A to Z of technology in 2025: A for Agentic AI, B for Blackwell, C for Compute crisis, and more

Twenty—six letters, twelve months, and more twists than a season finale. That was 2025 in technology. Chinese AI labs surprised Silicon Valley. The thinnest iPhone ever made couldn't find buyers. Humanoid robots began folding laundry. A bankruptcy court took one of the most loved names in consumer robotics. Elon Musk bounced between politics and tech empires. Mark Zuckerberg wrote record paycheques to fuel AI goals, while his metaverse dreams faded. A meme about tiny bananas somehow captured the spirit of AI's creative boom. There were shockwaves from new players, familiar faces making big changes, and technologies that went from niche to essential. Here’s how the year unfolded, one letter at a time.The shift was subtle at first, then suddenly everywhere. Generative AI—the technology that could write emails and create images—began evolving into something more autonomous and more proactive. Enter agentic AI: artificial intelligence that doesn’t just respond to prompts, but anticipates needs, takes actions, working as a companion rather than a tool.Everyone pivoted hard toward this vision of agentic AI, positioning their chatbots not as a mere chatbot but as a digital companion. Microsoft pushed Copilot as a work partner, Google reimagined Gemini as a helper, and OpenAI hinted at agents capable of handling complex tasks across multiple platforms. The promise was simple: AI that works for you, not just with you.Nvidia’s Blackwell architecture turned out to be the dark horse nobody saw coming—or rather, one everyone underestimated. What began as the next iteration of Nvidia’s GPU lineup quickly became the backbone of the AI revolution. Blackwell chips found their way into everything: powering high—end gaming PCs with unprecedented performance, driving AI workstations for creative professionals, and, most importantly, serving as the engine room of massive AI data centres worldwide. The architecture’s ability to handle both training and inference workloads with remarkable efficiency made it indispensable. As tech giants scrambled to build out their AI infrastructure, Blackwell was what they were scrambling for. And as Nvidia’s market cap soared, Jensen Huang’s leather jacket somehow became even more iconic.All those AI breakthroughs came at a cost—literally. The already tight supply of computing resources turned into an all—out crisis. GPU prices had already surged earlier, but 2025 saw the crunch spread to storage and memory in ways few anticipated. Big Tech’s race to build AI data centres created unprecedented demand for virtually every component imaginable. Memory giants like Samsung, SK Hynix, and Micron faced an impossible choice: continue supplying consumer markets or divert production to tech giants waving massive purchase orders. They chose the latter. Memory and storage prices hit record highs as supply for regular customers dried up. The message was clear: in the age of AI, if you’re not building a data centre, you’re at the back of the line.China's DeepSeek didn't just enter the AI race—it shocked everyone watching. The company emerged seemingly from nowhere with models that rivalled, and in some cases outperformed, Western counterparts at a fraction of the reported cost. Their technical papers suggested training methods that were more efficient, their architecture choices that were unconventional yet effective. Silicon Valley had grown comfortable assuming the AI lead was secure; DeepSeek shattered that assumption. The implications rippled across the industry: Could China leapfrog in AI development? Were Western companies overspending on compute? The geopolitical dimensions of artificial intelligence suddenly felt very real, very urgent.What a year for Elon. Musk went from tech entrepreneur to playing politician, advising on government efficiency and policy with DOGE, only to pivot back to his tech empire with renewed vigour. Between steering Tesla through production challenges, pushing SpaceX toward Mars, and step up xAI's ambitions, he somehow found time to be everywhere at once. Whether you admired his audacity or questioned his judgment, ignoring Musk in 2025 was impossible. He also grabbed himself the biggest paycheque for CEO worth $700 billion.2025 was the year when foldables stopped being a novelty and started getting serious. Samsung and Huawei led the mission, but Chinese manufacturers too pushed boundaries with designs that were increasingly durable and practical. The crease problem that plagued early foldables became less pronounced, and battery technology finally caught up to power these expansive displays. And if you though bi—folding devices are the future, the year saw tri—fold devices move from concept to commercial reality, with