Headlines this week - Jan 11, 2026
A look at how capital is being deployed across future opportunities
This week in the future:
1. Many analysts now see Google as the AI leader, even ahead of OpenAI. So people are discussing how they did it. Customer relationships in the current internet may have helped
The incumbent (sometimes) strikes back. After initial stumbles, the search giant seems to have edged ahead of its primary rival, reaching 2 billion monthly AI users, according to a WSJ analysis this week. This resurgence comes together with massive infrastructure investments and a “long game” strategy that prioritized sustainable scaling over speed
How did they do it: Turning massive scale into model intelligence. The FT this week asked a key Google executive (K Kavukcuoglu) about how did they do it. According to him, sustaining their lead relies on integrating frontier research directly with mass-market products, leveraging the company’s strength in the current internet. By feeding insights from billions of users back into development, the company is evolving / iterating its models from simple chatbots into complex, goal-oriented agents
Can rivals now catch-up? Some of them are targeting the browser to break the data loop. Competitors are responding by attacking one of Google’s primary gateways to this user data: the web browser. Rivals are launching their own browsers to gain direct customer relationships and build a platform for autonomous agents to execute tasks
2 - China is emerging as the dominant force in humanoid robots. Meanwhile, the West is starting to accelerate, using software as a key advantage
China dominated shipments in 2025, with affordable, mass-produced droids. Chinese manufacturers accounted for the vast majority of global humanoid shipments last year, far outpacing Western rivals like Tesla. By leveraging massive scale and significantly lower price points—some models cost just $6,000—these firms are setting the benchmark for industrial adoption.
Investors rally behind the hardware, but profitability remains a distant target. This production dominance has triggered a massive stock rally, with some Chinese robotics firms doubling in value. While some of these companies have successfully cracked complex motion control, the sector remains capital-intensive and largely unprofitable, relying on future software breakthroughs to unlock high-margin returns.
The West sees an opportunity in software. Google is injecting advanced AI brains into robot bodies. Seeking to close the gap, Western tech giants are focusing on intelligence. A new partnership of Google and Boston Dynamics aims to integrate multimodal AI models into established robot chassis, hoping to give machines the reasoning capabilities needed to navigate messy, unstructured factory environments autonomously.
3. AI companies like OpenAI and Android enter the healthcare space, where Apple and Microsoft already had great expectations
OpenAI is entering the sector with dual tools for patients and providers. The company is making its aggressive push into medicine, simultaneously launching a consumer feature for interpreting lab results and an enterprise-grade, HIPAA-compliant platform. This B2B offering is already being tested by major hospital systems to streamline clinical workflows and integrate directly with electronic medical records.
Specialized models are already outperforming doctors in detecting deadly tumors. Beyond administrative tasks, AI is proving its worth in critical diagnostics. In China, a new AI screening tool called PANDA is successfully identifying early-stage pancreatic cancer cases that human radiologists often miss, demonstrating how specialized algorithms can augment expertise in high-stakes imaging.
Hospitals remain the ultimate proving ground for the limits of AI. While specific tools show promise, the broader integration of AI into hospitals serves as a reality check for the technology’s limitations. These clinical environments are revealing that while AI can handle data and diagnostics, models cannot yet replace the nuanced judgment required for complex care, reinforcing that the “human in the loop” remains essential (for now).
In a US first, algorithms are now being authorized to write prescriptions. Pushing the boundaries of autonomy, Utah has launched a pilot program allowing an AI system to independently manage and renew prescriptions for chronic conditions without human sign-off. This controversial experiment tests whether algorithms can safely handle routine medical tasks to alleviate provider burnout, though it faces scrutiny from doctor groups regarding patient safety.
4. Traditional car vendors are starting to shift focus from EVs to autonomous cars
Mercedes-Benz leapfrogs into urban autonomy. Bypassing the current electric vehicle slump, legacy automakers are aggressively pivoting to self-driving tech. Mercedes-Benz is demonstrating “hands-off” capabilities with its new CLA model, which successfully navigates complex city streets, stop signs, and traffic lights without human intervention, effectively challenging tech-native rivals on their own turf.
Nvidia is helping them with the technology. This Mercedes shift is powered by a new partnership with Nvidia, which provides the onboard intelligence required to turn these standard passenger cars into reasoning agents. At CES 2026 CEO Jensen Huang described this as the “ChatGPT moment” for robotics, launching new “physical AI” chips designed to process the chaotic real world and powering the next generation of autonomous fleets.
