Astera Labs: our view and what our model says about 2030
Our view
Astera Labs is a fast-growing, highly profitable maker of the chips that connect processors inside AI server racks. It also depends on one or two big cloud customers and on an AI-spending cycle it does not control. Our simulation's middle outcome for the end of 2030 is about $465 a share. That is about a 19% yearly return from $223, well above our 9% hurdle. It assumes investors keep paying a premium valuation, and the range is wide.
- Price, Nov 4 close
- $223.27
- Hurdle, end-2030
- $319
- Median value, end-2030
- $465
- P10–P90 range, end-2030
- $146–$1,307
- Simulations beating hurdle
- 67%
- End below today's price
- 20%
Prices: official close, Nasdaq, Nov 4, 2026 · Company data: 10-Q, Sep 30, 2026 · What these numbers mean
Market data as of the Nov 4, 2026 close ($223.27). The hurdle and the probabilities depend on that price. Values per share at Dec 31, 2030, undiscounted, from 50,000 simulations.
Snapshot at the Nov 4, 2026 close
- Market value
- $39.0B
- Enterprise value
- $37.6B
- Price/sales
- 25.1×
- EV/sales
- 24.2×
- P/E
- 84.6×
- Price/FCF
- 113.2×
- Revenue growth
- +115%
- FCF margin
- 22%
- Net cash
- $1.4B
Our calculation from the company's filings (latest: 10-Q for the quarter to Sep 30, 2026, filed Nov 4, 2026); not investment advice. Revenue, earnings and cash flow: 12 months to Sep 30, 2026. Price $223.27 (official close, Nasdaq); 174.8 million shares (common shares on the cover page of the 10-Q filed Nov 4, 2026). n/m = not meaningful (e.g. negative earnings).
Earlier versions
| Version | Price used | Horizon | Median value | P10–P90 | Beat chance |
|---|---|---|---|---|---|
| Nov 4, 2026THIS VERSION | $223.27 (Nov 4, 2026) | end-2030 | $465 | $146–$1,307 | 67% |
No earlier versions yet. Each row is frozen as published: when we update this view, the previous numbers stay here.
Version 1.22 · Nov 5, 2026
Editors hold this stock, and also Nvidia, which this article mentions. Updated for the Q3 2026 results. Market data as of the Nov 4, 2026 close ($223.27); market value $39.0B on 174.8M common shares (Q3 10-Q cover). Including options, stock units and warrants on the treasury method (about 187M diluted shares, from the Q4 guide): $41.8B. Latest filings: the 10-K for 2025 and the Q3 2026 10-Q.
Labels: Fact (filings or market data), Company (management claims), Analyst (named or consensus views), Our view (our reasoning). Each section ends with a confidence level.
Simulation result (value per share at Dec 31, 2030, undiscounted; 50,000 simulations):
- P10 $146 · P50 $465 · P90 $1,307 (middle half $260–$805; mean $635).
- Hurdle $319 (the end-2030 price that earns 9% a year from today): 67% of simulations beat it (about +19% a year at the median); 20% end below today’s price.
P10 and P90 are the tails: one replay in ten ends below P10, one in ten above P90. Hurdle: $223.27 × 1.09^4.16 = $319, roughly the stock market’s long-run return to Dec 31, 2030.
What would move the result
- Toward the bull case:
- No deep AI-spending pause before 2031 (with a calm economy, the median would move to about $530).
- Astera’s share of AI spending keeps rising after the Scorpio X ramp.
- Its open scale-up switches (UALink) win customers beyond the lead program.
- Investors pay a higher exit multiple than our middle case.
- Main downside risks:
- An AI-spending pause, which hits a parts supplier harder through customer destocking.
- The lead customer’s program shrinks or moves (one end customer was over 70% of 2025 revenue).
- Ethernet wins scale-up networking.
