can create synthetic data as a shortcut.

The company has 20 researchers that are developing a proprietary engine, which will be used to simulate the real-world and quickly collect that synthetic data.

While this is a more difficult way to go about it, the move could give them a clear advantage over their competitors - as many are relying on Nvidia’s tools for updates.

Genesis has offices in both Silicon Valley and Paris, which allows them to hire the top researchers in both continents. Whereas, most startups are only able to choose one location in their early years.

In an ambitious move, they’ve also promised to release their model by the end of 2025. It’s pretty encouraging to see this level of funding in such an early stage, so Genesis has done a great job at attracting attention from investors.

This Week’s Art

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

• Why Google is betting on fusion power

• The EU’s new plan for quantum computing and the challenges involved

• How websites can charge AI companies for scraping their content

• Tesla’s sharp drop in vehicle sales

• Why Samsung has delayed its $44 billion chip factory

• DeepMind’s work on AI video games

• Moves by the US and China to dominate the energy industry

• And how Genesis AI is developing foundation models for robots

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.

About the Author

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.