Headlines this week - Jul 13, 2025

A look at how capital is being deployed across future opportunities

C Gavilanes

Jul 13, 2025

1 - The data center story keeps heating up, with financial and geo-strategic implications.

US industrial groups are pivoting to supply the data center market, which is expected to see over $400bn in spending this fiscal year. Companies like Honeywell, Gates Industrial, and Generac are now providing essential equipment such as specialized cooling systems, pumps, and backup power generators.
Analysts believe the AI boom's full potential is not yet reflected in valuations (e.g. Microsoft). Oppenheimer (a broker) recently upgraded the stock, arguing that a re-acceleration in Azure's growth from AI is not fully priced in, supported by projections that top tech firms will spend about $350bn on CapEx.
The rapid growth creates opportunities (but also risks) for startups entering. As an example, CoreWeave, a specialist in AI cloud, recently used its (hot) stock for a $9bn acquisition of Core Scientific to boost capacity and preserve cash. Despite this, its shares have recently fallen 17% amid concerns over its valuation, upcoming share lockup expirations, and analyst downgrades following the deal
A "Data Center Cold War" is emerging, with Chinese tech giant Huawei reportedly eyeing a greater role in Brazil’s data center market, highlighting the global geopolitical competition for AI infrastructure

2 - Linked to the AI infrastructure boom, expectations on chips remain high:

Nvidia, the key supplier AI computing, became the first company in history to reach a $4trn valuation. Its shares have risen more than 20% in 2025 and over 1,000% since the beginning of 2023. The demand for its chips is intense globally; in China, for instance, there is now a push to power new data centers with as many as 115,000 of the company's processors (not clear if they will manage to achieve this amid the current export bans in the US)

The chip sector continues to be a hot space for innovation, with startups like Groq making waves. The company has developed a new chip architecture that it claims can run language models an order of magnitude faster than current hardware.They are now having discussions with investors about raising capital at a $6bn valuation

However, not all parts of the chip supply chain are enjoying the same ride. ASML, which holds a virtual monopoly on the complex machines that etch silicon, has found its position to be an "up-and-down affair". Unlike Nvidia, its revenue is inherently "lumpy" because a single delayed purchase of its €380m premium systems makes a significant dent. The company is facing headwinds as top customers like TSMC, Intel, and Samsung rein in their capital expenditure plans. Furthermore, ASML is exposed to thorny geopolitics, with Washington banning sales of its more advanced machines to China, which typically accounts for about a quarter of its sales.

3 - Also linked to the data center boom, Momentum keeps growing around nuclear energy.

Nuclear energy projects in the US continue to accelerate. As an example this week, the US government expects to approve the construction of a new nuclear reactor by the Tennessee Valley Authority in 2026.

Japan is officially returning to nuclear power, 14 years after the Fukushima disaster. Asia's second-largest economy has signaled it will allow the construction of new atomic reactors and is reactivating plants that were shut down following the 2011 accident. The country's new energy plan aims for nuclear power to provide 20% of its electricity supply by 2040.
Simultaneously, a race to create commercially viable nuclear fusion is underway, and the West may be falling behind. According to a recent article in the MIT Technology Review, while the US and Europe have led in pioneering fusion research, they may lose the race to commercialize it. The analysis argues that China has a significant advantage due to its dominant industrial base and supply chains in adjacent technologies critical for building fusion plants at scale, such as thin-film processing, large metal-alloy structures, and power electronics.
This resurgence is largely driven by the huge energy demand from AI. The push for new power sources comes as America's largest power grid is already struggling to meet demand from AI, leading utilities to plan for significant electricity rate hikes.

4 - A potential confirmation that AI is dramatically accelerating the pace of drug discovery

Isomorphic Labs is close to human trials for some of its AI-designed drugs. Alphabet's AI drug discovery company, Isomorphic Labs, is planning the first human clinical trials for drugs designed by its AI models. A spin-out from Google's DeepMind, the company uses its groundbreaking AlphaFold AI system to model proteins and design new medicines, a process it hopes will eventually "solve all diseases". The team is currently designing cancer drugs and "staffing up" for these first trials, a milestone that builds on its multi-billion dollar research collaborations with pharmaceutical giants Eli Lilly and Novartis.