- The lead customer’s builds pause between generations. An analyst asked on the Q3 call whether that could happen around 2027; the CFO declined to discuss customer volumes but cited continuing deployments. Our model does not include such a gap (section 4).
- The valuation falls to chip-peer levels.
Our view in brief
- Astera Labs makes the chips that connect AI accelerators inside racks, with revenue more than doubling (Q3 2026 $585M, +154% year on year; Q4 guided to $730–750M) and a non-GAAP operating margin of about 45%.
- It depends on very few buyers: one end customer was over 70% of FY2025 revenue, one direct buyer was 53% of Q3 2026 revenue, and the Scorpio X ramp is led by one hyperscaler, which RBC analysts link to Amazon’s Trainium3; the filings do not name it.
- We found no restatement or open control weakness in the filings through Q3 2026, but stock pay is about 70% of free cash flow, tax benefits inflated GAAP profit earlier in the year, and receivables and inventory jumped with the ramp.
- At $223, our simulation gives $260 / $465 / $805 at the end of 2030 (bear/middle/bull), against a $319 hurdle: 67% of replays beat it and 20% lose money.
- The result leans on the exit multiple: because the last four quarters’ average multiple is below the longer-run anchor, our model blends them, 75% average and 25% anchor, and fades the premium only when growth lags expectations. The middle case ends at about 44× operating profit after stock pay, just below that 44.5× average; on peer multiples the result would be far lower.
- Watch whether the second Scorpio X customer reaches volume, whether the lead customer’s builds pause between generations, 2027 hyperscaler capex and the 2027 product launches.
Summary box: what changed, pending events, and the price against our middle case
| Item | |
|---|---|
| What changed since the last report | First published report on ALAB. It reflects the Q3 2026 results: revenue of $584.9M against a $540–560M guide, and Q4 guided to $730–750M. A quarter ago the question was whether the Scorpio X ramp would arrive; now it is what follows it. The stock moved −4% on the day after the results, by our calculation on Nasdaq closes. |
| Pending transformative event | None (no merger, spin-off or financing). The next results, for Q4, are due around February. The 2026 Amazon warrant (up to 3.26M shares at $142.82, vesting with up to $6.5B of Amazon purchases) is the main contract-linked item. In March 2026, Astera filed an S-3ASR registering resale of all 5,578,474 Amazon warrant shares; this is not a financing (Astera gets no proceeds), and exercise can be cashless. |
| Price implies vs our middle case | Middle simulated outcome $465 against a $319 hurdle (9% a year over 4.16 years to Dec 31, 2030): about 46% above it, about 19% a year. It also clears a 10% hurdle ($332). The 2030 revenue middle case is about $5.9B. |
1. The business in one paragraph
Astera Labs designs fabless connectivity chips and modules that move data between processors, memory and network cards inside AI servers; Scorpio X is its bid to become a supplier of AI switches, not just signal-repair parts. Our model ties its revenue to its share of AI spending, plus five new product lines.
The full facts are in Astera Labs’ Long View → /stocks/alab/long-view/
The facts behind this
- Products: four families, all managed by its COSMOS software. Aries retimers clean up and boost PCIe signals over circuit boards and cables. Taurus modules do the same for Ethernet copper cables. Leo controllers attach extra memory over CXL. Scorpio fabric switches come in two lines: P-Series connects mixed devices, and X-Series links accelerators to each other (“scale-up”).
- Manufacturing: TSMC makes all of its chips, and ASE and Amkor package them.
- Position (our read of the press coverage we reviewed): in PCIe retimers it is the market leader.
Confidence: high.
2. Where growth comes from (3–5 years)
The next year rides on the Scorpio X ramp at one lead hyperscaler and on PCIe 6 upgrades. Growth after 2027 depends on product lines that mostly start shipping in 2027 and on whether open protocols win share in scale-up networks, the smallest slice of independent forecasts. Our simulation starts from the top of the Q4 guide, holds Astera’s share of AI spending about flat on average and adds new products, with AI spending growing 12%, 9% and 8% in 2029–31 as our judgment: its middle case is about $3.4B in 2027 and $4.5B in 2028, above the consensus path below, then about $5.3B in 2029 and $5.9B in 2030.
