Artificial-intelligence has been the hottest investing trend of the past few years, and the semiconductors that power the technology have been one of the most lucrative areas.
Each year brings an increasing array of AI chips to the market, so investors need to be on top of innovations.
These are the biggest developments to watch in the AI chips sector in 2026:
Nvidia
The only place to start is with the current sector leader. The company’s annual improvements to its processors set the pace for the AI chips sector. This year isn’t likely to be any different as the company launches its Vera Rubin hardware.
Nvidia CEO Jensen Huang has said that a Vera Rubin AI server, scheduled for launch in the second half of 2026, is set to be 3.3 times faster than the current top-of-the-range Blackwell Ultra equivalent. It will be powered by Rubin graphics-processing units and Vera central-processing units.
One development to watch out for is Nvidia’s release of a new class of GPU called the Rubin CPX, which the company has said is built for “massive-context processing,” enabling it to speedily handle software coding and generating video. It is designed to be cheaper than Nvidia’s other GPUs, which could help it blunt the challenge from rival chip makers.
Before the Vera Rubin launch, watch out for the release of AI models trained on Nvidia’s current-generation Blackwell hardware. Another positive factor for the stock could be the GTC event in March, the annual conference where Nvidia updates its product schedule.
Another thing to monitor is Nvidia’s integration of technology from chip startup Groq. Nvidia is paying $20 billion for a nonexclusive license for Groq’s tech, along with compensation packages for many of the company’s employees who will join Nvidia, The Wall Street Journal reported. The deal is expected to strengthen Nvidia’s position in the market for chips to support inference, or running AI models, which is Groq’s specialty.
Nvidia shares were up 3.1% in early trading on Friday. The stock is up 35% over the past 12 months.
Advanced Micro Devices
It has struggled to make serious inroads into Nvidia’s market share, but 2026 could give Advanced Micro Devices its best shot yet with the launch of its Helios server racks.
Designed to hold 72 of AMD’s MI450 Series GPUs, the Helios racks are scheduled to arrive on the market in the second half of 2026. AMD has lined up Oracle and OpenAI as customers and has built some racks according to specifications from Meta Platforms, which could also lead the social-media company to become a future client.
The factor to monitor will be how smooth the first large-scale deployments of AMD server racks are. Oracle has said it would deploy 50,000 AMD chips starting in the third quarter of 2026 and that it plans to expand the partnership in 2027. The size of any follow-on order will likely depend on next year’s deployment.
AMD shares were up 5.6% in Friday trading, having risen 78% over the past 12 months.
Broadcom
Broadcom is harder to forecast than its peers because it designs custom AI chips for other companies, meaning it depends on their development schedules. But its most important work will continue to be with Alphabet’s Google, working on the search company’s Tensor Processing Units.
The major change is the potential ramping up of sales of TPUs to external clients. AI start-up Anthropic has made two major orders for TPUs, totaling $21 billion, and Meta is also in talks to use the processors, according to The Wall Street Journal.
Confirmation of Meta’s engagement and further details on whether Anthropic is getting performance comparable with Nvidia hardware from its TPUs would be major boosts for Broadcom. Conversely, watch for any indications that Google is bringing more of its design work in house or giving it to rival chip designers such as Taiwan’s MediaTek for threats to Broadcom.
Away from TPUs, Broadcom has also been eager to stress that it is continuing to win other custom chip work with major clients such as Meta and ByteDance, TikTok’s parent. One continuing question is its expected 10-gigawatt deal with OpenAI. Broadcom executives warned in recent earnings that they don’t expect much revenue from the agreement in 2026, but the market would appreciate confirmation of a firm contract.
Broadcom shares were up 4% Friday, and are up 49% over the past 12 months.
Intel
Very much the outsider in the AI chip stakes currently, Intel is hoping to change that this year under the leadership of CEO Lip-Bu Tan.
Intel has said it plans to introduce a new data-center GPU named Crescent Lake to customers in 2026, although a wider launch might have to wait until 2027. Early indications are that it is designed to be on the cheaper end of the market, designed for AI inference—the process of producing answers from models rather than the more intensive training tasks—and optimized for air-cooled servers rather than liquid-cooled infrastructure.
The biggest question hanging over Intel is whether it can convince any significant AI chip maker to manufacture its processors with Intel rather than Taiwan Semiconductor Manufacturing.
A $5 billion investment from Nvidia has raised hopes but so far there has been no indication that the AI chip leader will commit to using Intel’s foundry. Nvidia tested manufacturing its chips using Intel’s production process but stopped moving forward with plans, Reuters reported recently, citing people it didn’t identify. Nvidia and Intel didn’t respond to requests for comments.
Intel shares were up 6.3% in morning trading and have surged 82% in the past 12 months.