devices that could morph between phone, tablet, and even laptop—like experiences.It was another year of disappointment for Grand Theft Auto fans. This was the year that was supposed to be the year of GTA VI. But as months passed, that release window kept sliding, Rockstar dropped the bomb: the GTA VI is being pushed to 2026. The game have already become soothing of a legend even before its launching, winning not one but two awards at The Game Awards. But, the wait continues, the hype persists, and the question remain: would 2026 finally be the year?This was the year when Tesla’s Optimus made its first real—world appearances, serving cocktails. China, too, entered the humanoid robotics race with remarkable speed and scale—showing robots that could do all Optimus could but at price points that would make Western competitors nervous. All of this made it clear that good old vision of humanoid robots, doing our chores, manufacturing at factories, and even doing the waiting at restraints now suddenly felt less like science fiction and more like a near—future reality.Apple’s iPhone Air finally arrived this year, and it was everything design enthusiasts hoped for: impossibly thin, gorgeously crafted, a marvel of engineering. At mere 5.5mm thick, the Air changed people’s expectations of what a phone could feel like. But consumers didn’t bite. The compromises required for that thinness— a smaller battery, no physical SIM slot, and limited camera capabilities—proved too much for most buyers. People admired it from afar, but sales so far this year have been modest.The layoffs that began during the pandemic never really stopped, but it only got worse in 2025. Amazon, Microsoft, Salesforce, and IBM—to name a few—all announced job cuts with tens of thousands being sacked, and the reason behind them all was one: AI. Roles in customer service, data entry, content moderation, and even some coding positions evaporated as AI tools proved capable of doing the work faster and cheaper.Companies framed it as “transformation” and “optimisation,” but the human cost was real. The conversation around AI’s impact on employment shifted from theoretical to urgent, as white—collar workers began facing the same kind of disruption that earlier technological waves had brought to manufacturing.The phrase took on a far grimmer meaning. In 2025, a few incidents in took place that involved AI systems making decisions that resulted in deaths. Autonomous vehicles were involved in fatal accidents where AI failed, medical systems delivering incorrect diagnoses that led to patient deaths, and automated defense systems misidentifying targets. Each case triggered intense scrutiny.The technology wasn’t malicious, but it was consequential in ways abstract debates about AI safety had never fully captured. Regulators scrambled to respond. Tech companies issued familiar statements about improving safeguards. Yet the incidents exposed an uncomfortable truth: AI had reached a point where its failures could be lethal, and society was unprepared for that responsibility.Out of nowhere, Apple shook up iOS this year - version 26 arrived with something called "Liquid Glass." Not quite flat, not quite real glass, the new look added slick movement and texture to each screen. Layers seemed to hover, shifting slightly when you tilted your phone. It pulled ideas from older styles like soft edges and frosted panels, then pushed them forward with shimmer and depth. Some said it wasn’t bold enough, just small steps disguised as leaps. Yet others found joy in the polish, calling their phones sharper, smoother, somehow alive again.Mark Zuckerberg went all—in on AI in 2025, writing cheques that made even Silicon Valley veterans do double—takes. Zuckerberg turned Meta’s AI division into a magnet for top talent, with compensation packages reaching unprecedented levels. He personally recruited researchers from rivals, offering not just money, but the resources and freedom to push boundaries.Meta’s AI models improved rapidly, its infrastructure expanded at scale, and Llama grew increasingly competitive with closed models. The metaverse, however, told a different story. The Quest headsets saw only incremental improvements, while the company leaned more heavily than ever into smart glasses. Yet Zuckerberg’s vision of immersive digital worlds where people would live and work still failed to materialise.For a company that once changed its name to signal that ambition, cutting 30% of the budget for the very team building Zuckerberg’s metaverse dream was telling. The pivot was clear—even if it was never openly acknowledged that the metaverse vision was fading.The Nano Banana phenomenon took over the internet, turning AI image generation into a full—blown cultural moment. What began as an internet trend revealed something more serious: