May 20, 2026 · Money Circuit
THE LARGEST UTILITY MERGER IN US HISTORY JUST HAPPENED AND AI CAUSED IT: NEXTERA BUYS DOMINION FOR $67 BILLION
SOURCE: Fortune – May 18, 2026 | CNBC
Here is a story that starts with utility companies and ends with artificial intelligence taking over the eastern seaboard’s power grid. NextEra Energy announced on May 18 that it is buying Dominion Energy in an all-stock deal worth $67 billion, creating the largest utility company in the world. And if you are wondering what any of this has to do with AI, the answer is: absolutely everything.
Dominion Energy is not just any power company. It is the company that lights up northern Virginia, which happens to be home to the largest concentration of data centers on the planet. When you prompt ChatGPT to write your wedding speech, there is a decent chance the electricity powering that transaction flows through Dominion’s wires. Now NextEra owns all of that. And NextEra is not buying this because it loves power cables.
Here is what is actually happening. The AI industry has a problem most people do not think about. Running the models is extraordinarily expensive not just in compute costs but in raw electricity. A single large language model training run can consume as much power as a small city for weeks. And inference, meaning actually answering your questions, never stops. Data centers need guaranteed electricity around the clock, and they need vastly more of it than exists today.
The hyperscalers, meaning Microsoft, Google, Amazon, and Meta, have been quietly panicking about power availability for the last two years. They have been signing long-term contracts with nuclear plants. They are talking to coal plants about staying open. They are building their own power generation. And now the utility companies are merging with each other to get big enough to serve them.
That is what this deal is. NextEra is already the world’s largest producer of wind and solar energy. Dominion brings nuclear, natural gas, and, most importantly, Virginia. By combining the two, you get a company that can realistically supply the power needs of the next decade of AI infrastructure buildout. The combined entity will be the world leader in renewables and battery storage, the US leader in natural gas generation, and second in nuclear.
The deal is all-stock. NextEra shareholders will own 74.5% of the combined company. Dominion investors will own the rest. Dominion’s stock jumped more than 9% on the announcement. NextEra’s fell more than 4%, which is what happens when you announce you are paying $67 billion for anything.
This is the largest utility acquisition in US history. The largest energy acquisition this century. And nobody is making a big deal of it because there is no product demo, no chatbot, and Jensen Huang was not on stage. But the AI companies themselves will depend on this deal more than they depend on most of the headline-grabbing announcements you have read about this week. You can have all the models in the world. Without power, they are expensive paperweights. The electric bill for the AI revolution just arrived, and it came in at $67 billion.
BLACKSTONE AND GOOGLE BUILD A $5 BILLION AI CLOUD COMPANY TO DETHRONE NVIDIA’S STRANGLEHOLD ON COMPUTE
SOURCE: CNBC – May 19, 2026 | Bloomberg
For two years now, CoreWeave has been the name in AI cloud computing. You want raw GPU power? You rent from CoreWeave. Every AI lab, every startup, every enterprise trying to run inference at scale has been knocking on CoreWeave’s door. And CoreWeave has been getting very rich. Google and Blackstone have apparently noticed this and decided they would like to be CoreWeave, but bigger.
On May 19, Blackstone and Google announced they are jointly creating a new AI infrastructure company. Blackstone is putting in $5 billion in equity, which, when leveraged, gives the venture $25 billion in total firepower. The goal is to build 500 megawatts of compute capacity by 2027 and then keep growing from there. The thing that makes this interesting is what powers it: Google’s own TPU chips, not Nvidia’s GPUs.
Every major AI compute company today runs on Nvidia hardware. CoreWeave runs on Nvidia. Oracle’s AI cloud runs on Nvidia. Microsoft’s Azure AI runs on Nvidia. The entire ecosystem is, somewhat uncomfortably for everyone involved, dependent on a single chip designer out of Santa Clara. Google has been developing its Tensor Processing Units for years as an alternative, and they are genuinely excellent at inference workloads. But nobody has built a large-scale cloud infrastructure business around them. Until now.
If Blackstone and Google can make this work, they create the first serious alternative to the Nvidia-dominated cloud compute market. They would also give every AI lab a second option when negotiating compute contracts, which means the whole industry’s cost structure gets more competitive. Blackstone brings the money and the real estate relationships. It manages more than $1.3 trillion in assets and already owns data center infrastructure around the world. Google brings the chips, the software, and the engineering talent.
The broader context matters here. Blackstone earlier in May announced a separate joint venture with Anthropic to deploy AI in private equity portfolio companies. So in the space of about two weeks, Blackstone has established itself as one of the most aggressive institutional investors in the entire AI ecosystem. They are not making small bets. They are writing $5 billion checks and calling it a Tuesday.
For Google specifically, this is a strategic necessity. The company has been spending tens of billions of dollars building AI infrastructure, training Gemini, and competing with OpenAI. But it has never quite cracked the third-party cloud market the way AWS and Azure have. Partnering with Blackstone gives Google a credible path to monetizing its chip investment with customers who are not Google. The first 500 megawatts come online in 2027. The AI compute market is about to get its biggest institutional player yet, and it arrives with Google’s technology and Blackstone’s money. That is a combination that does not lose very often.
