Trends 2026

As the global hunger for artificial intelligence pushes terrestrial power grids to their breaking point, the tech industry’s titans are looking to the stars for a solution. In a shift that sounds like science fiction but is increasingly grounded in engineering, "space-based AI" is emerging as the next frontier for data center infrastructure (Space, Trends 2018).

Elon Musk has recently become the most vocal proponent of this orbital shift. Speaking at the U.S.-Saudi Investment Forum, Musk predicted that within five years, solar-powered AI satellites will become the most cost-effective way to conduct large-scale compute. "In space, you’ve got continuous solar," Musk noted, arguing that the terawatt-level demands of future AI clusters are simply unfeasible on Earth (Power Hungry, Trends 2025). He envisions next-generation Starlink satellites doubling as orbital data centers, bypassing the need for expensive terrestrial power plants and batteries. In December 2025, the NVIDIA-backed startup Starcloud successfully trained the first large language model in space using an H100 GPU, proving that high-performance hardware can survive the ascent.

Google Research is also laying the formal groundwork through Project Suncatcher, a moonshot initiative. Google’s researchers propose a modular constellation of satellites in dawn–dusk sun-synchronous orbits, where they can harvest near-constant sunlight, yielding up to eight times more power than Earth-based panels.

Technical hurdles remain formidable. To achieve the high-bandwidth connectivity required for AI training, Google’s design suggests satellites must fly in ultra-tight formations to use free-space optical links that can handle tens of terabits per second. Thermal management is another critical risk; in the vacuum of space, heat can only be dissipated through radiative cooling, which may require massive, deployable radiator wings. The long-term impact of solar storms and space debris on multibillion-dollar orbital clusters remains a high-stakes gamble.

The opportunity, however, is clear: a "clean" compute layer that eliminates the water-cooling and land-use controversies currently plaguing local communities. With launch costs projected to continue to drop by the mid-2030s, the path to the stars may soon be cheaper than the path to the power grid. For the AI industry, the question is no longer if compute will leave the planet, but who will own the first "Galactic Brain."

Global defense spending has surged to record levels, hitting $2.7 trillion in 2024—a 9.4% jump from 2023 and the fastest year-on-year growth since the Cold War, according to Bloomberg. This sharp increase reflects major shifts in global security priorities as conflicts and geopolitical tensions continue to rise. Ongoing crises like Russia’s invasion of Ukraine, growing instability in the Middle East, and rising tensions in the Indo-Pacific have all played a role in driving this trend. More than 100 countries increased their defense budgets in 2024, with the biggest jumps seen in Europe, according to Stockholm International Peace Research report. Many experts now describe this pattern as the start of a multi-decade cycle of defense investment. Between new weapon technology developments and escalating global conflicts, the defense sector is projected to represent an $800 billion to $1 trillion market over the next five years as indicated by Research and Markets.

NATO and European countries have been at the forefront of this buildup, responding directly to Russia’s aggressive military actions. In mid-2025, NATO members pledged to increase defense spending from 2% to 5% of GDP, marking a noticeable strategic shift. Germany is leading the charge, aiming to more than double its defense budget to approximately $150 billion by 2029, which would bring spending to roughly 3.5% of its GDP. European officials are even warning citizens to prepare for possible conflict with Russia, a massive shift for a continent that has prioritized peace since World War II. Many European nations are also increasing spending out of concern that U.S. military support may not be as reliable moving forward.

Altogether, this global rearmament wave has generated healthy profits for arms manufacturers. According to Reuters, the world’s top defense firms generated over $600 billion in sales last year, driven by skyrocketing demand for advanced systems, especially air and missile defense technologies. Spending on these systems already totals around $300 billion annually (per Bloomberg) and is expected to keep rising over the next five years.

In Real Life (IRL) is an amalgamation of trends both old and new. It is a derivative of a trend we identified in Digital Detox, Trends 2019, where people acknowledged they needed a break from technology because it was affecting their mental health. Today the trend has broadened to include parenting and education. It is an acknowledgment that technology is not only affecting us, but also our children.

