It was no April Fool’s joke – Meta informed its employees that it would be cutting 8,000 jobs in May. It also abandoned plans to fill several thousand open positions, as it intends to invest $135 billion into artificial intelligence projects. In April, in just 10 days, Nvidia’s shares jumped by more than 18 per cent, driven by a significant increase in demand for its AI chips from Meta, Amazon, Google, and Microsoft.
“Those four companies recently announced that they will collectively invest over $725 billion this year, which is nearly double their investments from last year. These are funded from the companies’ current revenues and, given their scale, they must be justified to investors – as a rule, through the announcement of future cost savings,” explains Borivoje Nikolić, who, as he tells Radar, found himself at Berkeley “with a bit of luck”, where he has now spent 27 years. Today, he is one of our leading experts in chip design, a full professor of electrical engineering and computer sciences at Berkeley, and holder of a prestigious engineering chair established by National Semiconductor. In addition, Nikolić co-manages two laboratories at Berkeley: the Wireless Research Center and the Specialized Computing Systems Ecosystems.
“Leading hyperscalers are investing enormous resources into the development of artificial intelligence, which requires a powerful computing infrastructure. The situation somewhat resembles the arms race of the past; participants fear for their own survival if they fall behind the competition. At the same time, to avoid alarming investors, companies are resorting to layoffs, expecting that increased employee productivity will enable continuous revenue growth. Furthermore, it is generally believed that large companies hired too many people during the post-pandemic boom, so a certain amount of rationalisation was expected anyway. Meta is probably in the most delicate situation and must invest particularly rationally because, unlike Amazon, Google, and Microsoft, it does not have business branches that can lease surplus capacity to others,” Nikolić explains to Radar.
Yet, there are companies today that are living their finest hour?
Primarily hardware companies, like Nvidia, are experiencing powerful growth precisely on the back of these investments, as hundreds of billions of dollars are flowing directly to them. It is particularly significant that the stock value of the world’s most valuable company is rising. Sudden, frequent spikes in share prices are not uncommon for smaller firms, but are rarely seen among the largest ones. Further growth in hardware demand is expected, and this is driving up share values. We do not know when a new equilibrium will be established in the market. Currently, all hardware shares, especially those of memory companies, are rising rapidly on expectations of their future revenue growth.
According to some estimates, the semiconductor industry will face a shortage of one million workers by 2030?
The semiconductor industry is exceptionally large and encompasses various specialities, from design and manufacturing all the way to servicing finished products. On one hand, new chip factories are opening, each employing two to three thousand predominantly highly qualified workers. These factories rely on supply chains for raw materials and semi-finished products, which must also grow in tandem with rising demand. On the other hand, chip design represents a significantly more diverse branch of the industry, given that different chips are developed for completely different purposes – modern mobile phones contain dozens of different chips, which differ fundamentally from those installed in cars or AI servers. This segment of the industry employs more people than the factories and their suppliers themselves. I cannot claim with certainty that there will be a shortage of a million workers across all these domains, but the shortage of engineers is currently very noticeable. Many believe that jobs in semiconductors and hardware in general will change more slowly with the development of artificial intelligence, which is fueling student interest in these fields.
Alongside representatives from Nvidia, Microsoft, Intel, and AMD, you participated in a panel discussion on the impact of AI on chip design, organised by Synopsys. What impact does the development of chip design and new processor architectures have on the explosive growth of artificial intelligence?
The development of more powerful chips was a key prerequisite for the rise of artificial intelligence. Repurposing graphics cards towards the parallel processing of matrices and vectors – necessary for the efficient execution of AI algorithms – enabled this accelerated development happening now. Every month we witness new achievements that improve specific parts of the overall process, meaning these very technological shifts are essential for further development. For technology to keep advancing, hardware must simultaneously become both faster and more energy-efficient.
There is a global hunger for chips. How can it be satisfied?