The facts behind this
Drivers, from most to least visible:
- Scorpio X at the lead customer (Company). Scorpio became the largest product family in Q3, with X-Series revenue passing P-Series and the family above $1B annualised. X-Series shipped to two customers in Q3 across multiple platforms. The second, an additional hyperscaler, is in pre-production and should help revenue in 2027. Backlog from more customers ships in Q4, with revenue from them expected in the second half of 2027. It claims more than 10 customer engagements. The filings do not name the lead customer. Amazon’s February 2026 warrant vests with purchases of exactly these product types (fabric switches, signal conditioning and optical engines, up to $6.5B), but we cannot confirm Amazon is the lead Scorpio X account. RBC analysts tie the ramp to Amazon’s Trainium3.
- PCIe 6 upgrades. Aries retimer growth in Q3 was led by PCIe Gen 6, and management expects the same in Q4, with Taurus 800G ramping (Company). PCIe 6 products were over half of Q2 revenue, the last share it gave.
- 2027 launches and later (Company): a Leo CXL volume ramp from about mid-2027; optical products sampling now, with first revenue late in 2027 and co-packaged volumes mostly in 2028–29; custom parts, including NVLink Fusion designs, with first shipments late in 2027 and volume in 2028; UALink-based Scorpio X; and Orion rack designs through partners, with about $100,000 of Astera content per rack.
Size of the opportunity. The independent estimates are smaller than management’s. Company claims include content that can already exceed $1,000 per accelerator in some systems, an aim to double it with each generation, a Taurus market above $4B by 2030, and scale-up optics of about $40B in 2030. Two independent views:
- 650 Group, a networking market researcher, forecasts scale-up switching above $30B by 2030: NVLink $25B+, Ethernet $8B+, and PCIe/UALink, the protocols Scorpio X uses today, only $3B+.
- Dell’Oro, a telecom and data-centre market researcher, sees AI back-end switch spending near $1T cumulatively over 2026–30, with scale-up more than half of it by 2030. It expects Ethernet to become the leading scale-up technology by the end of the decade.
Our view: the end market is large and growing. But the open-protocol slice that Astera is strongest in is the smallest piece of the independent forecasts, so the growth case depends on UALink adoption, Astera making Ethernet or NVLink Fusion products, or both.
Consensus path (Analyst; S&P Global via StockAnalysis, Nov 4, 2026; 2028 from TIKR, October 2026, thin and pre-results). The FY2026 figure is still pre-results on Nov 4: nine months plus the Q4 guide already imply about $2.03B. Consensus beyond 2028 is too thin to use. TIKR reports that the two or three analysts who publish 2029–30 numbers sit below the 2028 consensus.
| FY2024 | FY2025 | FY2026E | FY2027E | FY2028E | |
|---|---|---|---|---|---|
| Revenue ($B) | 0.40 | 0.85 | 1.92 | 3.04 | 3.83 |
| Growth | 242% | 115% | ~125% | ~58% | ~26% |
| Non-GAAP EPS | 4.09 | 6.57 | n/a |
Track record (Fact). Financial guidance has been beaten every quarter since the IPO, by 3–13% above the top of the range, with no steady trend (Q1 beats have been the smallest; Q3’24 +13%, Q4’24 +9%, Q1’25 +3%, Q2’25 +10%). Product timelines are mixed: Scorpio is early, while CXL keeps sliding.
| Quarter | Revenue guide ($M) | Actual ($M) | vs top of range |
|---|---|---|---|
| Q3 2025 | 203–210 | 230.6 | +10% |
| Q4 2025 | 245–253 | 270.6 | +7% |
| Q1 2026 | 286–297 | 308.4 | +4% |
| Q2 2026 | 355–365 | 392.4 | +8% |
| Q3 2026 | 540–560 | 584.9 | +4% |
On products, Scorpio X arrived ahead of plan. CXL (Leo), which management has presented as an upcoming driver since the IPO, is now a 2027 volume story (Our view, medium confidence on the earlier timing). Optical and custom now point to first revenue late in 2027, with volume in 2028.
