Astera Labs (ALAB) achieved 115% revenue growth to $182.5M in 2025 with a 75.7% gross margin by supplying PCIe and CXL chips that eliminate data movement bottlenecks inside AI server clusters. Credo Technology (CRDO) more than tripled revenue to $407M in Q3 with a 67.8% trailing gross margin by providing optical and electrical connectivity solutions between servers, with data center revenue representing over 90% of sales.

Amazon’s $200 billion AI infrastructure spending this year to scale Anthropic’s frontier models drives surging demand for high-speed connectivity both within servers and across distributed systems, creating a tailwind for companies supplying the internal and external plumbing that hyperscalers depend on.

The analyst who called NVIDIA in 2010 just named his top 10 AI stocks. Get them here FREE.

The AI arms race isn’t just about who builds the smartest chatbot — it’s about who supplies the plumbing. When Anthropic expanded its partnership with Amazon (NASDAQ:AMZN) to a reported $100 billion scale, the headlines focused on cloud dominance and model training. But here’s the real question investors should be asking: who quietly benefits every time more compute gets turned on?

That’s where the opportunity lies. Let’s look beyond the obvious winners and zero in on two under-the-radar stocks — Astera Labs (NASDAQ:ALAB) and Credo Technology Group (NASDAQ:CRDO).

Amazon’s latest commitment to Anthropic builds on prior investments disclosed in Amazon filings and Anthropic announcements, where Amazon pledged billions in cloud credits and infrastructure support through AWS. That’s a fancy way of saying: more GPUs, more networking gear, and more data center buildout.

READ: The analyst who called NVIDIA in 2010 just named his top 10 AI stocks

Amazon’s capital expenditures hit $131.8 billion in 2025, with management explicitly citing AI infrastructure as a major driver. It plans to spend a massive $200 billion this year. With Anthropic scaling toward frontier models, that number likely trends higher going forward.

What the numbers tell us is that AI workloads are bandwidth-hungry:

Training large language models requires tens of thousands of GPUs per cluster.

Networking speeds are moving from 100G to 400G to 800G connections.

Data center interconnect demand is projected to grow between 17% to 20% annually through 2028

That surge doesn’t just benefit Amazon. It lifts the entire supply chain — especially the companies making chips that move data faster and more efficiently.

Astera Labs sits in a niche most investors overlook — connectivity solutions that link CPUs, GPUs, and memory inside AI servers. Its products, like PCIe and CXL-based chips, help data move faster between components. In short, it removes bottlenecks.

That matters because AI clusters are only as fast as their slowest connection. In 2025, Astera Labs’ reported revenue grew 115% year over year to $852.5 million, gross margin came in at 75.6%, and customers include hyperscalers — widely understood to include Amazon’s AWS.

Now compare that to peers:

Company

Revenue Growth

Gross Margin

Astera Labs

115%

75.6%

Marvell Technology (NASDAQ:MRVL)

42%

59%

Broadcom (NASDAQ:AVGO)

24%

67.8%

Let’s translate that. When Anthropic scales workloads on AWS, more GPUs get clustered together. Those clusters require high-speed interconnects — exactly what Astera sells. Surprisingly, the more powerful AI models become, the more they stress internal data pathways. That plays directly into Astera’s strength.

Granted, the company is still relatively small and turned consistently profitable in 2025, generating $219 million in net income last year. That introduces volatility. But when all is said and done, its growth rate ties directly to hyperscaler AI spending — and Amazon has signaled that spending isn’t slowing.

If Astera manages traffic inside the server, Credo builds the highways between servers. Credo specializes in high-speed connectivity solutions — particularly optical and electrical connectivity for data centers. Its chips enable faster transmission over longer distances with lower power consumption.

Credo’s revenue tripled year over year to $407 million in the third quarter with management guiding to it doubling for the full year. Data center revenue accounted for over 90% of total sales, while trailing 12-month gross margin is expanded to 67.8%.

Now stack that up against competitors:

Company

Revenue Growth

Data Center Exposure

Credo

201.5%

90%+

Marvell

42%

74%

Broadcom

24%

43.5%

Here’s why that matters. Anthropic’s models don’t run on a single server — they run across massive distributed systems. That requires moving data between racks, buildings, and sometimes regions. Credo’s chips handle that transmission efficiently.

That said, there’s a catch. Credo’s customer concentration is high — three of them account for 90% of revenue. If AWS shifts vendors or delays spending, growth could wobble. Regardless, the direction of travel is clear. As Amazon deepens its AI footprint, it needs faster, more efficient networking — and Credo sits right in that demand stream.

In short, the Anthropic-Amazon partnership isn’t just about AI models — it’s about infrastructure scaling at a pace the market is still digesting.

Astera Labs benefits from inside-the-server demand — connecting chips and memory. Credo benefits from between-the-server demand — moving data across networks. Different roles. Same tailwind.

That’s the key insight. Investors often chase the headline names — Amazon, Nvidia (NASDAQ:NVDA), Anthropic. But the real leverage sometimes sits a layer deeper in the stack.

That said, these are not risk-free plays. Both companies depend heavily on hyperscaler spending cycles. If AI budgets tighten — something we’ve already seen hints of with companies overshooting AI budgets — growth could slow.

In any case, if Amazon follows through on scaling Anthropic to the level implied by a $100 billion partnership, the demand for high-speed connectivity doesn’t just rise — it compounds. And that’s exactly where Astera Labs and Credo operate.

Wall Street is pouring billions into AI, but most investors are buying the wrong stocks. The analyst who first identified NVIDIA as a buy back in 2010 — before its 28,000% run — has just pinpointed 10 new AI companies he believes could deliver outsized returns from here. One dominates a $100 billion equipment market. Another is solving the single biggest bottleneck holding back AI data centers. A third is a pure-play on an optical networking market set to quadruple. Most investors haven’t heard of half these names. Get the free list of all 10 stocks here.