Chip production is a lengthy process; it begins with design and ends with fabrication and packaging in specialised enclosures. Depending on the complexity of the product, this process takes a year or longer. Currently, bottlenecks are everywhere: there are not enough design engineers to speed up the process of designing new chips, while, on the other hand, the manufacturing capacities of factories (TSMC, Intel, GlobalFoundries, Samsung…) are also limited. Launching a single modern factory line for advanced technologies costs more than $20 billion, and the construction itself takes two to three years. It is very difficult to reliably predict demand so far in advance. Investing tens of billions of dollars into capacities that might not run at full capacity later would be a poor investment. Even now, we see that manufacturers like TSMC are hesitant to make excessive investments. They are opening new lines, but not at the pace demanded by the rest of the industry. Ultimately, with a limited supply, TSMC and others are in a position to dictate higher prices.
Japan, South Korea, and Taiwan are offering tax incentives, while Brazil, Mexico, Vietnam, Kenya, Hungary, and Poland are entering the semiconductor arena. Serbia is also offering various public and secret incentives?
I believe that the state should intervene only when it is truly necessary. Companies will, of course, gladly accept tax incentives when considering expanding into a new environment. However, the state’s role in creating conditions where businesses can develop smoothly is far more important. In Serbia, there is a whole range of factors that make drawing in industry difficult, which are almost impossible to compensate for through tax incentives alone. The lack of a highly qualified workforce is probably among the most significant. The best long-term investment would be the modernisation of universities, which could become the pillar of long-term development in this field. But, as with all long-term investments, the fruits of such a move would not be felt immediately.
For years, Serbia has relied on the IT outsourcing model. What is the alternative?
The primary reason for establishing such organisations was to enable cheaper operations, with 24-hour coverage from different locations around the world. The introduction of artificial intelligence into the software system development process is now changing the internal dynamics of these organisations. It is far from the case that software development will disappear as a human activity, or that the IT service model will vanish entirely, but they will inevitably change. The application of AI in programming, testing, optimisation, and system maintenance reduces the need for human resources on any individual project. However, that same principle opens up the possibility to quickly and efficiently develop many systems with significantly smaller teams that previously weren’t even considered – a phenomenon known as Jevons’ paradox. I believe the number of people working with these technologies will grow, but their knowledge and skills will have to evolve at the same time.
A considerable number of well-known global chip design companies operate in Serbia – Elsys, Veriest, Infineon, Chip Globe, Next Silicon, Tenstorrent, and AMD. What are the effects and scope?
It is true that many multinational companies have opened their subsidiaries in Serbia, which is indeed of great significance. Verification represents an exceptionally important phase in chip development and must never be underestimated. Judging by my admittedly limited contact with these and similar companies, further recruitment in Serbia is primarily limited by the shortage of qualified engineers.
You work at Berkeley, one of the world’s prestigious universities. What are you currently working on, and what can we expect in the future?
My research group is predominantly engaged in chip design for various purposes. Some of our current projects relate to the implementation of chips that accelerate artificial intelligence. In parallel, we are working on developing AI that will assist us in the design process itself. It is precisely in this second direction that we expect very interesting results, where we will demonstrate how an engineer, with the help of AI, can very quickly design analogue electronic components like amplifiers, or how AI can contribute to the development of computer architecture.
The technology world, and indeed the world at large, is under a latent threat of power concentration, where five or six companies control global intelligence, and thus the future?
Large tech companies already control a vast amount of natural resources. Each of these companies feels the need to invest more than its competition to avoid dropping out of the race. The reason is the fear that if someone else develops a more powerful artificial intelligence before them, their future business prospects will be seriously jeopardised. This applies not only to companies, after all, but also to states and governments. However, as long as there is healthy competition – both among companies and among states – the balance will be maintained somehow.
