Astera Labs, Inc.

Thesis

Astera Labs is the purest merchant-silicon play on inside-the-rack AI connectivity — PCIe 6 retimers today, Scorpio scale-up fabric switches ramping in 2H 2026 — and we expect revenue to roughly quadruple from ~$1.0B TTM to ~$4.2B by mid-2031. At $326.97 (~55x EV/TTM sales), that outcome is already over-discounted: our base case is worth ~$123/share today and even a bull case (~$6.6B 2031 revenue) returns under 4%/yr. This publishes as a watchlist thesis, not a buy — the business passes every quality test, the price fails both valuation gates; our pre-registered entry zone is roughly $95-100.

Why Now

The inflection is documented, not hoped for. Q1 FY2026 revenue was $308.4M, up 93% year over year and 14% sequentially, with GAAP net income of $80.3M (+152%) and a 76.3% gross margin (10-Q for the quarter ended 2026-03-31, filed 2026-05-05). That follows a fiscal 2025 in which revenue grew 115% to $852.5M and net income reached $219.1M (10-K FY2025, filed 2026-02-20). Guidance for Q2 FY2026 is $355-365M — another 15-18% sequential step — meaning the run-rate exits 2026 well above $1.4B annualized (Q1 FY2026 earnings release, 2026-05-05).

The mix shift is the real story. Astera was born as a retimer company riding Nvidia GPU-attach; it is becoming a fabric company. Scorpio fabric switches plus Taurus Ethernet cables reached 30% of revenue in Q4 FY2025, up from effectively zero at the start of that year (Q4/FY2025 earnings release, 2026-02-10), and PCIe Gen 6 products crossed one-third of revenue in Q1 FY2026 (Q1 FY2026 earnings call, 2026-05-05). The Scorpio X-Series 320-lane scale-up switch — supporting PCIe 6, UALink, and Nvidia's NVLink Fusion — enters production ramp in 2H 2026, targeting a merchant scale-up switch market the company sizes at $20B by 2030 (company announcements, May-June 2026). Scale-up fabric is the highest-value new silicon socket in AI racks as accelerator clusters move from 8-GPU nodes to 72+ GPU coherent domains.

The commercial anchor is contractual, not rhetorical: Astera's expanded agreement with Amazon (announced November 2025) ties warrants for roughly 3.3M shares to up to $6.5B of future purchases — a hyperscaler pre-committing its custom-silicon rack roadmap to Astera's fabric. The cost shows up in filings as a roughly two-point quarterly gross-margin drag as tranches vest, and Q2 FY2026 gross-margin guidance of ~73% (down from 76.4%) marks the hardware-heavy mix arriving. The why-now for the theme is genuine; as we show below, the why-now for the stock is not.

Business Quality

The unit economics are among the best in semiconductors: 76.3% GAAP gross margin, 36.2% non-GAAP operating margin, and $74.6M of quarterly operating cash flow on $308.4M of revenue (Q1 FY2026 10-Q). The balance sheet is pristine — $148.3M cash plus $1.04B marketable securities against zero debt — and the model is fabless with only 756 employees at year-end 2025 (10-K FY2025), so revenue per head exceeds $1.1M and scales without heavy capital. Growth is entirely self-funded.

The moat is real but narrower than the multiple implies. Astera's edge is co-design intimacy and speed: it shipped PCIe 6 retimers and fabric switches into production roughly a generation ahead of merchant rivals, and its COSMOS software layer (fleet-scale link diagnostics and telemetry) creates switching costs at hyperscalers who wire it into their rack-management stacks. Protocol breadth — one vendor covering PCIe, CXL, Ethernet, UALink, and NVLink Fusion — makes it the default fabric partner for every custom-ASIC program that is not Nvidia. But the competition is the most dangerous in the industry: Broadcom owns PCIe switching incumbency and is pushing scale-up Ethernet (Tomahawk Ultra), Marvell and Parade sell competing retimers, and Nvidia keeps NVLink largely captive. The step-down in guided gross margin to ~73% is the first visible print of that pressure plus hardware mix.