Backlog quality (adapted). Astera does not report a backlog figure. Customers buy on purchase orders with no minimum volumes (10-K). The Amazon warrant gives Amazon an incentive to buy but no obligation. Beyond disclosed near-term backlog, including Scorpio X parts for additional customers shipping in Q4 (Company, Q3 call), the forecast rests on orders not yet placed. What exists beyond that is design wins, which in this industry tend to last a product generation (2–3 years).
Confidence: medium. The near term is strongly guided; 2028 and beyond is thin consensus plus unproven products.
3. The edge (moat)
Astera’s edge is speed to market, software, and trust earned inside hyperscaler data centres. It is real but young: it rests on a few design wins at a few customers, against much larger rivals, and the patents are thin, so the moat is execution, not IP. That is why our model lets Astera’s share of AI spending drift up or down each year rather than locking it in.
The facts behind this
Why customers choose it (Company/Our view). It was first to market with PCIe 4, 5 and 6 retimers. The COSMOS software gives customers fleet-wide monitoring of links, which matters at hyperscale. Its interoperability lab tests its parts against every major CPU, GPU and accelerator, and its products sit in many protocol camps at once (PCIe, CXL, Ethernet, UALink, NVLink Fusion). Patents are thin: 27 issued and 40 pending at end-2025 (10-K).
Signs it is widening:
- It moved from a single retimer product to four families. Scorpio became the largest family in Q3 2026, about two years after launch.
- Amazon signed a second, much larger warrant in 2026 (Fact). Nvidia, AMD and Arm feature Astera in reference designs (Company).
- The non-GAAP operating margin rose to 44.7% in Q3 through the mix shift.
Signs it is narrowing or capped:
- Broadcom competes in PCIe switches and retimers and leads Ethernet scale-up switching. Marvell, Credo (cables), Montage (retimers, including a Gen6 part) and Microchip compete in parts of the portfolio (10-K; press).
- Nvidia’s own NVLink dominates scale-up today (Dell’Oro). Moving more traffic onto proprietary links on its boards could reduce retimer demand (Our view).
- Management targets a lower long-term gross margin of about 70% as switches and modules grow (Company). GAAP gross margin was 72.3% in Q3, and the 10-Q ties the fall to inventory-related charges, manufacturing overhead and the warrants. The 2026 warrant costs roughly 4% of Amazon-related revenue in contra-revenue.
- If hyperscalers standardise on Ethernet for scale-up, as Dell’Oro expects, Scorpio X’s PCIe/UALink base becomes a niche unless Astera follows. Management says it aims to address the entire scale-up switching market over time, including other protocols as customer interest grows (Q3 call).
Market share: not disclosed and not meaningful at product level, so we skip it.
Confidence: medium.
4. The worst realistic risk
The worst realistic case is a customer and cycle problem, not a solvency one. The lead hyperscaler finishes its Scorpio X build-out or moves to an in-house or Ethernet design just as AI capex digests in 2027–28, so revenue stalls near today’s run rate with costs built for twice the size. Our model carries this as a lead-customer loss event and AI-spending pauses; with $1.38B of cash and securities and no debt, survival is not in question.
One version of this is already being asked about. On the Q3 call, Trip Smith, speaking for Barclays analyst Tom O’Malley, asked whether the lead customer’s current builds peak around early 2027 with the next generation closer to 2028. The CFO declined to discuss customer volumes but cited continuing deployments, new platforms and additional customers. Our model holds Astera’s share of AI spending about flat on average through 2027–28 and does not include such a gap, so this is a stated risk to the 2027–28 figures here.
