Headlines this week - Jul 20, 2025

A look at how capital is being deployed across future opportunities

C Gavilanes

Jul 20, 2025

This week in the future:

1 - Meta has a new AI strategy, and the new “Superintelligence” unit is just the tip of the iceberg

Talent and infrastructure are the key priorities. Frustrated by the quality of past AI efforts, Mark Zuckerberg is personally driving an aggressive strategy to build momentum in the race for superintelligence. This involves poaching elite talent, such as two key researchers from Apple this week, and a plan to invest "hundreds of billions of dollars" into massive, multi-gigawatt data centers like the upcoming "Prometheus" cluster.

The company has an aggressive plan for new data centers, using a brand new design. Meta's infrastructure plan is a complete strategic reinvention aimed at leapfrogging competitors. Beyond just building massive facilities—one touted as nearly the size of Manhattan—the company is adopting an "'All Of The Above' Infrastructure Strategy," reportedly shifting from traditional buildings to more agile "tent" structures to accelerate deployment and build the world's largest training clusters.

Is Meta renouncing to open-source? The company’s new strategy (and some of its recently hired executives) may be driving a major philosophical shift. According to the NY Times, the new Superintelligence lab is discussing abandoning Meta's powerful open-source model, Behemoth, in favor of a closed-source approach. This potential "shake-up" comes after the unreleased model showed poor internal performance and represents a significant departure from the company's long-held public advocacy for open-source AI development.
"Acqui-hiring" looks like a successful strategy, and it could even be good for everyone else. Meta (ScaleAI) and Google (Windsurf) have recently used "acqui-hiring"—specialized deals that scoop up a startup's key talent while avoiding traditional acquisitions and antitrust scrutiny. At Bloomberg this week, P Olson claimed that a VC investment framework based on this could lead to more sustainable future startups, and even be beneficial for society.

2 - Nvidia gets approval to resume key chip sales to China

The US government seems to have accepted Jensen Huang’s argument that a ban would benefit Huawei. The policy reversal follows an intense lobbying campaign by Nvidia’s CEO, Jensen Huang. According to the WSJ, Huang had warned Washington that a continued ban risked the U.S. ceding its AI leadership to Chinese rivals like Huawei, which would benefit from the absence of American competitors.

Nvidia’s investors have received these news very positively (maybe too much?) The news fueled the latest leg of Nvidia's powerful stock rally, which has seen shares rise over 80% since April. However, a Bloomberg analysis suggests the rally shows signs of overheating, with some analysts describing market sentiment as "borderline giddy" and technical indicators suggesting buying is at extreme levels.

However, political uncertainty remains, as the Trump administration is unpredictable. A WSJ analysis cautions that "Nvidia's China Troubles Aren't Over," suggesting the celebration may be short-lived. The article points to the unpredictable nature of the Trump administration, noting that Nvidia remains a "pawn in the bigger geopolitical game" and the ban could easily be re-imposed or replaced with high tariffs.

3 - The “nuclear renaissance” in the US keeps gathering momentum

Constellation Energy is planning a new project in New York Responding to a push from the State Governor, the largest US nuclear operator is planning a new reactor project in upstate New York. As reported by Bloomberg, the plan seems to be to build at one of their 3 already existing nuclear sites in the state.

Westinghouse expects to build 10 new reactors in the coming years, using AI to improve the construction process. At a conference on energy and AI, Westinghouse's interim CEO told Trump of the company's plan to build 10 large AP1000 reactors, with construction to begin by 2030. CNBC reports that to avoid past project delays, Westinghouse is partnering with Google to use AI to make construction an "efficient, repeatable process".

Momentum is also building for Small Modular Reactors (SMRs). The advanced reactor startup Oklo has hired a lead constructor for its first small nuclear plant at Idaho. In an interview on Bloomberg TV, Okla’s CEO stated that the plant may be complete by late 2027, a sign of tangible momentum (and timeline advantages) for next-generation nuclear technology.

4 - OpenAI is looking for a monetization mechanism

The company is integrating e-commerce into ChatGPT's answers. In a strategic hunt for new revenues, OpenAI plans to take a cut from online sales made directly through ChatGPT. The FT reports the company aims to integrate a checkout system into the platform, allowing it to earn a commission and, for the first time, monetize its millions of free users.

It also plans to add slide and document creation features, in direct competition with Microsoft. OpenAI has unveiled a new agent that can create spreadsheets and presentations, a move that puts it in direct competition with its key partner, Microsoft. The WSJ describes these new features, which mimic the core functions of Microsoft Office, as a significant strategic step by OpenAI into the business software market.

5 - Genome sequencing is starting to deliver tangible, though controversial, applications

Startups like Orchid are sequencing human embryo DNA to screen for disease risk. Orchid is part of a Silicon Valley push to engineer healthier future generations by sequencing the DNA of IVF embryos. This allows prospective parents to screen for genetic risks and select which embryos to implant, moving the technology from a niche service to a well-funded venture.

