Astera Labs (ALAB) finished Tuesday up roughly 13.3%, one of the few names in the S&P 500 that closed meaningfully higher on a day when the broader tape was red and Treasurys were getting bid. The catalyst stack:
- RBC Capital Markets reiterated its Outperform rating and raised its price target from $225 to $250.
- CEO Jitendra Mohan presented at the J.P. Morgan 54th Annual Technology, Media & Communications Conference, where he laid out the company’s positioning for the next 18 months of hyperscaler capex.
- The presentation followed Astera’s Q1 quadruple beat — record revenue of $308.4M (+93% year-over-year, +14% sequentially) — and a Q2 guide at least 16% above consensus.
- The company’s newly-launched Scorpio X-Series, a 320-lane AI fabric switch unveiled earlier in May, gave RBC the architectural data point to defend the multiple expansion.
That all happened on a day the S&P closed down 0.67%, the SOX semi index was lower, and Micron, NVIDIA, and the rest of memory and broad-line chips traded slightly to moderately negative. ALAB decoupled.
Why custom silicon is acting differently
Astera’s product line sits in a specific corner of the AI infrastructure stack: connectivity. PCIe Gen 6 retimers, Active Electrical Cables (AECs), and now the Scorpio fabric switch — the components that move data between AI accelerators inside a server rack and between racks inside a hyperscaler data center.
This matters because connectivity scales with cluster size, and cluster size is going UP, not down. When NVIDIA, AMD, or a hyperscaler’s in-house accelerator gets bigger, the connectivity bill grows faster than the chip bill. Astera doesn’t care which silicon brand wins; it sells the picks and shovels to all of them.
That’s the bull-case argument that RBC just defended at $250. They’re saying: AI capex is not slowing, the connectivity attach rate is rising, Scorpio opens a brand-new product category (the “fabric” layer between racks), and ALAB has effectively no direct competitor in the PCIe Gen 6 retimer or AEC market for at least another six quarters.
The valuation problem nobody can dodge
At roughly $250, ALAB trades for around 165x forward earnings.
That’s the most expensive name in the AI infrastructure stack, and it’s the bull case AND the bear case in one number.
Bull case: If the Scorpio X-Series ramps as RBC models, Q3-Q4 revenue accelerates above the current $1.4B run rate, and EPS jumps faster than the share price, the multiple compresses NATURALLY into 2027. That’s how Nvidia compressed from 60x to 30x while the stock went up — you just have to grow earnings faster than the price.
Bear case: If any one of the hyperscalers signals a capex pause — a single Meta, Microsoft, Amazon, or Google headline about “rationalizing” AI spending — the multiple compresses VIOLENTLY. 165x to 90x is a 45% drawdown in a week, and the technical setup doesn’t care that the fundamentals are still intact.
The other bear risk is customer concentration. Astera has not disclosed a clean 10-K breakdown of how much of its revenue comes from its top one or two customers — but the analyst community generally assumes the top two are responsible for the majority of revenue. If either one diversifies away or insources connectivity, that’s a one-quarter air pocket.
The cleaner trade
If you believe the custom-silicon thesis but don’t want to wear the 165x multiple solo, the pair trade is the more elegant expression:
Long ALAB / short SOX (or short MU, short broad-line memory). This isolates the “custom AI silicon beats commodity chips” thesis without making you net long the entire semiconductor cycle. On a day like Tuesday — ALAB up 13%, memory and broad-line chips down 1-2% — that pair generated a 14-15% spread.
That’s not a typo. The thesis is working at full strength right now.
The Play
The setup into the FOMC minutes at 2 PM ET cuts both ways for AI infrastructure. Hawkish minutes mean higher yields and a discount-rate hit to all high-multiple growth names, including ALAB. Dovish minutes mean lower yields and a multiple-expansion bounce — which would extend the run.
Three concrete moves:
- If you’re already long ALAB: Tuesday’s 13% rip is a position you trim into, not chase. The setup is intact, but a 13% gap is usually not a continuation; the next leg comes after a consolidation.
- If you’re building a position: Use the pair trade (long ALAB, short SOX). It expresses the thesis with cleaner risk than naked long.
- If you’re looking for the next 18 months of catalysts: Watch for Scorpio design-win announcements in Q3 and the first datacenter deployment in late Q4. Those are the two prints that take ALAB from “most expensive connectivity name” to “most expensive name with proven new-product traction” — a category that historically commands premium multiples for two more years.
Custom AI silicon kept working on Tuesday. The question for today is whether 2 PM lets it keep working through the end of the week.