GOLDMAN SACHS, BLACKSTONE, AND ANTHROPIC FORM A $1.5 BILLION VENTURE TO PUT AI INSIDE EVERY PE-BACKED COMPANY ON EARTH
SOURCE: CNBC – May 4, 2026 | Fortune
If you own a medium-sized company and a private equity firm owns a chunk of it, there is now a very well-funded operation whose entire business model is to walk through your front door and restructure everything around artificial intelligence. Anthropic teamed up with Goldman Sachs, Blackstone, and Hellman and Friedman to launch a $1.5 billion joint venture, and the target is every portfolio company in the PE universe.
Goldman, Blackstone, and Hellman and Friedman each contributed roughly $300 million. Goldman Sachs kicked in $150 million. Apollo Global, General Atlantic, Leonard Green, Singapore’s sovereign wealth fund GIC, and Sequoia Capital also joined. The new firm will embed engineers and AI consultants inside mid-sized businesses to redesign their workflows around Claude.
This is Anthropic’s most aggressive enterprise move to date. They have been known as the safety-focused, thoughtful alternative to OpenAI. They have a chatbot people genuinely like. But this venture is something different. This is Anthropic saying: we do not just want to sell API access. We want to be inside your company, restructuring how you do business.
The venture targets a very specific gap. Every large PE firm in the world has a portfolio of companies in healthcare, manufacturing, financial services, retail, and real estate. These are not tech companies. They do not have AI teams. They have operations managers and accountants and people who have been doing their jobs the same way for twenty years. The joint venture will provide the engineers to go in, assess the workflows, and rebuild them around AI agents.
Goldman and its partners plan to use their own portfolio companies as the initial proving ground before expanding to other mid-sized businesses. That means the first wave of clients is essentially a captive audience. Goldman-backed company? Here come the Anthropic engineers. You are not being asked if you want this.
What Anthropic gains is distribution at industrial scale. Instead of waiting for companies to discover Claude’s API, they now have Goldman Sachs walking through the front door of every business they own and saying: you are going to use this. That is a go-to-market strategy that no amount of marketing spend could replicate. The foundational model race is largely settled for now. The winners are known. The real money in 2026 is in deployment. Goldman’s answer involves $1.5 billion and the entirety of private equity’s existing portfolio. When Goldman Sachs puts together a $1.5 billion answer to anything, it tends to be worth paying attention to.
DEEPSEEK WANTS YOUR MONEY NOW: CHINA’S STATE CHIP FUND BACKS THE LAB THAT SHOOK SILICON VALLEY AT A $45 BILLION VALUATION
SOURCE: TechCrunch – May 6, 2026 | Bloomberg
About a year ago, DeepSeek dropped its R1 model and the entire American AI investment world had a brief existential crisis. How could a Chinese lab produce a model that competitive, for that little money? Nvidia’s stock lost $600 billion in a single day. Venture capitalists spent a week updating their theses. And then everything went more or less back to normal because that is what markets do. But now something genuinely different is happening.
DeepSeek, which has never taken a dollar of outside investment, is about to raise its first funding round. And the valuation being discussed is $45 billion. In three weeks, that number went from $10 billion to $20 billion to $45 billion. At this rate, by the time you read this it might be $60 billion. The target raise is up to $7.35 billion, which is a lot of money for a lab that makes its models open-source and does not charge for API access the way its Western competitors do.
The lead investor in talks is the China Integrated Circuit Industry Investment Fund, which is basically China’s state-backed chip investment vehicle. Tencent and Alibaba are also reportedly in discussions to participate. So what you have is a company that shocked the AI world with its efficiency and open-source approach now potentially receiving enormous sums from Chinese state and corporate investors.
Here is why the founder, a hedge fund billionaire named Liang Wenfeng who controls nearly 90% of the company, decided to do this now. Competitors have been poaching DeepSeek’s researchers. You build a world-class AI lab, you attract world-class talent, and then every other lab in China starts showing up with bags of money. The only way to keep your people is to give them equity. And to give them equity, you need a company valuation and outside investors. The fundraising is driven as much by retention as by ambition.
The business model question is genuinely interesting. DeepSeek runs an API service and a consumer app, but it is not a venture-scale revenue machine yet. This round mostly signals something beyond business logic. It injects state money into a company that has become a geopolitical asset for China. The US has export controls on high-end Nvidia chips going to China. DeepSeek’s claim to fame is building competitive models on less compute. Now China’s chip investment fund is backing them. The strategic logic writes itself.
For the American AI industry, this is not the story that breaks anything overnight. But it is a signal that the AI race is not just between OpenAI, Anthropic, and Google anymore. The Chinese government is now a venture capitalist. Its portfolio company might be the most efficient AI lab on earth. And they just decided to put real money behind it.
