Headlines this week - Jan 18, 2026
A look at how capital is being deployed across future opportunities
This week in the future:
1 - Google becomes Apple’s partner of choice for AI (and Siri)
Apple chooses a rival to power its AI future. Ending months of speculation, the iPhone maker has selected its longtime competitor to underpin its new intelligence features. This pragmatic alliance allows it to bypass the exorbitant costs of training frontier models, relying instead on a partner’s established infrastructure to deliver generative capabilities to billions of devices.
The deal breathes new life into a stagnant assistant. The integration will specifically overhaul the company’s voice assistant, replacing its limited command responses with advanced reasoning capabilities. By leveraging these external models, Siri will evolve from a simple task executor into a conversational agent capable of summarizing communications and managing complex user requests.
Markets reacted very positively, turning Alphabet into a new $4trn “tech titan”. Investors immediately cheered the union, pushing the search giant’s market capitalization past the $4trn threshold. The rally reflects renewed confidence in its technical leadership, validating its massive infrastructure bets and confirming the vision that it is catching up in the generative race (or even winning it).
Some people believe this might also reinforce Apple, now playing as kingmaker between OpenAI and Google. By choosing to license rather than build, the device giant avoids the “CapEx wars” while retaining control over the user experience. This decision deals a strategic blow to early leaders like OpenAI, suggesting that owning the distribution channel might ultimately beat owning the model.
2 - Google and OpenAI are moving to improve AI monetization, using “traditional” levers (ads and e-commerce)
Google is introducing personalized ads to monetize its chatbot reach. The company is moving to monetize its hundreds of millions of free users by introducing personalized advertising into its AI shopping tools. This shift moves beyond traditional search links, allowing advertisers to present exclusive offers directly within conversational responses to capture high-intent shoppers. Let’s see what this does to customer experience…
A second stream would come from helping major retailers deploy AI agents to capture sales directly. Beyond simple ads, Google is launching purpose-built “AI agents” that handle complex transactions for third-party merchants. Major chains like Kroger and Papa John’s are already testing these tools, which allow customers to order groceries or pizza via natural language, effectively bypassing traditional apps.
For Google, “deep personalization” could reinforce its moat built on user history. To make these interactions stickier, the company’s models are tapping into the company’s vast data ecosystem. By “reasoning” across a user’s emails, photos, and video history, the AI is evolving from a transactional bot into a context-aware partner that knows its user better over time.
OpenAI is responding by also testing targeted ads to boost revenue. Meanwhile, OpenAI is also embracing the ad-supported model to diversify its income. The startup will soon test targeted advertisements for users on its free and lower-cost tiers, arguing that sponsored results will help users make better shopping decisions.
In this context, AI chatbots / marketing wars are also heating up, with OpenAI playing defense. This commercial pivot comes as OpenAI faces intense pressure from its rapidly growing rival. With Gemini now boasting 600m monthly users, OpenAI is reportedly purchasing expensive Super Bowl airtime for the second year in a row to reassert its dominance and brand value.
3 - Competitors to Nvidia might be getting stronger, also in China (but news are contradictory)
Is Nvidia vulnerable in inference? OpenAI bets billions on Cerebras (a new challenger). The industry leader has signed a massive $10bn agreement with Cerebras Systems, an inference chip specialist, aiming to build specialized infrastructure for AI inference. This strategic move signals one more shift away from total reliance on Nvidia, betting that alternative architectures can deliver faster, cheaper performance for running models.
Meanwhile, China (and Huawei) may have achieved a symbolic breakthrough in domestic self-sufficiency. Zhipu AI, one of the Chinese AI labs that recently had their IPOs, claims to have successfully trained a major open-source model entirely on Huawei processors, proving that US silicon is no longer strictly required. This milestone demonstrates that domestic hardware stacks are finally maturing enough to support complex, large-scale training runs without foreign technology.
Yet, the broader Chinese AI industry admits they do still need Nvidia. Despite individual successes, the consensus among Chinese developers remains grim, according to people interviewed by the WSJ (including a Zhipu founder). Many top engineering teams privately concede that the hardware gap is widening, preventing them from matching the scale and speed of US models and effectively leaving them unable to compete at the true frontier.
But Beijing seems to be prioritizing self-sufficiency vs. short-term progress, by blocking sales of Nvidia’s advanced chips . Complicating matters further, Chinese customs officials have blocked shipments of Nvidia’s compliant H200 chips. This unexpected freeze suggests a government strategy to force domestic adoption, even if it means short-term pain for local companies desperate for advanced compute.