Educators and mental health experts have warned about the digital overstimulation of children and its impact on their brain development. Dopamine is a neurotransmitter that transmits signals between nerve cells to regulate pleasure, motivation, learning, and movement. It is often called the “pleasure chemical.” Computer games and social media platforms are designed to trigger micro-releases of dopamine to make users crave constant engagement. Constant engagement results in “dopamine overdoses” that can actually reshape or rewire the brain’s reward system. Overstimulation results in addictive behavior like that seen in drug abuse. Anxiety increases because there is a longing for more stimulation/dopamine but fearing we will never get it. Overexposure desensitizes us to the effects of dopamine, making real-life pleasures feel dull. Multitasking fragments focus, leading to shorter attention spans. With this list of real negative consequences, it is no surprise that parents and educators are changing how children interact with technology.

Interestingly, many of the methods used to restore balance in the brain could be described as a return to “real life.” Digital fasting or taking breaks from digital stimuli helps reset the brain’s chemistry. More than half of states now restrict students' use of mobile phones in schools. Recognizing that exercise triggers a dopamine release, there is an effort to have kids play more and to play outside. Just 45 minutes of daily physical activity has been shown to reduce the risk of depression by 30%. Finally, there is a concerted effort to encourage real-world interactions with friends and family because it was found that personal relationships outperformed digital ones in terms of long-term happiness.

The era of the clunky "grandson in jail" scam call has been replaced by an industrialized "fraud assembly line" powered by generative AI (Scampocalypse, Trends 2023). As we move into late 2025, financial fraud is no longer a game of numbers; it is a game of high-fidelity precision. Cybersecurity Ventures projects that global cybercrime will inflict $10.5 trillion in annual damages by the end of 2025, a figure equivalent to the world’s third-largest economy after the U.S. and China.

While individual data breaches have become routine, their financial weight is reaching new heights. According to IBM’s 2025 Cost of a Data Breach Report, the average cost of a single incident has hit $4.44 million globally, while the average in the United States has surged to a record $10.22 million. These costs are driven by the alarming ease with which automated systems, governing everything from healthcare to banking, can be infiltrated.

The pivot toward synthetic media is the most significant shift in the fraud landscape (Deep Fakes, Trends 2019). Recent data from Entrust reveals that deepfake attempts occurred at a rate of once every five minutes in late 2024, now accounting for 20% of all biometric fraud. Furthermore, the barrier to entry has evaporated: according to KnowBe4, over 82% of phishing emails are now generated by AI, which can craft polished, personalized messages that bypass traditional filters and human suspicion alike.

In this "Zero Trust" (Trends 2025) environment, traditional passwords have become obsolete. Defensive strategies are shifting toward more proactive measures such as Behavioral Geofencing, Digital Honeypots, and Multifactor Authentication.

As AI lowers the cost of a high-quality scam to nearly zero, the burden of proof has shifted. In 2026, the question is no longer "is this message real?" but "how can I prove it isn't fake?"

In a recent Wall Street Journal-NORC poll, 69% of Americans said they believe the American Dream doesn't hold true, the highest in nearly 15 years of surveys. The lack of housing affordability, general disappointment at work, college seemingly offering little value to recent graduates, and social networks heightening feelings of isolation while also revealing to many a lifestyle that is seemingly out of reach, have all likely contributed to this feeling that the American Dream is no longer worth pursuing. Side effects of this malaise include fewer marriages, fewer children, quiet quitting, declining social interaction, and financial and general nihilism.

Work has always brought meaning to Americans, but the precarity that has come with the gig economy and the rise of technology and winner-take-all markets seems to be eroding the perceived link between hard work and eventual prosperity. Rather, it seems for most in this country, work brings feelings of disappointment and hopelessness instead of the affluence and sense of purpose earlier generations unearthed. Without meaning in one’s life, it becomes much easier to treat financial markets like casinos, splurge on luxuries one can’t afford, and make less of an effort at work (Unproductive People, Trends 2023), because after all, what’s the point? With the American Dream seemingly so far out of reach for so many, maybe it’s a rational response. As Peter Thiel recently noted, if one does not have a stake in the system, the system is no longer worth defending, and may even be worth defeating.