There are significant public initiatives around this topic at both sides of the Atlantic This technology is also being integrated into national healthcare systems. The UK's National Health Service is launching a national program to offer genomic screening to every newborn, building on a successful 100,000-infant pilot study. Similarly, in the US, Florida has passed a "Sunshine Genetics Act" to establish its own newborn screening pilot program.

This progress is exciting, but it also sparks moral objections and a potential political battle. This rapid progress is creating a fierce ethical debate, which The Washington Post frames as a choice between "Russian roulette or a jet to the future". While proponents see a chance to eradicate devastating diseases, critics raise moral objections and point to challenges around privacy, equity, and the potential for widening health disparities, setting the stage for future political battles.

6 - AI labs valuations keep growing. More data points this week

Anthropic could be worth more than $100bn OpenAI-rival Anthropic is receiving "pre-emptive funding offers" from investors at a valuation of over $100bn, even though it is not formally fundraising. Bloomberg reports this interest is fueled by surging revenue from its chatbot, Claude, with annualized revenue climbing from $3bn to $4bn in the last month alone.

Perplexity now has a valuation of more than $18bn The AI search engine's valuation has jumped to $18bn, just two months after a funding round valued it at $14bn. According to the FT, this rapid increase was driven by new investors who approached Perplexity after its last funding round had already closed, highlighting the intense investor appetite for top AI startups.

7 - AI is disrupting the cloud market, with OpenAI as a "kingmaker"

Oracle is now a key player, after becoming a partner of choice for OpenAI No longer an AI "dark horse," Oracle's stock has surged after its cloud backlog swelled with AI deals, including a key partnership to provide Microsoft with additional capacity for OpenAI. The WSJ notes this has dramatically boosted Oracle’s credibility and market value in the AI infrastructure race.

OpenAI's deal with Google Cloud could significantly increase Google’s market share. OpenAI will now use Google's cloud infrastructure in several countries to meet heavy demand for ChatGPT, adding Google to its list of suppliers alongside Microsoft and Oracle. CNBC frames this as a significant "win for Google," whose smaller cloud unit gains a major AI client and crucial validation for its infrastructure.

Meanwhile, Meta is working with AWS to bolster its Llama ecosystem. Meta and Amazon Web Services are teaming up to support startups building on Meta's open-source Llama models. As reported by CNN, the program provides technical support and $200,000 in AWS credits to each startup, a strategic move by Meta to strengthen its AI ecosystem while bringing new, high-growth customers to AWS.

8 - Robots and drones are consolidating as a key investment theme

They are already playing a key role in geo-strategy. E.g. Helsing (a German startup) is profiting from Europe’s commitment to become strategically self-sufficient, in fields like drones and autonomous weapons. According to The WSJ, Helsing is developing AI-powered systems like "drone walls," with a vision that Europe has "no choice but to build anew" its defense capabilities, creating a significant new market.

The US is concerned about its weakness in drone manufacturing relative to China. This is also aligned with robots becoming a geo-strategic priority The NY Times highlights a significant US strategic vulnerability in drone manufacturing compared to China. It notes that even the best US alternative to China's DJI assembles drones by hand, producing only about 1,500 units a month, a stark contrast that underscores the geo-strategic challenge facing the American industrial base.

“Self-sustainable robot ecologies” are on the way Researchers at Columbia Engineering have developed a process called "Robot Metabolism," allowing robots to grow and heal by consuming parts from other machines. Researcher H Lipson explains that for true autonomy, "robots must ultimately learn to take care of themselves," paving the way for self-sustaining robot ecologies in space exploration or disaster recovery.

9 - Managing the impact of AI on jobs is becoming a priority

This might keep policymakers busy in the coming years, as they’ll need to grapple with the AI's impact on the labor market. An FT article this week warns that a fall in white-collar employment could hollow out the middle class, a traditional bulwark of democratic stability, and fuel a new wave of populism.
The current Silicon Valley fight for top AI talent could make the problem worse While cutting staff in some areas, Big Tech is creating premium jobs “for the fortunate few" in AI, exacerbating inequality. An FT Lex column notes the paradox of companies like Meta and Google having fewer total employees than in 2022, while simultaneously spending billions to "acqui-hire" elite AI researchers whose work will "further diminish staffing needs".
This could lead to a (less than ideal) scenario where top AI experts are paid like Hollywood actors or NBA players. As discussed by VC investor J Luttig this week, these dynamics could have plenty of negative consequences for society.

10 - The ability of AI models to accelerate science may be overhyped

In the short term, AI could actually slow down scientific progress. A post this week by A Narayanan and S Kapoor argues that while the current mood is optimistic, AI adoption could initially have a negative impact. The authors suggest that AI might prolong reliance on flawed theories and that science's poor software practices are not ready for complex AI, predicting a "bumpy ride" before the long-term benefits are realized.