Deck

Astera Labs · ALAB · NASDAQ

Astera Labs designs the connectivity silicon — PCIe retimers, fabric switches, active cables, CXL controllers — that moves data inside AI server racks, selling fabless chips and software to a handful of hyperscalers building AI infrastructure.

$417
Price (6/18/26) All-time high
$75B
Market cap
$853M
Revenue FY25 +115% YoY
76%
Gross margin FY25
Listed March 2024 at a $62 first-trade close; dipped to $36 by August 2024, ran to $252 by September 2025, drew down to ~$100 by late March 2026, then more than quadrupled to $417 by June 2026 — roughly 7x from IPO across two violent 55%-plus drawdowns.
2 · The tension

The multi-year thesis runs through one design-win race at one customer in 2026.

  • The moat resets every PCIe generation. Aries — Astera's PCIe retimer line — holds roughly 55% share at AI accelerators today and has shipped more than 1M Gen 6 ports (management said "millions" on the Q1 FY26 call). Switching-cost lock-in expires when the spec ratchets, and the Gen 7 design-win race is being decided through 2026, with revenue impact in 2027–2028.
  • Broadcom just entered the franchise. The incumbent in PCIe switching — 75x Astera's revenue, four custom AI ASIC programs at hyperscalers — launched a Gen 6 retimer portfolio in October 2025 and will contest Gen 7 from inside the same customer conversations.
  • One customer is more than 70% of revenue. A Broadcom Gen 7 win at the lead customer would simultaneously break the highest-margin franchise, the moat-extension premium, and the concentration story. The race resolves quietly through 2026; the revenue impact does not land until 2027.
Every other debate — gross margin, warrant accounting, multiple compression — eventually collapses into Gen 7 at the lead customer.
3 · The money picture

FY25 was the inflection — profit emerged, cash compounded, and the balance sheet now funds the next two cycles.

$853M
Revenue FY25 +115% YoY
76%
Gross margin held across 17 quarters
$282M
Free cash flow 33% margin
$1.19B
Net cash zero debt

Revenue scaled 10.7x from $80M in FY22 while gross margin held in a 74–79% band — software-like pricing power through a hardware ramp. GAAP operating margin flipped from −29% in FY24 to +20% in FY25; capex runs at 4% of revenue on a fabless model. The cleanest counterpoint: stock-based comp ran $160M (19% of revenue, 92% of operating income), so SBC-adjusted free cash flow is closer to $122M than the headline $282M.

4 · The customer book

Over 70% of revenue from one buyer — and a ~$900M warrant paid to keep them.

  • One customer, more concentrated than at IPO. The top end customer drove more than 70% of FY25 revenue and the top three drove 86% (FY25 10-K). The Q1 FY26 10-Q put the largest direct buyer at 29% on a customer basis, up from 12% a year earlier — concentration is moving the wrong way.
  • The Amazon warrant locks the relationship in equity. A February 2026 agreement grants Amazon up to ~3.3M shares at a $142.82 strike — roughly $900M intrinsic at spot — vesting against $6.5B of cumulative purchases of Smart Fabric Switches, signal-conditioning, and optical engine products through the milestone schedule. The cost lands as contra-revenue, guided to compress reported gross margin by about 200 bps each quarter from Q2 FY26 forward.
  • The decisive disclosure is a year out. The FY26 10-K, filed February 2027, will show whether a second 10%-plus end customer has emerged alongside the lead buyer. Bulls and bears name the same line as the read that resets the multiple.
5 · The price

Spot $417 sits above every published price target, on a parabolic tape.

  • Three multiples, all stretched. ~74x trailing-twelve-month EV/sales, ~48x FY26 consensus sales, ~139x forward earnings. The nearest pure-play comparable, Credo Technology, trades at ~25x sales while growing faster. The premium is real but exceptional in semiconductors.
  • Analysts sit below the tape, not above. Street average price target is ~$245 — 41% below spot. The Street high is Evercore at $297, 29% below today's close. Ratings remain Buy, but published estimates have not been walked up to the FY26 run-rate.
  • The tape is extended on every reading. +319% over the last twelve months, +71% in the last 30 days, RSI ~70, price ~124% above a rising 200-day average. 252-day realized volatility sits near 96%, with the trailing 30-day reading above 100% — above the 80th percentile of the stock's own history. Two prior drawdowns inside this same 27-month chart cut 55–60%.
6 · The insider tape

230 insider transactions in six months — every one a sale.

  • $200M-plus sold, zero bought. CEO Jitendra Mohan and President/COO Sanjay Gajendra each cashed roughly $45–70M in May 2026 alone under pre-arranged 10b5-1 plans, executed in the $213–235 zone. Chair Manuel Alba's programmed plan covers another 1.412M shares (about $580M at spot) running through August 28.
  • The plans are legal — but the direction is one-way. Both founders still hold more than 4% each and the equity is essentially their entire net worth; this reads as diversification, not exit. The signal is the absence of any offsetting open-market buy as the stock tripled into the print calendar.
  • The CFO seat is fresh. Desmond Lynch arrived from Rambus in March 2026, holds no equity, has not yet headlined an earnings call, and inherits a bonus framework that paid out at 181% of target under his predecessor.
7 · Both sides

A high-quality franchise priced for two more clean cycles — with the deciding race still on the table.

  • For the bull. Eight straight beat-and-raise quarters since IPO, a 76% gross margin held through a 10.7x revenue ramp, $1.19B of net cash, and management commentary that silicon content per AI accelerator has stepped from 'tens' to '$1,000-plus' since IPO. Nasdaq-100 inclusion takes effect June 22 and management expects Scorpio fabric switches to become the largest product line by year-end 2026.
  • For the bear. SBC-adjusted free cash flow yield is well under 1%, the Street consensus price target sits 41% below spot, one customer drives more than 70% of revenue, and Broadcom — with 75x the revenue base — is now in PCIe retimers. Two prior 55%-plus drawdowns are on the same chart.
  • What both sides agree on. The Gen 7 PCIe design-win outcome at the lead customer is the single observable that decides the multi-year case. Every other tension — warrant drag, gross margin durability, multiple compression — collapses into that one race, and it does not resolve in the next quarter.

Watchlist to re-rate: Q2 FY26 print on August 4 — whether gross margin defends 73% with the warrant drag, and what the customer-concentration line in the 10-Q says. Q3 FY26 in November — whether Scorpio reaches its 'largest line' milestone. Broadcom transcripts through 2026 — any named hyperscaler Gen 7 win. FY26 10-K in February 2027 — whether a second 10%-plus customer is disclosed.