The last three years have been marked by the rapid development and deployment of new artificial intelligence technology. From the release of ChatGPT in late 2022 to today, large language models have seen significant improvements, driven by innovative training and model techniques, larger models and context windows, and advances in hardware capabilities. But based on comments and commitments from some of the largest AI companies in the world, there's still a long way to go.
In 2026, we can expect many more advances in generative AI. Here are three of the biggest trends to watch, along with how investors can position their portfolios to capitalize on the continued opportunities in artificial intelligence.
Image source: Getty Images.
1. Custom AI accelerators will take market share
The four biggest hyperscalers have been developing their own custom silicon for years. Alphabet (GOOG +1.82%) (GOOGL +2.20%) has utilized its Tensor Processing Units (TPUs) for AI development for over a decade, and it's been offering them through Google Cloud since 2018. Likewise, Amazon launched its Inferentia chip (for inference) in 2019 and Trainium (for AI training) in 2021. Microsoft and Meta Platforms (META -0.06%) make their own chips as well.
There's a growing demand for these custom chips as they offer better price performance than Nvidia's market-leading GPUs. The hyperscalers can acquire the chips for less, and they consume less power than GPUs. However, they're not able to run broad workloads like GPUs, making them only suitable for specific tasks.
However, continued advancements in custom silicon have led an increasing number of AI developers to adopt these solutions. Anthropic is using Amazon's chips and plans to use Google's chips starting next year. OpenAI is using TPUs, and it's in talks with Amazon as well. Even Meta is reportedly in talks with Google to use TPUs, as the two look to port the PyTorch framework (the most widely used AI software framework) to its architecture.
Alphabet stands out as a leader in the space as it's seen great results using TPUs in-house, and it's gaining traction with third parties. But investors should look one level deeper at the companies behind the designs: Broadcom (AVGO -0.46%) and Marvell (MRVL -7.80%). Broadcom designs Alphabet's TPUs, but Marvell may be the better stock to buy now based on valuation. It's behind the designs for Amazon and Microsoft's chips, which are also expected to see strong demand in the coming years.
NASDAQ: MRVL
Key Data Points
2. On-device artificial intelligence
Most artificial intelligence applications make a call to a server to access a large language model. That can be slow and energy-intensive, and it requires an internet connection. One significant trend to watch in 2026 is the advancement of on-device generative AI capabilities, also known as edge AI.
Apple (AAPL +1.02%) is expected to deliver on its long-promised Siri revamp, featuring AI-powered improvements. Apple's on-device AI model, Apple Intelligence, will handle many new tasks and ensure data privacy. It may use a third-party model for less sensitive tasks and questions. Many analysts have high expectations for the revamp, as it's been a long time coming.
If Apple impresses with a new Siri, it could push the industry forward in on-device AI capabilities. That would be a boon for Qualcomm (QCOM -4.21%), which makes high-end mobile chips. Faster mobile processors are necessary for on-device AI, which could drive more sales toward high-end devices equipped with Qualcomm's chips.
NASDAQ: QCOM
Key Data Points
Both companies could benefit if new on-device AI capabilities prompt consumers to upgrade their phones and other devices, resulting in strong sales growth. Apple could further benefit from more AI-capable apps in the App Store, which is a major source of services revenue for the iPhone maker.
3. The rise of the AI agents
Many businesses have been working on agentic AI capabilities, leveraging the power of large language models to get AI to perform multistep tasks. The area has been making steady progress, gaining momentum in the latter half of the year.
Salesforce (CRM +4.50%) released its Agentforce platform in late 2024, but it's just starting to gain momentum now. It has expanded the platform to make it easier for businesses to create AI agents that can tap into their unique data and utilize Salesforce's platform to execute business goals, such as handling customer service requests and warming up sales leads before passing them off to a human specialist. Salesforce's dominant position in enterprise software puts it in a unique position to benefit from advancements in agentic capabilities.
NYSE: CRM
Key Data Points
Likewise, Meta Platforms could benefit greatly from agentic AI for small businesses. Instead of hiring a social media ad manager, Meta's AI agent could handle developing and running an ad campaign for a small business, bringing more advertisers onto its platform and expanding advertising budgets. Likewise, AI agents in WhatsApp and Messenger could handle incoming sales leads and service requests, providing another key feature to put small businesses on the same playing field as massive corporations.
Both companies' stocks look attractive at today's price, and they could see strong revenue growth in 2026 as they develop more agentic AI capabilities that they can monetize both directly (by selling the agentic platform) or indirectly (by selling services the agentic AI uses).
