The global tech giants, NVIDIA, AMD, and Intel, are locked in a high stakes poker game, and the chips on the table are literally the chips that power artificial intelligence. This is not some abstract Silicon Valley drama, this is a very real contest that ripples across the world, influencing everything from autonomous vehicles to, crucially, medical diagnostics. Here in Serbia, we watch this battle with a pragmatic eye, understanding that access to this cutting edge hardware can make or break our own burgeoning AI ambitions, particularly in the health sector.
For years, NVIDIA has been the undisputed king of AI hardware, their Cuda platform a formidable walled garden. Jensen Huang, NVIDIA's CEO, has built an empire on GPUs, and his company's H100 and upcoming Blackwell B200 chips are the gold standard for training large language models and complex AI algorithms. Data from market analysis firm IDC, for instance, showed NVIDIA holding over 80 percent of the AI accelerator market share in 2025, a figure that speaks volumes about their dominance. But AMD and Intel are not simply standing by. They are pouring billions into their own solutions, recognizing that the AI boom is too big to ignore.
AMD, under Lisa Su, has been making significant strides with its Instinct MI300X accelerators. These chips are designed to compete directly with NVIDIA's offerings, boasting impressive memory bandwidth and computational power. Intel, with Pat Gelsinger at the helm, is pushing its Gaudi series, particularly the Gaudi3, which they claim offers competitive performance for specific AI workloads at a potentially lower cost. This competition is fierce, and it is a good thing for everyone, including us. When giants clash, innovation accelerates, and prices, eventually, might become more accessible.
"The Balkans have a different relationship with technology," remarked Dragan Jovanović, head of AI research at the Serbian Institute for Oncology and Radiology, during a recent conference in Novi Sad. "We do not have the endless capital of Silicon Valley, so every investment in hardware must be strategic. We need performance, but we also need value. The current chip war, while intense, offers us options we did not have five years ago." He highlighted how their institute, working with local startups, is exploring AI models for early cancer detection using medical imaging, a computationally intensive task that demands powerful accelerators.
Indeed, Belgrade's tech scene is real, not hype, and its impact on health tech is becoming increasingly evident. Companies like Anari AI, a Serbian startup, are even exploring custom chip design, though that is a long game. More immediately, our researchers and developers are relying on the availability and performance of these commercial chips. A recent study published by the Serbian Academy of Sciences and Arts indicated that AI adoption in Serbian healthcare institutions grew by 15 percent in 2025, largely driven by advancements in diagnostic tools and personalized medicine. This growth, however, is bottlenecked by hardware access.
One of the most critical aspects of this chip war for health applications is the software ecosystem. NVIDIA's Cuda has a massive head start, with a vast library of tools, frameworks, and a deeply entrenched developer community. This makes it easier for researchers and developers to get their AI models up and running quickly. AMD and Intel are working hard to build their own ecosystems, like AMD's ROCm and Intel's oneAPI, but catching up to CUDA's maturity is a monumental task. For a small country like Serbia, where developer resources are precious, choosing a platform often comes down to ease of use and existing community support.
"We are seeing a growing interest from our partners in Europe for more diverse hardware options," stated Jelena Petrović, CEO of 'MedAI Solutions,' a Belgrade based company specializing in AI powered medical diagnostics. "While NVIDIA remains dominant, the benchmarks for AMD's MI300X are compelling, especially for certain types of inferencing workloads. We are actively testing both, trying to find the optimal balance between cost, performance, and integration with our existing software stack." She added that the ability to run models efficiently on diverse hardware could significantly lower the cost of deploying AI in smaller clinics and hospitals, making advanced diagnostics more accessible across the region.
This is not just about raw power, it is about democratizing access to powerful AI. Imagine an AI system that can analyze thousands of patient records and medical images to identify rare diseases, or predict patient outcomes with higher accuracy than human doctors alone. Such systems require immense computational muscle. If the cost of that muscle remains prohibitive, its benefits will be limited to the wealthiest institutions. The competition between NVIDIA, AMD, and Intel, therefore, has direct implications for health equity.
Let's talk about what's actually working. In Serbia, several medical facilities are piloting AI solutions for radiology and pathology. The Clinical Center of Serbia, for example, has been experimenting with NVIDIA-powered AI to assist in early detection of lung nodules from CT scans. Initial results, presented at the Balkan Medical Informatics Conference last year, showed a 12 percent improvement in detection rates for early stage nodules compared to traditional methods. These are tangible benefits, not just theoretical promises.
However, the supply chain remains a concern. The global demand for these high-end AI chips far outstrips supply, leading to long lead times and inflated prices. This affects everyone, but smaller economies feel the pinch more acutely. "Securing allocation for the latest chips is a constant challenge," explained Marko Stojanović, procurement manager for a major Serbian data center provider. "We are often competing with hyperscalers and national AI initiatives from much larger countries. This makes long term planning difficult for our clients, especially those in critical sectors like health." The geopolitical tensions surrounding chip manufacturing, particularly in Taiwan, only add to this uncertainty, a concern that resonates deeply in a region familiar with supply disruptions.
Looking ahead, the chip war will likely intensify. Intel is betting big on its foundry services, aiming to become a major contract manufacturer for other chip designers, which could diversify the supply chain. AMD is pushing hard on software integration and strategic partnerships. NVIDIA, meanwhile, continues to innovate, not just in hardware but also in full stack solutions, from chips to software to entire data center architectures. For us in Serbia, the outcome of this struggle is not just a matter of technological curiosity, it is a fundamental determinant of how quickly and effectively we can integrate life saving AI into our healthcare system.
The stakes are incredibly high. It is not just about market share for these companies, it is about shaping the technological future of nations. For Serbia, a country that has often had to innovate with limited resources, the ability to choose from a competitive landscape of powerful, accessible AI hardware is not a luxury, it is a necessity for advancing public health. We will continue to watch, and more importantly, to build, with whatever tools this global chip war provides. The health of our people depends on it.
For more insights into the global AI chip market, you can refer to reports on Reuters Technology or Bloomberg Technology. You can also find technical deep dives on Ars Technica.