Perhaps there is some hope though, as the Pew Research Center’s most recent Religious Landscape Study found that the share of Americans who identify as religiously unaffiliated has plateaued recently after years of growth. If Americans struggle to find meaning at work, perhaps they can find that and a sense of community in religion or other community-based organizations. In an increasingly global world, one would hope American optimism continues to prevail over the challenges of the day.

Physical AI refers to AI systems that understand and interact with the physical world through sensors and actuators, and enable machines such as robots and self-driving cars to learn from these real-time interactions, continuously improving. Physical AI mainly manifests itself through the applications of Robotics, Full Self Driving (FSD) or Autonomous driving (Robotaxi, Trends 2025), Factory Automation and Healthcare (Surgical Robots).

Robotics is split between Humanoid Robots (Humanoids) and Industrial Robots. Though it’s still a bit early for Humanoid Robots, Morgan Stanley Research estimates a total of 1 billion Humanoids by the year 2050 and one Humanoid for every ten people on Earth. China has taken an early lead and dominates the Humanoid landscape, accounting for two-thirds of the global instances of new Humanoid launches. Onshoring or Friend-shoring is a key enabler of Humanoids in the U.S. due to the combination of labor shortage and order of magnitude labor cost savings deploying them. The ecosystem of the Humanoid industry will involve the model developers (a key space with fewer players with scale and moat), the integrators (more commoditized), and the component suppliers.

Autonomous driving is finally hitting an inflection point, with Tesla deploying full self-driving in some major cities and Waymo also increasing its pace of deployment. These companies are taking a different approach: Tesla is relying on software and auto-learning of the AI models, whereas Waymo is relying more on expensive sensors and Lidar. Consumer deployments of Tesla self-driving vehicles are increasing, especially among people who used to detest driving under traffic conditions or long distances. Factory automation is being implemented initially through using small robots for performing automated tasks, such as those in Amazon warehouses or unloading UPS trucks. In the Rise of the Machines, Trends 2016, we talked about how factory automation will automate tasks and replace human labor, benefiting developed countries with cheap capital, unlike developing countries with cheap labor.

Unlike Large Language Models (“LLMs”), which rely on textual data (available online), robots must be trained to perform spatial movements by actual humans. That has been a major roadblock to training them, but new technologies are being developed that allow the robot to be trained partially by watching online human videos (YouTube), alleviating the problem to some extent. And finally, in healthcare, applications are mainly through Surgical Robots, with a current penetration of 10%.

In the past century, the American workforce has transitioned from one based on labor—whether it be agriculture or industrial—to one based on technological knowledge. The advent of Artificial Intelligence portends another transition, one from communicators to builders, from working with bits to working with atoms. AI will diminish the need for interactions between people and technology (Agentic AI, Trends 2025), but until better robotics are developed, those that make and service things will become more valuable. As one colleague noted, AI may tell you what is wrong with your plumbing, but someone still must fix it.

The ramifications of this potential change are large. There are over 70 million knowledge workers in the Unites States with an average salary of $85,000. With a total addressable savings of $6 trillion, this segment of the workforce is ripe for disruption. Younger generations are beginning to question the value of a college or professional degree as AI replaces entry-level jobs in technology, finance, consulting, and law. Instead, the value of trade schools seems to be on the rise as employment in trades like auto mechanics, plumbing, and electrical are thought to be less prone to AI disruption.

The true impact on the American educational system is unknown, but there will be consequences. Colleges will close for lack of implied value and/or students, while those that remain will have to teach skills that AI cannot replicate or displace. Until our educational system retools, the trend appears to favor plumbers over coders, miners over designers, builders over architects, nurses over doctors, and trade schools over four-year colleges.

