Initiated 2026-10-02 · Price $356.42 (as of 2026-10-01 close, stockanalysis.com) · Mkt cap $61.8bn basic / $65.3bn on 183.3mm diluted · Information technology / semiconductors (AI connectivity) · Model: verified

Rating: Sell — Conviction: High

Astera Labs executes superbly: Q2 2026 revenue rose 104% to $392.4mm, Q3 is guided to $540–560mm, and it holds $1.25bn of net cash. The problem is the price. At $356 the stock trades at ~359x FY2025 GAAP EBITDA (after stock comp) and above even the bull case, which has revenue reaching ~$7.5bn by 2030 at a 41% margin. One end customer was >70% of FY2025 revenue, compounding the duration risk.

Conviction tests (3a-v-c): T1 pass (Base −76.1% at +1pp, −73.6% at −1pp; Sell at both; Bull still −39.5% / −32.4%) · T2 pass (Gordon −88%, exit-multiple −62%; Sell on either) · T3 pass · T4 pass

The rule's High may overstate certainty: a five-year DCF can't fully credit the decade of AI growth the market is pricing.

Model value range vs price
Bear $16.46Base $89.49Bull $227.48Price $356.42

Business overview

Astera Labs is a fabless chip company whose connectivity silicon moves data between accelerators, CPUs and memory inside AI servers and racks. The products are Aries PCIe/CXL retimers, Scorpio smart fabric switches (P-Series PCIe 6 and the X-Series 320-lane scale-up switch), Taurus Ethernet smart cable modules and Leo CXL memory controllers. In FY2025 one end customer took more than 70% of revenue and the top three about 86% (10-K). Earnings rest on three variables. The first is AI server and accelerator build volumes at that end customer. The second is Scorpio's ramp: the company says fabric switches become its largest product family in Q3 2026, a quarter early. The third is operating leverage on a fast-growing R&D base, with GAAP opex guided at $232–236mm for Q3 alone. SBC was $160.0mm in FY2025.

Competition

Of the competitors the 10-K names, Broadcom and Marvell matter most for switches and custom-accelerator connectivity, Credo in cables and retimers. Pressure would show first in gross margin (Q3 guide ~72%, down from 73.3% in Q2 as switches gain mix), then in next-generation socket awards at the largest customer. Competitors are growing as fast as Astera: Credo's revenue for the quarter to 2026-08-01 was $479.0mm, up 115% year on year (Credo Q1 FY2027 release), against Astera's 104% in Q2. The listed comparables are far cheaper: Credo trades at 71.4x trailing EV/EBITDA and 28.8x forward P/E, Marvell at 83.0x and 49.1x, and Broadcom at 32.1x and 19.8x.

Bull case

  1. Scorpio ramp — the X-Series 320-lane switch takes Q3 revenue up ~40% quarter on quarter, and P-Series ramps broadly in 2027, with UALink and NVLink Fusion adding sockets. Plays out if a second hyperscaler adopts Scorpio X in volume and revenue compounds 75%/45%/30% after 2026. Model: rev_growth
  2. Operating leverage — the GAAP operating margin rose from 21.6% in H1 2026 to ~29.5% implied by the Q3 guide. A larger revenue base on fixed R&D could carry the GAAP EBITDA margin past 40%. Plays out if gross margin holds near 72% as switches dominate the mix. Model: ebitda_margin
  3. Premium holds — the Amazon warrant, which vests with up to $6.5bn of purchases of switch, signal-conditioning and optical products, signals a multi-year commitment. That keeps a growth multiple well above Broadcom's at year five. Plays out if purchases track the warrant tranches and beta normalises toward 2.0. Model: exit_ev_ebitda, wacc

Bear case

  1. Duration at a 5.2% risk-free rate — at ~359x FY2025 EBITDA, almost all value lies beyond year five. CAPM on a peer-average beta of 2.3 gives a 16.7% cost of equity, and a de-rating toward chip peers is the base path. Plays out if the 10Y stays above 5% or AI leadership unwinds. Model: wacc, exit_ev_ebitda
  2. Single-customer concentration — more than 70% of FY2025 revenue came from one end customer. A platform transition, a second-source decision or a capex digestion there would stall growth outright, not trim it. Plays out if FY2027 growth slows to ~15% and revenue falls in 2028–29. Model: rev_growth
  3. Mix and pricing — switches carry lower gross margin than retimers, and Broadcom and Marvell compete for the same sockets. Q3 opex annualises to ~$0.9bn, which deleverages hard on flat revenue. Plays out if gross margin falls toward the mid-60s. Model: ebitda_margin

Valuation & balance sheet

Metric (definition) Current Own history (range or 5y avg) Peers Source, as-of
EV/EBITDA (model basis: diluted mkt cap − cash & securities + warrant; GAAP EBITDA after SBC) 359x FY2025 · ~228x LTM (est.) n/a (listed since March 2024) CRDO 71.4x · MRVL 83.0x · AVGO 32.1x (trailing, aggregator) Model; stockanalysis 2026-10-01
EV/Sales (aggregator, trailing) 50.4x n/a (unverified) CRDO 24.4x · MRVL 25.0x · AVGO 18.8x stockanalysis 2026-10-01
Forward P/E (consensus) 63.5x n/a (unverified) CRDO 28.8x · MRVL 49.1x · AVGO 19.8x stockanalysis 2026-10-01
FCF yield (TTM FCF / basic mkt cap) 0.4% ($276.7mm) FY2025: OCF $319.3mm − capex $37.5mm = $281.8mm n/a stockanalysis; Q4 2025 release

