The world’s biggest semiconductor and chipmaker companies recorded over $400 billion in combined sales in 2025 due to the rapid growth of artificial intelligence, making it the largest chip year ever. Analysts and industry executives anticipate that even that milestone will be exceeded next year.
This quick growth, driven by what analysts and CEOs refer to as the “insatiable demand” for processing power, has also led to growing difficulties. In addition to shortages of essential components, chipmakers and their clients are increasingly unsure of how soon AI firms will be able to make profits consistent enough to maintain their rate of chip purchases.
AI chipmakers ramping up for a bigger year in 2026
Hardware designers like Nvidia supply a large portion of the vital infrastructure supporting the AI boom, and Nvidia more than doubled its revenue from the previous year. Although competition has increased, Nvidia has dominated this stage of the market. As the competitive landscape moves from training AI models to effectively running them, Alphabet’s Google and Amazon.com are now posing a more direct threat to Nvidia.
Nvidia and Groq, a startup that creates chips and software to speed up AI inference—the process by which trained models produce responses—signed a $20 billion licensing agreement last week. Tech companies are now competing to provide quicker and more affordable inference as training gains traction. Following the transaction, Bernstein analysts wrote, “Inference workloads are more diversified and may open up new areas for competition.”
Nvidia’s cutting-edge H200 and B200 graphics processing units are still in high demand from customers ranging from data center operators to AI labs. Amazon gains popularity with its Trainium and Inferentia chips, while Google draws users with its custom TPUs. In order to create custom chips, software firms like OpenAI have collaborated with designers like Broadcom. With a new GPU targeted directly at Nvidia’s AI processors, Advanced Micro Devices intends to join the competition in 2026. When Microsoft announced in October that it would double its data-center footprint over the next two years, indicating increased chip demand in 2026, it added to the momentum.
A year of records
Every indication points to yet another year of records. According to Goldman Sachs, Nvidia alone will sell $383 billion worth of GPUs and other hardware in 2026, a 78% increase from the previous year. FactSet analysts expect Nvidia, Intel, Broadcom, AMD, and Qualcomm to generate more than $538 billion in combined sales.
However, 2026 also brings with it previously unheard-of challenges. Data center construction has been slowed by shortages of server components, gas turbines, and electrical transformers. As inference workloads become more “memory-bound” than training, memory chips and specialized silicon substrates continue to be hard to come by. “We’re significantly short of our customers’ needs and it’s going to persist for a while,” Micron Technology’s Sumit Sadana stated.
Although it takes years to build new fabrication plants, Micron, Samsung, and SK Hynix have profited from the crunch through increased investment and higher prices. Investors are also skeptical about AI leaders’ ability to fund the enormous expansion of data centers. Those worries triggered a widespread selloff in AI stocks this fall, highlighting how quickly optimism can turn when growth shows any indication of slowing.
Featured Image Credit: Ivan Chumak; Pexels: Thank you!