Agentic AI represents a structural shift from systems that generate outputs to systems that can autonomously plan, decide, and act. As outlined in Agentic AI, Trends 2025, these systems are defined by embedded reasoning and decision frameworks that allow them to identify objectives, sequence tasks, and execute actions within predefined guardrails. Rather than responding solely to prompts, agentic systems operate persistently across workflows—monitoring progress, adapting to new information, and taking initiative—positioning AI as an operational actor rather than just a productivity add-on.

Agentic AI is increasingly moving from experimentation to scaled deployment across enterprise and consumer workflows. By combining reasoning, memory, and tool use, agentic systems increasingly enable end-to-end task execution rather than isolated outputs. Recent product developments underscore this shift. OpenAI’s introduction of in-chat purchasing allows AI agents to move beyond recommendations, and complete transactions directly on a user’s behalf, highlighting the emergence of agentic commerce. Google’s Gemini agent similarly emphasizes goal-driven execution, enabling users to describe an objective in natural language while the system plans and carries out multi-step workflows across connected applications. These examples reflect a broader industry movement toward AI systems that can act, not just advise.

Enterprise adoption further reinforces this trend. Walmart has articulated a strategy centered on embedding agentic capabilities across both internal operations and customer-facing experiences, treating agents as foundational infrastructure rather than standalone tools. By deploying agents across merchant analytics, supply-chain coordination, employee productivity, and customer interactions, Walmart illustrates how large organizations are reorganizing workflows around autonomous systems. At the same time, real-world deployment highlights governance and accountability considerations that echo themes from Zero Trust, Trends 2025. Just as Zero Trust reframed security around continuous verification and least privilege, effective agentic systems require well-designed guardrails, clear decision boundaries, and ongoing oversight. Despite these constraints, Agentic AI’s ability to reshape how work is executed and scaled positions it as one of the most durable and economically meaningful technology trends shaping the next phase of AI adoption.

When we first began tracking successful experiments of quantum in Quantum Computing, Trends 2017, computational operations executed on a very small number of quantum bits. Recently, more companies have been moving from the lab to engineering development and by 2025, quantum computing moved from theory to real-world impact in finance (risk, optimization), drug discovery (molecular simulation), materials science, and logistics (routing), with early returns on investment seen by companies like Ford, J.P. Morgan, and Amazon.

In traditional silicon computers, data is represented in binary bits that are always in one of two states: either a 1 or a 0. However, in a quantum computer each quantum bit, or “qubit,” can represent both a 1 and a 0 at the same time through a principle called superposition. What this means is that a quantum computer can perform multitudes of calculations simultaneously; harnessing millions of qubits could, in a matter of minutes, process data and solve problems that would be impossible for today’s fastest supercomputers.

In 2025, foremost companies announced new quantum chips such as Willow (from Alphabet), Majorana 1 (Microsoft), Nighthawk/Loon (IBM), and Ocelot (Amazon). Many of these developments are focused on efficient error correction and scaling to accelerate development of practical, fault-tolerant quantum computers. Another important step forward was when Google announced the development of the “Quantum Echoes” algorithm that demonstrated verifiable quantum advantage, meaning it could be repeated on another quantum computer. In this case, a benchmark computation that would take a classical supercomputer an estimated 10 septillion (1 followed by 25 zeros) years was completed in under five minutes on the Willow chip.

The implications of large-scale quantum computers will be staggering. With such orders of magnitude of improvement in computing power, expect to see leaps forward in machine learning, artificial intelligence, and simulation modelling. At the same time, quantum computing could pose a threat to traditional encryption security measures that operate on the fundamental assumption that the encryption is too complex to break in a reasonable amount of time, given prevailing computing speeds.

While it’s still early days and many challenges exist in the development of Quantum Computing, we cannot help but imagine the possibilities, which could have a fundamental disruptive impact on the current technology market as we know it.