You say that upon leaving for your doctoral studies in the US in 1996, you realised you had a very solid foundational knowledge of mathematics, physics, electronics, automation, telecommunications, and computers, but lacked knowledge in applied areas, primarily practical electronics and computing. If that was not insurmountable for a student back then, how far behind are we today?
I haven’t had the opportunity to look at current curricula, but based on occasional contact with students who completed their undergraduate studies at the Faculty of Electrical Engineering (ETF), one gets the impression that practical knowledge, unfortunately, still belongs to the previous century. The reasons for this are numerous, lying primarily in staff policy and the criteria for faculty promotion. Unlike the last century, when a large number of ETF professors had the opportunity to pursue professional development at leading global institutions, today such a practice is rare. In the field of electronics, for example, I very rarely encounter professors or students from Serbia at leading international conferences. I haven’t met anyone from the Belgrade ETF in decades. The exceptions are colleagues from Niš, the late Prof. Ninoslav Stojadinović, and from Kragujevac, Prof. Vladimir Milovanović. The consequence of this practice is the isolation of the Serbian scientific and technical community and further stagnation.
In Serbia, at least according to announcements, at least one chip fabrication factory should have been well underway. First Intel’s, then one from the Mubadala fund from the Emirates…
I was informed about the talks with Mubadala; I even participated in a consultation organised by the Academy of Engineering Sciences of Serbia (AINS). Even back then, it was obvious that we lacked the personnel ready for such a factory, as well as any realistic possibility of training them within a reasonable timeframe. On that occasion, I appealed for universities to get seriously engaged and for teaching to be modernised so we could seize the next opportunity, but unfortunately, it seems nothing concrete has moved forward. With relatively small investments, we could catch up in just a few years, but there is little point in just talking about it; we need to finally start modernising.
Artificial intelligence in Serbia is talked about in the way that shortening mandatory military service used to be. We have an AI Development Strategy until 2030, and we are preparing for the “Leap into the Future 2027″…
I am genuinely not familiar with those initiatives. In the past, government delegations, as well as representatives from state companies and universities, regularly visited Berkeley and Silicon Valley, but lately, there have been almost no such visits, so I can only find out a bit about it over the internet.
To what extent do top researchers from the academic community resist the calls and offers of multiple-times higher salaries, and how does their drain affect higher education institutions in the US?
Working at universities in the US is exceptionally demanding, but the rewards at leading academic institutions go far beyond the monetary. It is a great pleasure to work with brilliant young people who develop ideas that no one before them had even thought of. In addition, most universities allow faculty members to consult for companies once a week or to take a leave of absence for a year or two to work in a firm during that time. Many of us are also engaged as advisors in smaller or larger startup companies. Of course, some of my colleagues decided at a certain point to leave their academic careers and, after an initial leave of absence, permanently move to a company they founded themselves or accept a position in a larger firm. I believe this most often happens due to a passionate interest in a specific project and the resources those companies can provide, though financial reasons sometimes play a role too.
The EU plans to subsidise the chip industry with €43 billion by 2030, the US with around $52 billion, while China will pump $150 billion into state R&D firms. Is China “running away” from America in chip design?
Semiconductor design and manufacturing are considered strategic resources in many countries, and that is precisely why we are witnessing such large investments and state subsidies. However, one should bear in mind that state funding makes up only a small fraction of total investment. The largest investments actually come from the end-users of the technology themselves. Hyperscale giants alone will spend hundreds of billions of dollars this year, mostly on buying chips. Total end-user spending significantly exceeds $1,000 billion this year. A significant portion of that money is then reinvested into further development, primarily in the US. China is advancing rapidly, both thanks to the development of its own IT industry and due to lavish state investments and subsidies. In addition, sanctions are further forcing them to rely more and more on their own strengths. Still, the US remains convincingly ahead of everyone else. In the field of manufacturing, the situation is different, and there Taiwan is currently ahead of everyone.
(Radar, 26.06.2026)
https://radar.nova.rs/ekonomija/ulaganje-srbije-u-obrazovanje/