CHINA’S MOONSHOT AI RAISES $2 BILLION AND THE LEAD CHECK CAME FROM A FOOD DELIVERY COMPANY
SOURCE: TechCrunch – May 7, 2026 | Bloomberg
Same week, different Chinese AI lab. Moonshot AI, the company behind the Kimi chatbot and the K2.5 model that Cursor quietly admitted to building on top of, just closed a $2 billion funding round at a $20 billion valuation. The lead investor was Meituan’s venture arm. Meituan is the company that delivers food to your doorstep in Chinese cities. Your kung pao chicken delivery app is now backing frontier AI research. This is the world we live in and it is completely rational once you understand what is happening.
Moonshot AI was founded in 2023 by Yang Zhilin, a former Meta AI and Google Brain researcher. In less than three years it has become one of China’s most popular AI labs, hit over $200 million in annual recurring revenue, and built a model that an American AI startup worth $50 billion used as the foundation for its own new product. That startup is Cursor, which recently admitted its latest coding model was built on top of Kimi K2.5. The irony of Silicon Valley’s hottest developer tool relying on a Chinese lab’s backbone is the kind of thing nobody in Washington wants to think about too hard.
The K2.5 model is specifically known for long context and coding capability. It can process enormous amounts of text and is particularly good at understanding large codebases. That is the feature that made it attractive to Cursor, and it is also what makes Moonshot genuinely interesting as a research organization rather than just another chatbot company.
Moonshot’s backers now include Alibaba, Tencent, HongShan (formerly Sequoia China), IDG Capital, and several other heavyweights. The $2 billion round pushes the valuation from around $18 billion in March to $20 billion in May. That is roughly $2 billion in new valuation per month, a pace that would make most Western startups feel like underachievers.
Meituan processes hundreds of millions of food delivery transactions. It has supply chain logistics, consumer behavior data at massive scale, and a business model that touches everyday life in China. Why would a food delivery company back an AI lab? For the same reason every major company in the world is backing AI labs right now: the expectation that this technology is going to restructure every industry, and you want a position in the company that does the restructuring. Meituan has watched Alibaba and Tencent already make their moves. Sitting out the AI wave while your rivals collect equity stakes in frontier labs is not a strategy any serious Chinese tech executive accepts.
At $20 billion, Moonshot is worth roughly as much as Snapchat. But Snapchat is not building frontier AI models that Silicon Valley’s best coding tool is quietly depending on. The Chinese AI ecosystem has a depth that Western analysts consistently underestimate, and this round is another data point in that argument.
SAP DROPS $1.16 BILLION ON AN 18-MONTH-OLD GERMAN AI LAB BECAUSE EUROPE IS NOT GOING TO SIT THIS ONE OUT
SOURCE: TechCrunch – May 5, 2026
While American tech giants were busy counting zeros and Chinese labs were becoming geopolitical instruments, Europe quietly made one of its most interesting AI deals of the year. SAP, the German enterprise software company that runs the back office of half the world’s corporations, announced it is acquiring Prior Labs, an 18-month-old AI startup based in Germany, and committing one billion euros over four years to grow it into a dedicated AI research lab. This is not a story that generates trending topics. But it is a story about what happens when a 50-year-old German company decides to take the technology seriously.
Prior Labs was founded by Frank Hutter, Noah Hollmann, and Sauraj Gambhir. Hutter is a professor at the University of Freiburg who has been working on machine learning since before the term was fashionable. The lab built a model series called TabPFN, specifically designed for structured data, meaning rows and columns, databases, and spreadsheets, the kind of information that enterprise software lives and breathes. That model has been downloaded more than three million times by developers.
This is an important distinction. Most of the AI news you read is about large language models that can write essays and code and answer questions in conversational language. TabPFN does something different. It runs predictions on tabular data, the unglamorous backbone of every business on earth. Sales forecasts. Inventory models. Financial projections. If you run SAP’s software, you are generating tabular data constantly. And SAP just bought the team that arguably builds the best AI for understanding exactly that kind of data.
The acquisition price was not disclosed, but sources told TechCrunch the founders walked away with well over half a billion dollars in cash up front. For an 18-month-old lab, that is an extraordinary outcome. For SAP, it is a rounding error on its market cap, but a meaningful signal about strategy. SAP has been fighting a quiet war for relevance. The company generates most of its revenue from ERP software, but its customers are nervous. They are watching OpenAI and Microsoft build products that look a lot like replacements for what SAP does. The Prior Labs acquisition is SAP’s answer: we will not be replaced. We will build the AI that makes our platform indispensable.
The billion-euro investment over four years will turn Prior Labs into SAP’s internal AI research arm, available to the thousands of enterprises already running SAP systems. Europe has largely been a spectator in the AI race, contributing regulation more than innovation. This deal does not change that picture entirely. But it is a reminder that European technology money is still capable of making serious bets when the right opportunity appears. And apparently the right opportunity is a German professor who taught a machine to understand spreadsheets better than anyone else on earth. You have to start somewhere.
Keywords: NextEra Dominion acquisition AI power, Blackstone Google AI cloud TPU, Goldman Sachs Anthropic joint venture, DeepSeek $45 billion valuation first funding round, Moonshot AI $2 billion Meituan, SAP Prior Labs acquisition German AI, AI infrastructure investment May 2026