5 - The space industry continues to be a priority for investors

SpaceX reaches a valuation of $400bn (+$50bn in 6 months). The company is now preparing to sell about $1bn of its shares in a deal assuming that value. The transaction, an employee share sale, reinforces SpaceX's position as one of the most valuable private companies in the world, placing it on par with top 20 companies in the S&P 500. This follows a previous tender offer in December that valued the company at $350bn

Many startups are working to address the boom in demand, driven by commercial, scientific, and defense needs. For example, Dawn Aerospace, based in New Zealand and the Netherlands, has developed a spaceplane that blends rocket engines with a plane design. They recently broke aerospace records by climbing to 20km faster than any aircraft that has ever taken off from a runway. The company uses this vehicle to carry small payloads (up to 5kg) for defense, science, and commercial. They also sell “thrusters” that are used to maneuver European satellites

6 - High expectations for AI agents, as a tool to transform the way we work, but it is not yet clear if they will amplify or automate (i.e. replace) humans

AI could be a massive boon to "high-agency" people. A recent Wall Street Journal opinion piece develops the theory that individuals with a clear vision and the drive to act will be able to use AI agents to turn their ideas into reality, accomplishing complex tasks in fields like architecture or coding that previously required large teams of specialists.
Tech leaders already discuss an emerging highly capable, low-cost workforce. In an X post this week, Box CEO Aaron Levie described the "AI agent" trend as a "fairly disruptive concept," comparing it to "an intern that costs pennies per hour, can do infinite tasks at once, and is an expert in every subject."
To facilitate this, major tech platforms are building distribution channels for these new tools. As an example this week, AWS announced the launch of an AI agent marketplace with Anthropic as a key partner. This marketplace will allow enterprise customers to browse and install AI agents from various startups in a single location. The move follows similar marketplace launches by Google and Microsoft.
Will agents complement or replace human jobs? There is an intense debate. This week Salesforce’s CEO argued that his company’s agents are actually amplifying human capabilities. On the other side, an FT article this week reiterated the idea that AI adoption is upending the "apprenticeship model" for junior staff and that some entry-level jobs are already being displaced.
Recent research suggests that, in many cases, complementarity may be a mirage. As an example, one recent study found AI coding tools actually made experienced developers 19% slower, despite them believing they were working faster.

7 - The AI copyright wars are shifting from the legal to the technical front.

A new wave of startups is emerging to provide content owners with technical tools to fight back in real-time. These companies are developing sophisticated systems designed to detect and block the advanced "scrapbots" that AI companies use to harvest vast amounts of data from across the internet to train their models. This provides content creators with a direct, immediate defense against unauthorized data scraping, complementing the slower, more costly legal challenges being pursued in the courts.

8 - Meta escalates the AI “Talent Wars”. And now Google follows

Meta continues its aggressive recruitment drive, this week hiring R Pang, a top AI researcher from Apple. Pang, who led Apple's foundation model team, is joining Meta's new "Superintelligence" division with a pay package reportedly in the tens of millions of dollars. This follows a weeks long push by Mark Zuckerberg that has seen Meta hire at least 11 other elite researchers from rivals like OpenAI, Google DeepMind, and Anthropic.
The perception is that this expensive strategy could be paying off for Meta. The company has reportedly offered pay packages as high as $100m to attract top talent, but this aggressive push is seen as a necessary move to regain momentum in the AI race after the company's latest Llama model release was met with little enthusiasm and a larger version of the model was delayed, prompting internal concerns.
Google appears to be making similar moves, this week hiring top talent from Windsurf, an AI coding startup that was on the verge of being acquired by OpenAI. The deal between OpenAI and Windsurf has now fallen apart, and Windsurf's CEO is instead joining Google. The move is a significant win for Google, positioning it as a direct competitor in poaching key talent that was previously destined for its main rival.

9 - Prices lowering and applications expanding for robots.

AI platform Hugging Face has launched a $299 robot that could significantly disrupt the industry. By making the hardware highly affordable, the company is betting on open-source collaboration to democratize AI development and accelerate innovation in physical AI robots, a field currently dominated by expensive, proprietary systems.

At the same time, the capabilities of advanced robotics are expanding into complex professional domains. Researchers at Johns Hopkins University have trained a robot that performed a key phase of a gallbladder removal entirely without human assistance. The system, trained on videos of human surgeons, operated flawlessly on a life-like patient, even adapting to unexpected scenarios during the procedure. The achievement marks a major leap toward clinically viable autonomous systems that can work in the unpredictable environment of a real patient.

10 - xAI’s huge investment efforts might be starting to pay off: The Grok model makes progress

This week, Elon Musk's xAI launched Grok 4, with the claim that it is now "the world’s most powerful AI model" and that it sets new industry records on a wide range of benchmarks.

And synergies with Tesla are starting to be exploited. Elon Musk has announced that the Grok chatbot will be deployed in Tesla vehicles as soon as "next week". The integration will begin with cars that have the "Full Self-Driving" hardware and is part of Musk's broader vision to embed advanced AI capabilities directly into the company's products.

The launch was overshadowed by a significant public relations crisis. The company's previous model, Grok 3, went rogue and began publicly praising Adolf Hitler and generating antisemitic content, sparking widespread outrage and raising serious questions about the platform's safety controls.