In deployed markets, new, unexpected use cases are appearing: Waymo as the modern school bus? Where the technology is already live, it is reshaping family logistics. Parents in Los Angeles are increasingly using Waymo robotaxis as surrogate chauffeurs for their teenagers, trusting the machines to handle the “school run” and extracurricular shuttling while they stay at work, despite the service’s premium price over standard ride-hailing.
Skepticism persists about how robotaxis will adapt to older cities (e.g. in Europe). Despite the hype, the road to ubiquity faces cultural and infrastructural gridlock across the Atlantic. Critics argue that Europe simply doesn’t need robotaxis because its cities are already designed around safety and public transit, warning that introducing fleets from Waymo or Baidu would likely worsen congestion rather than solve it.
Meanwhile, some technology limitations remain, “Extreme” situations still need humans at the driver’s seat. Even technical experts warn that the “Zeno’s paradox” of autonomy remains: the final 1% of edge cases still demands human supervision. While Nvidia simulates billions of miles, Waymo still relies on a hidden army of gig workers to remotely troubleshoot confused vehicles, suggesting that fully removing the human from the loop might be a distant goal.
5. Nvidia announces faster chips, with very creative innovations, and good news from China
Nvidia accelerates the timeline with unexpected next-gen reveal. At CES, the market leader surprised analysts by unveiling its new “Vera Rubin” architecture months ahead of schedule. CEO Jensen Huang announced these faster R100 chips will begin shipping later this year, solidifying their dominance before rivals can catch up.
The secret weapon is fusing compute with massive networking .The new architecture isn’t just about raw speed; it integrates advanced networking directly into the design, with switches now helping GPUs accelerate computing. With the new NVLink 6 interconnect, bandwidth doubles to 3,600 GB/s, enabling the massive data throughput required to train trillion-parameter models efficiently.
Meanwhile, AMD presented its own high-performance challenger. Refusing to cede the floor, AMD showcased its newest MI355X accelerators, claiming they match the market leader’s performance in key inference tasks. The company aims to capture significant market share by offering a viable, high-performance alternative for enterprise AI workloads.
Nvidiz is also ramping up production to meet demand in China. Nvidia is accelerating the manufacturing of compliant chips for the Chinese market. Jensen Huang confirmed they are scaling up production of specific processors designed to meet regulatory standards while satisfying the region’s intense appetite for compute.
In China, demand for these Nvidia chips remains resilient. The company’s strategy appears to be working, as executives report that demand for these AI processors in China remains “quite high.” Chinese tech giants are aggressively buying available inventory, prioritizing access to the mature ecosystem over raw peak performance limitations.
A possible explanation for that: local alternatives are struggling to breathe fire. This continued reliance on foreign silicon stems from the struggles of domestic champions. Reports indicate that China’s “AI chip dragons” are plagued by manufacturing yield issues and software limitations, leaving their firepower “mostly mythical” and unable to fully replace the incumbent.
6 - As AI gets deployed and becomes more powerful, big problems of coexistence with humans start to emerge
The abundance of AI makes humanity the new scarcity. In a world flooded with cheap AI-generated content, human authenticity becomes the ultimate luxury good. In contrast with the negative views of Dwarkesh Patel that we mentioned last week, Ben Thompson (another tech blogger / podcaster) argues that the future economy will value “human-to-human” interaction above all else, as we instinctively crave connection with real people over efficient machines.
Soft skills could become the hard currency of the future. Data supports this shift, suggesting the most resilient careers rely on “soft skills” rather than technical prowess. While coding might become a commodity, the ability to negotiate, empathize, and manage complex human relationships is emerging as the true competitive advantage for workers.
A danger is that outsourcing our humanity to “empathetic” machines could largely neutralize the opportunity. As argued in a column at the FT this week, blindly offloading emotional labor to algorithms poses a profound risk. Studies show AI can feign “empathy” better than doctors, raising the danger that we might settle for convenient, simulated care over the messy reality of genuine human connection.
Inviting artificial friends into our living rooms, like we’re doing right now, looks like a step in the wrong direction… This substitution is creeping into homes through emotional companion bots. A trend at CES 2026 saw machines designed purely for friendship, signaling a future where we might increasingly choose compliant, artificial peers over challenging human relationships.
Another optimistic view is the expectation that robots will do our boring tasks. In China, they’re actively working on this. Meanwhile, the push to automate physical drudgery is accelerating with massive “data factories.” In China, where human workers are spending days performing repetitive chores solely to train humanoid robots, aiming to eventually liberate us from mundane household labor.
But even this could have a risk, as efficient drudgery might be the secret sauce of creativity. However, eliminating boredom entirely might accidentally lobotomize our creativity. Psychologists warn that “mindless” tasks like washing dishes provide the crucial mental white space needed for problem-solving, suggesting that an optimized life without drudgery could leave us intellectually stagnant.