Model-implied value range (from model-summary.json; generic module, Gordon-growth DCF and exit-EV/EBITDA DCF, midpoints): Bear $16.46 · Base $89.49 · Bull $227.48 per share, i.e. implied returns of −95.4% / −74.9% / −36.2% vs $356.42. These ranges show how the bull and bear drivers translate into value; they are not price targets. The price sits ~57% above the bull midpoint: the market pays for more than five years of 30–75% growth and 40%+ margins. The base 28x exit multiple is a ~92% de-rating from today's same-basis 359x. The methods differ widely (Base Gordon $41.96, exit $137.02): the exit method embeds the higher terminal multiple, is the fairer read for a business still growing at year five, and is still 62% below the price. One end customer is more than 70% of revenue, so I ran a tail sensitivity on the bear: FY2026 +105%, then −20%, −15%, −5% and +2%; EBITDA margin falling from 25% to 10–12%; NWC 15% of revenue; a 12x exit; and an 18.7% WACC. The bear midpoint falls to $7.15.

Balance sheet: not meaningful (net cash): $1,253.0mm of cash and securities at 2026-06-30, no funded debt, net leverage −7.0x FY2025 EBITDA. The February 2026 Amazon warrant (up to 3.26mm shares at $142.82) is carried as a $696.8mm claim at today's price, ~$3.80 a share. No ratings.

Model note: verified: LibreOffice recalculation matched all 3,206 formula cells. Unverified inputs: none. Assumptions without basis: none. No scenario CHECKs. WACC uses a peer-average beta of 2.3 (CRDO, MRVL, AVGO). That is a deliberate haircut from the own 3.78 aggregator beta, which covers only ~2.5 years and would give 24.1%. EBITDA is GAAP after SBC; NWC excludes prepaid items. The company guides one quarter ahead (Q3 GAAP EPS $0.87–0.92), so there is no annual EPS guide to test Base FY2026 EPS of $2.81 against; Q2 GAAP net income ($153.1mm) exceeded operating income on tax items, so reported GAAP EPS isn't comparable. Whether warrant shares sit in the diluted count is n/a (unverified); carrying them as a claim may double-count up to ~$3.80 a share.

Scenario stress test

Reasoned from the bull/bear drivers above. The model column comes from the scenario overlays (Base case + shock).

Scenario Effect Mechanism Magnitude Model Δ value vs Base ($/sh)
S1 Fast equity crash − Beta 3.78 on a ~359x multiple; de-risking hits the multiple, no operating channel Med −$18.35
S2 Slow bear / recession − Hyperscaler capex cuts, supply-chain inventory digestion, a ~$1bn opex base deleverages. Transitory, so about half the permanent Base − Bear gap ($73.03) High −$39.45
S3 Rapid rate shock − No debt; nearly all value beyond Y5 re-prices on discount rate and growth multiple Med −$19.06
S4 Slow rate grind − Same channel, grinding Low −$7.92
S5 Soft-landing cuts + Lower discount rate on a long-duration equity; capex intact Med +$11.19
S6 Recession-driven cuts − Capex cuts and growth de-rating outweigh lower rates Med −$25.14
S7a Credit liquidity shock − Net cash, no funding need; forced selling hits top-multiple equities Med −$11.01
S7b Slow default cycle − Debt-funded neoclouds and data-centre developers lose financing, trimming second-tier AI demand Med −$13.07
S8 Stagflation − Wafer, substrate and wage inflation plus higher real rates on a long-duration equity Med −$17.27
S9a Dollar spike − USD pricing into Asian ODM supply chains; modest Low −$0.79
S9b Dollar slide 0 No material effect, not modeled (USD pricing and costs) Low $0.00
S10 Melt-up + AI-leadership, beta-3.8 name: momentum extends the multiple Med +$18.35
S11 Energy supply shock − No direct cost channel; risk-off and rate pressure only Low −$3.67
S12 Mega-cap/AI derating − The direct channel: AI-leadership unwind de-rates the multiple and trims connectivity orders High −$38.96

Currently active/on watch per the playbook: S3 partially active (10Y ≥5% price-level leg met); S8, S10, S11 on watch.

Model value change vs Base, by scenario
S2 Slow bear / recession−$39.45S12 Mega-cap/AI derating−$38.96S6 Recession-driven cuts−$25.14S3 Rapid rate shock−$19.06S1 Fast equity crash−$18.35S8 Stagflation−$17.27S7b Slow default cycle−$13.07S7a Credit liquidity shock−$11.01S4 Slow rate grind−$7.92S11 Energy supply shock−$3.67S9a Dollar spike−$0.79S9b Dollar slide$0.00S5 Soft-landing cuts+$11.19S10 Melt-up+$18.35

What would change the call

Upgrades if: the price falls toward the bull midpoint while Q3–Q4 deliver the Scorpio ramp with GAAP gross margin ≥72%, or a second hyperscaler is disclosed above 10% of revenue, cutting concentration below 50%. Downgrades if: n/a below Sell. The conviction weakens if Q4 guidance implies sequential deceleration below 10%, or if the 10-K shows the largest customer's share rising further.

Watch items

Sources