how far image models had come in understanding scale, context, and visual coherence. The phenomenon spread rapidly, spawning endless images and becoming shorthand for AI’s creative leap. Everyone played along, artists squinted, and for a brief stretch, timelines were filled with AI—generated images. It was playful and ridiculous—but it also showed how accessible, intuitive, and genuinely advanced AI tools had become.2025 was the year the internet kept breaking. AWS went down for hours. Microsoft’s Azure suffered multiple major disruptions. Google’s infrastructure showed cracks. Cloudfare, a not so staple name, but have many websites up and running, also suffered a major outage, taking countless services with it. Each outage exposed just how fragile our cloud—dependent world had become.Discussions stopped on social media, productivity tools went offline, and smart home devices stopped responding. The causes have been different but the outcome was always the same: panic.The incidents reignited debates around over—centralisation and the need for stronger redundancy. As usual, the conversation was loud—but little fundamentally changed.Perplexity and its CEO Aravind Srinivas became impossible to ignore, generating headlines for both impressive achievements and controversial moves. The AI—powered answer engine gained serious traction as a Google alternative, offering cited responses and conversational search. Srinivas positioned the company as David against Google's Goliath, attracting significant funding and user growth. Then came the audacious move: attempting to acquire Google Chrome when regulatory pressure mounted. The bid was ultimately unsuccessful, but it demonstrated Perplexity's ambitions.Controversies around content sourcing and attribution followed, with publishers questioning how Perplexity used their content. Love it or hate it, Perplexity had made itself central to conversations about search's future.Quantum computing graduated from promising research to delivering actual breakthroughs. IBM, Google, and startups like IonQ announced quantum systems solving problems classical computers couldn't touch—drug discovery simulations, optimisation challenges, and cryptographic calculations that would take conventional systems millennia. The number of stable qubits increased dramatically, error correction improved, and practical applications emerged. It wasn't yet the quantum revolution that would break all encryption or revolutionise every industry, but the trajectory became clear. Quantum computing moved from "someday technology" to "in the next few years," with governments and corporations investing billions to lead the race.Bankruptcy hit iRobot in 2025, the company behind the well—known Roomba vacuum. At first glance, the story seemed predictable — lower—priced rivals appeared fast, while big tech players stepped in. Still, what truly hurt wasn’t competition at all.Stuck without a clear path forward, iRobot found itself frozen after Amazon walked away. Not even the fame of Roomba could save it — brand power didn’t fix broken fundamentals. Plans stalled, options narrowed. What looked solid on the outside cracked under pressure. Future moves? Too slow. Cash flow problems piled up while decisions waited.The Roomba label might survive under new ownership, yet iRobot itself, once its own driving force, had already faded into history.Something about OpenAI’s Sora clicked the moment people saw it move. Like Nano Banana once shocked image makers, this made video feel suddenly possible. A single line of text could now spin into scenes that held together, looked sharp, felt alive. Not just rough sketches—these played like real footage. Artists poked at its edges, brands watched closely, forums filled with tests and twists. What used to take teams and gear now emerged from words alone. Entire workflows blinked into question. Worries grew heavier when fake videos spread faster than facts could catch up. Yet people kept finding clever ways to make art out of code. Motion began flowing from machines—scenes dreamed by algorithms, shaped into something almost alive. Reality blurred a little more each time someone pressed play.TikTok's 2025 was a rollercoaster of bans, lifelines, and eventual sale. The app was banned, then granted temporary reprieves, then banned again as political and security concerns ping—ponged through Washington. After months of uncertainty affecting millions of American users and creators, a deal finally emerged: Oracle and a consortium of investors acquired TikTok's US operations, maintaining the app's functionality while satisfying government concerns about data and Chinese ownership. Out of nowhere, tensions across the Atlantic heated up again. Not long after one ruling settled, Brussels moved on another front — Google caught fresh legal