The structural flaw is concentration: one end customer exceeded 70% of FY2025 revenue and the top three were ~86% (10-K FY2025). The Amazon warrant deepens the dependence even as it secures it — the customer is being paid in equity to stay. Capital allocation is unproven: cash accumulates while stock-based compensation runs heavy (diluted weighted shares grew ~3.5% year over year to 181.2M), and there is no buyback offsetting it. Nothing here is disqualifying for the business; all of it matters for what you should pay.

Financial Base

Reference figures as of 2026-07-23: trailing-twelve-month revenue of $1,002M, 181.2M diluted shares, net debt of $-1,188M, and a share price of $326.97. Sources:

• 10-Q Q1 FY2026 (period ended 2026-03-31, filed 2026-05-08... filed 2026-05-05) — revenue $308.4M, net income $80.3M, cash $148.3M + marketable securities $1.04B, no debt, diluted weighted-average shares 181.2M, 171.4M shares outstanding as of 2026-04-30
• Q4/FY2025 earnings release, 8-K furnished 2026-02-10 — Q4 FY2025 revenue $270.6M (+92% YoY); FY2025 revenue $852.5M (+115%); Scorpio + Taurus = 30% of revenue
• Q3 FY2025 earnings release (Nov 2025) — revenue $230.6M (+104% YoY), GAAP gross margin 76.2%
• Q2 FY2025 revenue $191.9M — cross-checked: FY2025 $852.5M minus Q1 $159.4M, Q3 $230.6M, Q4 $270.6M; TTM = 191.9+230.6+270.6+308.4 = $1,001.5M
• 10-K FY2025 (filed 2026-02-20) — largest end customer >70% of FY2025 revenue, top three ~86%; FY2025 net income $219.1M; 756 employees
• Share price $326.97, Nasdaq close 2026-07-23 (stockanalysis.com); 52-week range $97.93-$499.48

Scenario Valuation

Five-year scenario ranges under the published methodology — the tables below are computed from these exact parameters at publication and sealed with this text. What each scenario assumes:

• Loss (10%): AI capex bust coincides with architectural displacement: Nvidia keeps NVLink captive, Broadcom wins the scale-up Ethernet standard, and the >70% end customer redesigns Astera out of its next rack generation — revenue halves and the multiple collapses. $70 is ~21% of today's price, well above the ~$6.60/share of net cash.
• Bear (25%): AI infrastructure spending digests in 2027-28 and Broadcom/Marvell price aggression caps share: growth decelerates hard after 2026 (5y CAGR 20% still reaches ~$2.5B), margins settle at 32% FCF, and the market pays 18x FCF — the pre-AI Broadcom/Microchip range for low-growth quality semis.
• Base (45%): Scorpio X ramps across three-plus hyperscaler custom-ASIC programs and PCIe 7 holds the retimer franchise: revenue ~4.2x to ~$4.2B by mid-2031 (a leading share of the merchant scale-up/connectivity TAM), 36% terminal FCF margin, and 27x FCF — where mature high-quality semis (ADI ~25-30x, NVDA-mature ~30x) trade with mid-teens growth.
• Bull (20%): Scorpio becomes the de facto merchant scale-up fabric standard (UALink + NVLink Fusion both ramp) and optics attach adds a layer: ~$6.6B revenue by 2031 — roughly a third of the projected $20B scale-up switch market plus the retimer base — at 38% FCF margins and a still-compounding 32x FCF.

Published range (present value): $20.56 – $70.00 · probability-weighted expected value $34.57 against $326.97 at drafting (0.11×) · base-case IRR −30.8%/yr. Gates: base IRR ≥ 15% and EV ≥ 1.3× — both passed at publication. Ranges are graded at the 1-year checkpoint and the 5-year horizon. This is a range, not a target.