4 - AI is creating a lot of stress in the supply of memory chips. The crisis is expected to impact consumer devices this year
The inevitable price hike: Shortages hit consumer electronics. As AI devours global capacity, the shortage of standard memory chips is trickling down to consumer goods. Analysts predict significant price increases for smartphones and PCs this year as manufacturers pass on the soaring component costs to buyers.
Past experiences make producers reluctant to expand capacity despite the AI-driven squeeze. The scarcity is exacerbated because the boom in AI-dedicated memory is cannibalizing production lines for standard chips. Yet, major players like Micron and Samsung are refusing to build new factories, fearing another bust cycle, to ensure that prices stay high.
Meanwhile, in China, startups are working to replace traditional leaders. Capitalizing on this caution, Chinese challenger CXMT is aggressively expanding capacity to capture the underserved legacy market. However, looming US export controls threaten to stifle this rapid ascent, potentially cutting off a critical relief valve for global supply chains.
5 - The coming wave: more signs of the massive impact that AI will have on jobs
Anthropic builds a viral tool for workers to automate the “boring stuff”. The company has released “Cowork,” a new agentic tool capable of independently building apps, managing spreadsheets, and sorting data. Notably, the software itself was largely written by the company’s own AI models, demonstrating a recursive loop where systems are now sophisticated enough to engineer their own successors.
McKinsey makes the ability to “collaborate” with AI a requirement for its new hires. Signaling a shift in what defines talent, the consulting giant is now requiring graduate applicants to use its internal AI, “Lilli,” during job interviews. The pilot program assesses candidates not just on their raw problem-solving abilities, but on their skill in prompting and collaborating with synthetic intelligence to refine their answers.
Goldman Sachs is preparing for a “dramatic”, efficiency-driven transformation. CEO David Solomon is staking his legacy on a new strategic phase dubbed “One Goldman Sachs 3.0,” which prioritizes extreme productivity gains. With revenue per employee already rising, the bank is betting that AI can radically reshape its workforce structure, allowing fewer bankers to generate significantly higher returns.
London is scared about the potential fallout of white-collar displacement. While banks eye efficiency, politicians fear the social cost. Mayor Sadiq Khan is warning that the capital faces a “new era of mass unemployment” as AI automates entry-level roles in finance and creative industries, potentially destroying traditional career ladders faster than the economy can create new ones. The impact on investment banking (as anticipated by Goldman -see above) could turn this into a key problem for London.
And the machines keep advancing: a new milestone for “AI-generated maths” happened this week. Compounding these fears, AI capabilities are jumping from imitation to genuine discovery. Google DeepMind has used its Gemini model to help prove a novel theorem in algebraic topology, demonstrating that algorithms can now act as creative intellectual partners in solving abstract problems previously thought to be the exclusive domain of human genius.
6 - However, there are still significant barriers to adoption (as with most general-purpose-technologies)
Simplicity might be the ultimate gatekeeper of digital success. History suggests that for new tools to gain widespread adoption, they must be intuitive rather than intimidating. Transformation efforts often fail when technologies make users feel incompetent; if a system requires complex explanations or disrupts workflow, it will simply remain unused.
Even high-stakes industries face the “last mile” problem. Nowhere is this friction more visible than in law, where adoption remains sluggish despite the hype. Concerns over accuracy and inertia around billable-hour models create headwinds, forcing firms to balance the promise of efficiency against the risk of costly hallucinations.
Some companies want to accelerate the transition by teaching machines to mimic us directly. To bridge this gap, companies are hiring experts to train algorithms on specific tasks, effectively codifying human intuition. Former professionals are finding lucrative work teaching models to replicate their old roles, creating a “direct substitution” path that could bypass complex integration hurdles, and accelerate adoption.
Yet the most durable skills could remain distinctly human. Despite the hype on “automation”, some analysts argue that long-term career survival won’t come from mastering specific tools, but from cultivating “human-centered” capabilities. As technical barriers lower, the ability to define problems and collaborate becomes more valuable than the coding itself.
For some people, even the “messy” middle management layer would be surprisingly safe. This reinforces why middle managers might actually be secure. Their role (translating high-level strategy into execution and managing human nuance) requires a level of empathy and context that current AI agents simply cannot replicate, making them the indispensable “glue” of the organization.