The facts behind this
| Risk | How likely (3–5 yrs) | How bad | Notes |
|---|---|---|---|
| Lead-customer shift or pause | Medium | Severe | One end customer was over 70% of FY2025 revenue. In Q3 2026, three direct buyers were 77% (53%, 14%, 10%; the largest direct buyer was 29% in Q2), and the largest held 58% of receivables. The 10-Q says certain of them are manufacturing partners buying for end customers, and that the percentages can move as end customers shift production among partners, so 53% is not a measure of one end customer’s demand. |
| A build gap at the lead customer in 2027–28 | Not estimated; not in our model | High for 2027–28 | An analyst’s question on the Q3 call; the CFO declined to discuss customer volumes but cited continuing deployments. |
| AI capex digestion | Medium | High | Beta about 3.9. Northland’s chip analyst Gus Richard has moved between Market Perform and Buy several times this year on valuation and data-centre spending, most recently back to Buy after the Q3 results. The stock fell 11% in one day in late October in a chip sell-off with no company news. |
| Protocol loss (Ethernet or NVLink win scale-up) | Medium | High for the growth case | Independent forecasts put PCIe/UALink at the small end of scale-up. |
| Competition and pricing | Medium | Medium | Broadcom is the biggest threat. Prices for chip generations typically fall over time (10-K). |
| China and Taiwan | Low–medium | High | 25% of Q3 revenue (30% over nine months) was billed to China addresses, and 53% to Singapore (29% in Q2); these are billing addresses, which include customers’ manufacturing partners. Management has said China end demand is single-digit percent. All wafers come from TSMC, and 65% of fixed assets are in Taiwan. |
| Valuation reset | Medium | High | The stock ran from about $98 to $499 and back to $223 in about eight months. |
Legal and regulatory docket (Fact, Q3 2026 10-Q, the contingencies note and Part II Item 1): the company says it is not currently a party to any material pending proceedings, and we found no securities class action in a web search this month. The filing records no indemnification liabilities and does not mention legal accruals. The open regulatory issue is US export controls on China, which is a risk factor, not a case.
Dependence map (Fact unless marked):
- Customers: one end customer over 70% (FY2025) and top three about 86%. Amazon holds warrants for up to 5.58M shares in total: 2.32M at $20.34 (2022/23) and 3.26M at $142.82 (2026). These are customer incentives booked as reduced revenue: $18.1M in Q3 2026 (Q2: $10.2M) and $30.4M in the first nine months. 1.87M warrant shares had vested by the end of September, and none had been exercised. No customer financing or related-party sales.
- Suppliers: TSMC (all chips), ASE and Amkor (packaging and test), and a few module and substrate makers. Purchase commitments are $163.6M, down from about $182M at June.
- Platforms: Nvidia, AMD, Arm and Intel ecosystems; UALink and NVLink Fusion standards.
- Government: US export controls and tariffs; Israel design centre (geopolitical).
Product-programme risk (Company/Our view). Scorpio X is ahead of plan. Everything else in the 2027–28 growth case (UALink switches, CXL volume, optical, custom NVLink Fusion, Orion racks) is still pre-volume. CXL has already slipped.
Survival check (Fact/Our view): passes easily. Cash and marketable securities were $1.38B at the end of September 2026, with no debt on the balance sheet. The business is profitable on a GAAP basis, and trailing free cash flow is about $345M. There are no customer prepayments inflating cash. Astera could get through a bad year, even a 50% revenue drop, without raising money. Opex of about $790M a year (non-GAAP, Q4 guide annualised) would be the pressure point.
Confidence: medium–high.