Sensitivity — present value across growth × exit multiple

The shaded cell is the one today's price of $326.97 most closely implies. Find your own cell — we show our work rather than assert a number.

Kill Conditions

1. 1. Share-loss signal: two consecutive quarters of sequential revenue decline while combined data-center capex at Microsoft, Amazon, Alphabet, and Meta (from their 10-Qs) is still growing quarter over quarter — demand exists and Astera is not winning it.
2. 2. Fabric transition stalls: Scorpio + Taurus fall below 30% of revenue (the Q4 FY2025 disclosed level) for two consecutive quarters after the 2H 2026 Scorpio X production ramp, or the company stops disclosing the mix metric entirely.
3. 3. Pricing power breaks: non-GAAP gross margin prints below 70% for two consecutive quarters (vs. 76.4% in Q1 FY2026 and ~73% guided) — evidence Broadcom/Marvell competition is repricing retimers and switches faster than mix explains.
4. 4. Concentration unwind without replacement: disclosed largest end customer drops from >70% of revenue (FY2025 10-K) toward <40% in FY2026-27 filings while total revenue growth simultaneously falls below 25% YoY — the anchor left and nobody filled the hole.

Risk Register

• Valuation — critical: ~55x EV/TTM sales means even our bull case ($6.6B revenue, 38% FCF margin by 2031) returns ~3.6%/yr; the multiple, not the business, is the dominant risk.
• Customer concentration — critical: one end customer was >70% of FY2025 revenue (10-K); a single architecture change at that customer invalidates most of the model.
• Nvidia vertical integration — high: NVLink stays largely captive; if hyperscaler custom-ASIC programs disappoint, the merchant scale-up TAM Astera targets shrinks structurally.
• Broadcom/Marvell competition — high: Broadcom owns PCIe switch incumbency and pushes scale-up Ethernet (Tomahawk Ultra); guided gross margin step-down to ~73% is the first print of pricing/mix pressure.
• AI capex cyclicality — high: 2025-26 is a historic build; a 2027-28 digestion phase hits a company priced for uninterrupted hypergrowth.
• Amazon warrant structure — medium: ~2pts of quarterly gross-margin drag as tranches vest, and the anchor customer is partly paid in equity, deepening dependence.
• Dilution — medium: heavy SBC with no offsetting buyback; diluted shares grew ~3.5% YoY.

Theme Position

Astera Labs is effectively a 100% theme-pure holding: the 10-K describes a company "built from the ground up for cloud and AI infrastructure," and substantially all revenue — Aries retimers, Taurus Ethernet cabling, Leo CXL, Scorpio fabric switches — ships into AI/cloud racks. Within ai-infrastructure it occupies the specific angle our theme research flagged: the inside-the-rack connectivity layer (PCIe 6/CXL signal integrity plus the emerging scale-up fabric socket), which grows with accelerator density per rack rather than just accelerator units.

Alternatives considered. Broadcom (AVGO): the incumbent giant in switching and custom ASICs and the safer way to own the theme, but AI is a minority of revenue, the market cap is ~30x Astera's, and the thesis is fully consensus — no edge in us writing it. Marvell (MRVL): cheaper (mid-single-digit EV/sales) with real electro-optics exposure, but its custom-ASIC narrative has been losing sockets and the connectivity attach story is muddier; it is a value case, not this theme's sharpest expression. Credo (CRDO): the closest pure-play comp in AEC cables/retimers with similar hyperscaler concentration, but a narrower product ladder (no fabric-switch equivalent to Scorpio) at a similarly froth-priced multiple — inferior business at the same style of price. Astera is the best asset on the angle; the entire debate is price.

Benchmark

Beating means total return above QQQ over the 5-year window from publication; our base case implies a negative absolute return from $326.97, i.e., clear expected underperformance versus QQQ at today's price. Success means beating QQQ's total window return from publication; the scoreboard grades this thesis against that bar, live, whether we like the answer or not.