7 - Is “merging” with machines the future? Merge Labs, a startup, is working to create brain-machine hybrids
OpenAI bets on the ultimate interface: merging minds with machines. Backed by Sam Altman, this startup is developing a non-invasive headset designed to fuse human thought with artificial intelligence. Unlike more “invasive” rivals (like Neuralink), they aim for a “mass market”, “easy to wear” brain-computer interface that seamlessly integrates digital capabilities directly into the user’s mind.
8 - A political debate keeps expanding about data centers, given the stress they generate on electric demand
America’s largest grid operator warns of a looming supply crisis. PJM Interconnection, which serves 65m people, warns that data center energy demand in hubs like Northern Virginia is rapidly outpacing supply. With retiring power plants and transmission bottlenecks, the grid faces imminent reliability risks that could trigger blackouts during peak usage.
Meanwhile, a “green energy rush” driven by chip manufacturers is triggering social conflict in Taiwan. In Taiwan, the aggressive rollout of wind farms to power semiconductor giants is upending rural coastal livelihoods. Farmers and fishers report that rapid infrastructure expansion is destroying oyster beds, fueling local resentment against the industry’s insatiable appetite for renewable energy.
US politicians are starting to act: New York demands data centers pay for their own surge. New York regulators are responding with strict proposals requiring large data centers to either generate their own electricity or pay premium rates. The initiative aims to shield working families from skyrocketing utility bills caused by the tech sector’s massive consumption.
Under public opinion pressure Microsoft is pledging to protect consumers from rising bills .Seeking to defuse this growing political backlash, Microsoft has vowed to “pay its way” for new infrastructure. The company pledged to cover grid upgrade costs and pay higher rates in specific regions to ensure its AI operations do not inflate residential electricity bills.
The company’s strategy also aligns with the White House’s push for lower energy costs. This proactive approach quickly garnered praise from Trump, who cited the pledge as a model for the industry. The administration views shielding voters from AI-driven energy inflation as a critical priority for maintaining economic stability and public support.
In this context, the need to secure power sources is driving tech firms to aggressively poach utility talent Beyond public pledges, tech firms like Google and Amazon are quietly building internal power utilities. For this, they are hiring grid experts and energy traders directly from traditional providers to secure complex power purchase agreements and manage their own independent infrastructure.
Nuclear energy (and Small Modular Reactors in particular) is expected to play a key role. While hiring sprees offer short-term fixes, the long-term solution likely lies in nuclear scalability. Many analysts argue that Small Modular Reactors (SMRs), despite longer construction timelines, offer the only viable path to generating the consistent, carbon-free gigawatts required by future models.
9 - Who will own the AI that countries deploy globally? Many people think it won’t be the US
Naver, from South Korea, pitches a “sovereign” alternative to the superpower duopoly. South Korea’s search giant is aggressively marketing its cloud services to the Middle East and Southeast Asia as a safe “third option.” By promising localized data control and customization, Naver aims to win over nations wary of becoming dependent on American or Chinese tech giants.
But even “native” South Korean efforts struggle to escape the gravitational pull (of China). The complexity of building truly independent infrastructure was laid bare when South Korea’s government-backed “sovereign AI” project faced backlash. Reports revealed the initiative relied heavily on Chinese open-source code to catch up, undermining claims of technological independence and highlighting the difficulty of building a purely domestic stack.
Meanwhile, according to Microsoft, China would be quietly capturing the Global South. While the West focuses on the high end, Microsoft warns that China is winning the AI race in emerging markets. By flooding Africa and the Global South with efficient, low-cost models like DeepSeek, Beijing is establishing the dominant technological standard for the next billion users before US firms can react.
10 - The US-China rivalry is moving to space
For some analysts, the US might be losing the race, due to internal slashing. While the White House has officially declared a contest for lunar dominance, aggressive budget cuts and staff reductions at NASA are creating chaos. Critics warn that this “General Sherman” approach to dismantling government capacity is effectively handing the lead to Beijing’s steady, state-backed advance.
Nuclear energy could become the critical path to staying power. Despite the turbulences around their space projects, the US is pressing ahead with plans to deploy a fission reactor on the lunar surface by 2030. NASA leaders argue that mastering off-world nuclear power is the absolute prerequisite for the permanent bases required to actually “win” the moon, rather than just visiting it.