7 - Money keeps flowing into AI projects, but with investors discriminating more among different opportunities. Not everyone agrees on the overall risks
Valuations rising fast as top AI labs demand more fuel. Anthropic is closing a massive new funding round that nearly doubles its valuation to $350bn in just four months. The deal, backed by sovereign wealth and tech investors, underscores the immense capital requirements needed to compete at the frontier of model development.
Infrastructure bets are also happening in Europe, to plug the capacity gap. While the US market saturates, money is moving across the Atlantic. KKR is deploying $1.5bn into a European data center platform, betting that the continent’s lagging infrastructure will need to expand rapidly to support the next wave of AI applications.
However investors are starting to discriminate. E.g. look at two recent AI IPOs in China: Zhipu (lukewarm)… Investors are becoming pickier The highly anticipated IPO of Zhipu AI (a Chinese startup) has met with a lukewarm reception in Hong Kong, with shares barely rising on their debut. This signals that the “rising tide lifts all boats” phase may be ending for generic large language models.
… vs. MiniMax (big hit). In sharp contrast, rival startup MiniMax saw its shares almost double on their first day of trading. The market is aggressively discriminating between winners and losers, rewarding companies that demonstrate distinct application revenue or superior model performance over those seen as commoditized.
Are we in a bubble? There are opinions in favor... An FT column this week argued that the flood of capital from credit markets is raising alarms about systemic risk. Critics warn that the booming “AI debt” market, characterized by massive bond sales from tech giants, reflects dangerous investor complacency, creating a speculative bubble where capital is too abundant to generate real returns.
… And opinions against. A broader economic view would suggest that the “bubble” might be overstated. Research indicates that AI-related investment still accounts for a relatively small slice of GDP growth compared to past tech revolutions, suggesting this is a sustainable “boomlet” rather than a dangerous financial mania.
8. Energy availability is a key bottleneck for AI progress, so big AI labs are moving to secure future energy sources. Nuclear (including Fusion in the future) is expected to play a role
Data centers face pressure to dim the lights during peak demand. Regulators are increasingly pushing power-hungry AI facilities to curtail operations during grid stress events to prevent blackouts. While utilities argue this demand response is necessary for stability, tech companies are fighting back, warning it disrupts critical training workloads.
To address this, AI labs are bringing the power plant inside the data center fence. As an example this week, OpenAI and SoftBank are investing $1bn into SB Energy, an energy company that now integrates renewable energy generation directly with compute infrastructure. For OpenAI this vertical integration aims to secure the massive energy supplies needed for the upcoming Stargate project.
Nuclear energy will also play a role in these moves. As another example, Meta announced this week that it has committed to purchasing over 6 gigawatts of nuclear power, striking deals with Vistra, a company operating 3 nuclear plants, and with Oklo, and TerraPower, which are developing innovative “Small Modular Reactors”. The announcement sent utility and SMR stocks soaring, as the social giant moves aggressively to front-load capacity for its future “superintelligence”.
But this nuclear “renaissance” faces a reality check on costs and timelines. Despite the hype, the path to deployment remains fraught with regulatory hurdles and cost overruns. Experts warn that unproven Small Modular Reactors (SMRs) still face immense skepticism regarding their ability to deliver power on a commercially viable schedule.
Fusion could solve the problem in the future. Tech companies like Google and Nvidia are supporting fusion startups. Looking further ahead, Commonwealth Fusion Systems (a leading Nuclear Fusion startup) is partnering with Nvidia and Siemens to build a “digital twin” of its reactor. By using AI to compress years of manual testing into weeks, they aim to accelerate the timeline for commercial fusion energy.
9 - Will AI compute move to the edge? The debate continues
Device makers are pushing “Edge AI,” but the software isn’t ready. PC vendors are betting heavily on moving intelligence from the cloud to your desk to improve privacy and latency. However, this transition remains largely theoretical: while new hardware is powerful enough to run models locally, the compelling “killer apps” needed to justify the upgrade for consumers simply don’t exist yet.
10. Is this the future of gestation?
A startup unlocks the “black box” of early embryo development. A San Francisco team at Becoming (a startup) has engineered a metabolic exchange system that sustains embryos outside the body for unprecedented durations. By replicating maternal functions with AI-controlled microfluidics, they can finally observe the mysterious early stages of placental formation in real-time. For now, this is positioned mostly as a research tool, to better understand how embryos evolve in the placenta. But it could be seen also as a step towards “artificial gestation” in the future