Recap 2025: The Year Artificial Intelligence Finally Learnt To Think

If 2023 was the year of chatbots, 2025 will go down in history as the year for 'digital colleagues' - your friendly (or not so friendly) AI helper who can do much more than just write bad poetry.

Indeed, we have crossed a definitive moment in AI evolution - moving away from simply AI-generated content through text or image to becoming a tool with much enhanced thinking capabilities.

In 2025, models broke the thinking barrier, and from being a tool for doing mundane tasks, it became an avenue for creativity, research and productivity.

As models broke the "reasoning barrier" and began executing complex workflows with diminishing oversight, the industry was forced to reimagine everything from software interfaces to cybersecurity architectures.

In 2025, perhaps the biggest shift in generative AI came in the form of 'thinking' before writing.

Almost every foundational model available online now has a 'think' mode, a trend spearheaded by Gemini 3.0 - the holy grail for AI foundational models right now.

Gemini's DeepThink architecture has allowed low-latency chats and high-level reasoning.

This approach proved decisive: on "Humanity's Last Exam," a benchmark designed to stump frontier models, Gemini 3 Pro’s "Deep Think" mode scored 41%, significantly outperforming standard configurations.

Gemini 3.0 wasn't the only flagship Google launched this year. Its Nano Banana Pro model took image generation a step further with superior text rendering - a feature that allows it to generate text within images and thus make infographics.

While Google stole the show with Gemini and Nano Banana, OpenAI's 'code red' response saw them release ChatGPT 5.2, which achieved a perfect 100% on the AIME 2025 math benchmark and 52.9% on the ARC-AGI-2 test of fluid intelligence, significantly outranking competitors.

Meanwhile, Anthropic's Claude 4 cemented its place as the developers' choice thanks to its reliability in long-term computing. The chatbot can indeed sustain focus on tasks for up to seven hours, equivalent to a full corporate workday.

It isn't just the software that has improved in 2025. Hardware to support the complex foundational models have also improved.

Perhaps the most interesting development was Google making strides in the chipmaking game through its Tensor Processing Units (TPUs).

The success of TPUs in not only improved Gemini but also serve as a marketing tool for Google to sell and threaten Nvidia's position, even though the Jensen Huang-run company remain the undisputed leader in this field.

Let's talk about phones now. AI had a massive role to play in most flagship smartphones this year, with Samsung revolutionising the idea of an AI phone by offering various features, while Apple fell behind.

Most phones in 2025 also launched with NPUs - Neutral Processing Units designed to store a neural network that powers the generative AI in a smartphone.

The AI revolution in 2025 did not just help individuals; it also redefined the corporate landscape.

From offering passive chatbot solutions, AI has helped corporate entities get hands-on access to a 'digital colleague' - an agentic shift that allowed enterprises to delegate complex, multi-step workflows.

These include refactoring software code, resolving billing disputes, understand data, with minimal human oversight.

This had a telling impact on the labour force as companies such as Amazon, Verizon, IBM and Starbucks all orchestrated major layoffs this year.

From a consumer perspective, the onset of the latest Gemini or ChatGPT models has greatly helped individuals as well as corporations.

But the growing demand for AI and data centres has had a significant impact on global supply chains, particularly in relation to chips.

The AI boom has indeed coincided with global RAM and SSD shortages. NDTV Profit reported earlier that companies producing these RAMs and SSDs are mass-producing hardware for big companies and data centres.

As a result, RAM prices have spiked up to 300% in the last two months, thereby significantly hurting consumers. Companies like Micron have even shifted away from the consumer tech business, leading to fears that this supply glut could remain for the foreseeable future.