The Next Big Moves in AI Chips Are Coming From These Small Players

Over the past year, big-name semiconductor stocks have been the stars of the market. NVIDIA, TSMC, Broadcom—you could almost buy them blindfolded and still make money. But recently, the market has started to shift. Investors are beginning to pour money into smaller semiconductor companies that are tightly plugged into the AI boom.

This week, a number of small-cap chipmakers saw huge gains. Astera Labs, which builds high-speed connectivity chips, jumped 19 percent. Power chip specialist Navitas surged 25 percent, and even kept climbing in after-hours trading. This wasn’t random. Both companies have deep ties to the current king of AI—$NVDA.

Why These Two Names Are Surging Right Now

Astera Labs makes PCIe, CXL, and NVLink connectivity chips—essential components that allow GPUs and CPUs in NVIDIA’s AI server platforms (like HGX and Blackwell) to communicate fast and with minimal latency.

Navitas focuses on high-efficiency power chips. It’s working with NVIDIA to develop the next-gen power systems for AI data centers. Its GaN (gallium nitride) and SiC (silicon carbide) chips support ultra-high-density 800V power architectures. In other words, they help AI systems run faster, cooler, and more energy efficiently.

These aren’t hype-driven companies riding AI buzzwords. They’re real players in NVIDIA’s core ecosystem.

The Market Is Shifting from Mega Caps to Small Caps

AI-related gains have mostly gone to large-cap names so far. But this latest round of buying tells a different story. Investors are starting to rethink where future growth could come from:

• Big names are already expensive, with limited room for surprise upside

• Smaller players are carving out critical niches with huge growth potential

• These companies are often tightly linked to industry giants, offering higher risk but also higher reward

Astera Labs and Navitas are perfect examples of this shift.

Astera Labs Could Be Supplying AWS’s Next AI Chip

One of the biggest potential catalysts for $ALAB  is its rumored role in Amazon Web Services’ upcoming Trainium 4 AI chip. According to reports, Astera provided up to 128 PCIe chips for AWS’s Trainium 2—each of which relied on their components.

Now it’s rumored that Astera will supply the I/O chips for Trainium 4, while another firm, Alphawave, will handle the SerDes (the tech that handles ultra-fast data transmission). The order isn’t confirmed yet, but the logic makes sense: AWS wants to keep customizing chips, and Astera delivers the high-bandwidth, low-latency connectivity to make that possible.

For investors, this isn’t just a partnership—it’s entry into the inner circle of a super-platform customer. That kind of stickiness is rare and valuable.

The New Scorpio Chips Could Become a Major Growth Driver

Astera Labs is also building out its own product line. It recently launched a new brand called Scorpio, a family of smart interconnect switch chips.

It comes in two flavors:

• Scorpio P is designed for server-to-server horizontal scaling, where each server gets a Scorpio chip

• Scorpio X is aimed at vertical GPU-to-GPU scaling, similar to NVIDIA’s NVLink—but with the ability to connect one chip to multiple GPUs, opening up more potential use cases

Scorpio chips are expected to start shipping in late 2025 and become a major revenue driver by 2026. Management says Scorpio could be a core component of next-gen AI servers, with hundreds of dollars in chip value per unit.

Navitas Is Quietly Powering the AI Energy Revolution

$NVTS , the other small-cap winner this week, is riding the AI wave from the power side.

Traditional servers use silicon-based power chips. But as AI models grow bigger and more power-hungry, data centers need better efficiency and higher power density. Navitas specializes in GaN and SiC chips—smaller, cooler, and more efficient—which are perfect for AI systems that draw kilowatts of power.

NVIDIA is rolling out new 800V server architectures, and Navitas is one of its key partners. That means Navitas isn’t just benefiting from a tech refresh—it’s part of a full-blown generational leap in AI hardware.

Why I Think These Small Players Still Have Room to Run

The broader picture is clear:

• AI adoption is happening faster than expected

• The biggest players—NVIDIA, AWS, Google—are customizing chips and building their own stacks

• Small-cap companies that own key pieces of the puzzle could see their revenue multiply over the next few years

Astera Labs and Navitas aren’t cheap, but their long-term potential is easy to see. Astera in particular is moving beyond simple connectivity chips and into full-fledged system components. That’s a powerful shift that could fuel sustained growth.