heat. Meanwhile, Meta ended up paying heavy penalties over how it handles personal information. Apple found itself squeezed by scrutiny around its control of app distribution. Over at Amazon, officials poked deeper into who really sets the rules on its platformA single situation meant staggering penalties—sometimes crossing into billions — alongside deep shifts in how firms operate. Fines on this scale sparked pushback from Silicon Valley giants who saw Brussels as meddling too far, accusing it of choking homegrown progress. Officials across Europe stood firm, framing their actions less as attacks and more as necessary shields for everyday users and fair marketsA clash like this one wasn't just about now—it pointed at long—standing friction over who controls online spaces, how tech firms pay taxes, yet also which laws should shape web activity. Still, each party stood firm.Out of nowhere, coding stopped being only lines and rules. A different kind of energy took over—some called it vibe coding. This way of building software leaned on gut feeling, smooth motion, deep focus. Beauty in the act mattered just as much as what the code did. It felt good when things clicked—not because they worked, but how they came together. People talked about their desks, lights, music, even coffee routines.These little habits helped them land in the zone before typing began. Smart tools stepped in to manage boring parts automatically. That left room for humans to explore, play, enjoy the making. Some rolled their eyes, said it was self—indulgent nonsense. Yet those doing it believed joy shaped better outcomes. Craftsmanship showed up not in slogans, but in clean, thoughtful output. The practice grew quiet, steady, without needing attention. What started as personal preference slowly changed daily rhythms across teams. Not faster, not flashier—but deeper, richer, somehow more alive.Even as artificial intelligence pushed wearables forward, they kept circling a missing breakthrough—something vital enough to shift them from optional to necessary. Devices like Meta's smart glasses edged ahead, while fitness bands began tracking bodily signals with sharper precision. Watches packed on new tools too, layering functions one after another. Smarter software brought deeper readings of body patterns, guessed what users might need next, offered help based on surroundings. Still, progress felt slow, step by narrow step. They stayed on the edges of daily life, useful but never central. Wearing smart gadgets didn’t turn out quite like they said it would. Helpful? Yes, better than what came before. Yet that big change—expected by many—never really arrived. Just hovered there, close but never real.Just like Elon Musk, his AI company—xAI—had a colossal year filled with drama, ambition, and controversy. The company built one of the world's largest AI training facilities, deployed it faster than anyone thought possible, and pushed Grok—xAI's chatbot—as a politically unfiltered alternative to "woke" AI. The facility's energy consumption and environmental impact drew criticism, while Grok's deliberately provocative tone generated both enthusiasm and concern. Musk positioned xAI as the counter to AI "censorship," though people noted his own platform moderation inconsistent, at times favouring Musk. The company raised massive funding, hired aggressively, and demonstrated that even in a crowded AI field, Musk's brand of chaos could carve out space.A surprise shift unfolded in the world of artificial intelligence. From deep within Meta, Yann LeCun—Turing laureate, widely seen as a pioneer—stepped away, launching his own startup. What followed stirred immediate debate: he claimed large language models aren’t living up to their hype. Behind closed doors, experts had whispered doubts; now one of them spoke aloud. Transformers, he said, have dominated progress too long, crowding out fresh ideas. His bet? That breakthroughs lie outside the mainstream path everyone else is racing down. This wasn’t just another opinion - it carried weight because of who voiced it. Reactions spread fast among labs and research teams. Maybe what comes next won’t resemble ChatGPT at all.Tech valuations reached stratospheric heights in 2025, with AI companies commanding eye—watering multiples that made even dot—com bubble veterans nervous. Startups with minimal revenue but AI in their pitch decks raised hundreds of millions at billion—dollar valuations. Nvidia now wears the crown as the planet’s priciest business, crossing $4 trillion in value. Hype pulsed through each announcement--earnings reports name-dropped artificial intelligence constantly, new gadgets claimed smart tech at their core, entire brands reshaped themselves around algorithm-driven promises. Still, quiet questions began spreading about how long it could last.