Category: Technology | Published: 2026-06-09
Most people never think about the chip inside the device they are using. They think about the app, the website, the AI assistant responding to their question. The hardware underneath is invisible, taken for granted in the way that electricity or running water is taken for granted, noticed only when it is not there.
That invisibility is changing fast. AI chips, the specialised processors that make modern artificial intelligence possible, have become one of the most strategically important commodities on earth. Governments are restricting their export. Nations are spending billions to secure domestic manufacturing. Alliances are being formed specifically around who gets access to them and who does not. What was once a technical back-room conversation between engineers has become a front-page geopolitical story.
Here is why.
The Numbers Are Extraordinary
Global semiconductor sales reached a record 791.7 billion dollars in 2025, a 25.6 per cent increase year-on-year and the fastest growth the industry has seen in years. Projections for 2026 put the market close to one trillion dollars as AI infrastructure spending continues to accelerate.
The concentration of value within that number is striking. High-value AI chips now drive roughly half of total semiconductor revenue, yet they represent less than 0.2 per cent of total unit volume. A tiny number of extraordinarily expensive, extraordinarily complex components are pulling the entire industry forward. The combined market capitalisation of the top ten chip companies globally stood at 9.5 trillion dollars by the end of 2025, up 46 per cent from 6.5 trillion dollars just twelve months earlier.
Those are the kinds of numbers that attract government attention.
Rethinking How Chips Are Built
To understand why AI chip technology has become so consequential, it helps to understand what is happening at the physical limits of semiconductor engineering.
For decades, the industry relied on Moore's Law, the observation that the number of transistors on a chip roughly doubles every two years. The method was straightforward: make the transistors smaller so you can fit more of them in the same space. That worked for a remarkably long time. It is now running into fundamental physical constraints as transistors approach the size of individual atoms.
IBM recently demonstrated one answer to that problem with a chip design it describes as the world's first sub-1 nanometre semiconductor technology. Its 0.7 nanometre design can pack close to 100 billion transistors onto a chip the size of a fingernail, delivering up to 50 per cent higher performance or 70 per cent greater energy efficiency compared with its earlier 2 nanometre technology.
The difference in approach is significant. Rather than continuing to push transistors closer together horizontally, IBM's NanoStack architecture builds upwards, stacking multiple layers of transistors vertically to achieve far greater density. As Jay Gambetta of IBM Research put it, the goal is not just smaller transistors but reinventing how chips are built entirely.
This architectural shift matters because it opens up a new dimension for AI chip performance improvements at precisely the moment when the demand for them is greatest.
Why AI Changed the Semiconductor Industry
The rise of large language models and other AI systems has transformed what the semiconductor industry is being asked to produce. Training and running modern AI requires processing power on a scale that has no real precedent in consumer or business computing history.
AI chips designed for these workloads are not simply faster versions of the processors in a laptop or smartphone. They are architecturally different, optimised for the specific types of mathematical operations that AI models rely on, and capable of running many thousands of parallel calculations simultaneously.
Energy efficiency has become as important as raw performance in this context. AI data centres consume enormous amounts of electricity. Every improvement in how much processing a chip can do per watt of power consumed has a direct impact on operating costs and on the environmental footprint of AI infrastructure. Recent AI chip announcements have focused as heavily on power consumption as on headline performance figures.
Companies Building Their Own AI Chips
Perhaps the clearest sign of how central AI chip technology has become is that the largest AI companies are no longer content to buy chips off the shelf. They are designing their own.
OpenAI recently unveiled Jalapeño, its first custom inference processor, developed in partnership with Broadcom. The chip is designed specifically for the workloads that power large language models rather than attempting to be a general-purpose solution. OpenAI President Greg Brockman described it as part of the company's long-term strategy to make compute more abundant, resulting in AI that is faster, more reliable, and more affordable.
OpenAI is not alone. Google has been running its own custom AI chips, called Tensor Processing Units, for years. Amazon has Trainium and Inferentia. Microsoft has the Maia chip series. Each of these companies has concluded that designing silicon specifically for its own AI workloads gives it meaningful advantages in performance, efficiency, and cost that buying commercially available AI chips cannot match.
For NVIDIA, which currently dominates the commercial AI chip market, this trend is both a validation of how important the market has become and a long-term competitive challenge.
When AI Chips Became a National Security Issue
The clearest measure of how strategically important AI chips have become is the extent to which governments have intervened to control who can access them.
The United States has spent several years building an increasingly complex set of export restrictions targeting China's ability to acquire advanced AI chip technology. The escalation has been relentless. In 2022, export bans were placed on NVIDIA's A100 and H100 processors for sales to China. In 2023, those restrictions were expanded. In early 2025, the US Bureau of Industry and Security introduced the AI Diffusion Framework, establishing quotas on advanced computing exports to countries outside the closest US allies. In April 2025, the H20 chip was banned from China entirely.
The restrictions have continued to tighten. By May 2026, new rules were targeting NVIDIA's most sophisticated processors, including the entire Blackwell series, requiring export licences for sales to Chinese and Macau entities. At the end of May 2026, the BIS closed a loophole that had allowed Chinese-parent-owned offshore companies to receive shipments without licences. The message from Washington has been consistent: advanced AI chips are not ordinary commercial products, and the US intends to control who gets them.
The Alliance Being Built Around AI Chip Access
The United States has also moved beyond unilateral restrictions to build a coordinated allied approach to semiconductor supply chains. The initiative is called Pax Silica, and it is designed to strengthen cooperation between allied nations on AI chip manufacturing, critical minerals, and advanced production capacity.
The European Union's decision to join Pax Silica is notable. The EU had been pursuing its own agenda of technological sovereignty, investing significantly in domestic semiconductor manufacturing through the European Chips Act. Joining a US-led initiative represents a recognition of how difficult it is to secure advanced AI chip supply chains independently, even for an economic bloc the size of the EU.
The European Commission summed up the reasoning plainly: as AI reshapes economies and societies, secure and resilient silicon supply chains are more important than ever. The countries and alliances that control semiconductor research, manufacturing capacity, and supply chain access are increasingly seen as holding significant strategic advantages over those that do not.
What This Means for Businesses
For most businesses, the global AI chip race plays out at a level of abstraction that feels remote from daily operations. But the effects are direct and practical.
Every cloud service, AI platform, cyber security product, and business application ultimately depends on the availability of increasingly powerful and efficient AI chips. When supply is constrained, whether by manufacturing capacity, export restrictions, or geopolitical disruption, the costs of cloud computing and AI services are affected. When new AI chip generations arrive, the capabilities of the platforms businesses rely on improve, and new things become possible.
The pace at which AI tools are evolving, and the pace at which businesses can realistically adopt and benefit from them, is shaped significantly by the semiconductor industry operating in the background.
If you are thinking about how AI tools can work for your business today, rather than waiting for the geopolitics to resolve, our AI Consultancy page is a good place to start.