5. Financial quality (read the notes)
Latest quarter (Q3 2026, from its 10-Q), first nine months of 2026 and trailing 12 months, with full years from the FY2025 10-K. We found no restatement in these filings, but non-GAAP profit excludes stock pay of about 70% of free cash flow, and tax benefits from that stock pay flattered GAAP profit earlier in the year. Our model therefore values Astera on operating profit after stock pay, not on the company’s non-GAAP earnings.
The full facts are in Astera Labs’ Long View → /stocks/alab/long-view/
The facts behind this
In the filings through Q3 2026 we found a clean audit opinion for 2025, conservative non-GAAP treatment of the warrant cost, and no financial engineering.
| FY2024 | FY2025 | 9M 2026 | Q3 2026 | Q4 2026 guide | |
|---|---|---|---|---|---|
| Revenue ($M) | 396 | 853 | 1,286 | 585 | 730–750 |
| GAAP gross margin | 76.4% | 75.7% | 73.5% | 72.3% | ~72% |
| Non-GAAP operating margin | 30.2% | 39.2% | 41.0% | 44.7% | ~45% |
| GAAP operating margin | −29% | 20.3% | 25.7% | 30.6% | ~35% |
| Stock-based pay (% revenue) | 59%* | 18.8% | 15.2% | 14.1% | ~11% |
| Free cash flow ($M) | 102 | 282 | 268 | 134 |
*Includes IPO-triggered stock pay.
Sources: FY2024 and FY2025 from the FY2025 10-K; nine months and Q3 2026 from the Q3 2026 10-Q and results release, with Q3 free cash flow calculated by subtracting first-half amounts (Q2 10-Q) from the nine months; Q4 guide from the Q3 results release (Company). Free cash flow and the Q4 GAAP margin and stock-pay share are our arithmetic on those filings.
What the notes show:
- Revenue vs gross profit (10-Q; our arithmetic). Gross profit is growing slower than revenue (Q3: +140% vs +154%). The Q3 10-Q cites higher inventory-related charges, manufacturing overhead and the warrants, and gives no amount; the Q2 10-Q cited a shift toward lower-margin hardware modules and the warrants. Management has guided to a lower margin as modules and switches grow (Company).
- Receivables and inventory (Q3 2026 10-Q balance sheet; collection days are our arithmetic using ending receivables and quarterly revenue). Receivables rose from $83M at December to $284M, a collection period of about 45 days, the same as at June and up from 28 at December. The 10-Q cites higher sales and the timing of customer payments, and one customer holds 58% of receivables. Inventory more than tripled since December ($59M to $201M) and rose 77% in the quarter, mostly in work in progress; the 10-Q ties it to anticipated demand. This is consistent with the guidance, but it becomes a write-down risk if the ramp slips. Astera has written down inventory before: $9.7M in Q1 2023 for a legacy customer system (company filings).
- Stock pay vs free cash flow (10-K and 10-Q cash-flow statements, trailing 12 months to the end of September 2026; our arithmetic). Trailing stock pay is about $237M against free cash flow of about $345M (69%), and it was 73% in the first nine months of 2026. Unrecognised future stock pay is about $834M (10-Q, stock-pay note). Non-GAAP margins exclude all of it. Our view: stock pay and related items took about a third off non-GAAP operating profit in Q3 ($262M against $179M), so non-GAAP earnings overstate what accrues to shareholders.
- Adjusted definitions (Q3 2026 results release). One small change this year: non-GAAP also excludes fair-value moves on private equity stakes. Warrant contra-revenue is not added back, which is also conservative. Non-GAAP tax is a normalised 11–12%.
- Tax flattering GAAP earnings. Q2 GAAP net income of $153M included a $50M tax benefit, mostly excess deductions from employee stock vesting; Q3 net income of $183M carried a $9M provision (10-Qs). A $52M income-tax receivable sits in current assets. The company still holds a full US valuation allowance. A future release would add a large one-time GAAP gain, which should not be mistaken for earnings. The trailing GAAP P/E (about 85× on common shares, 91× on about 187M diluted shares) is therefore not meaningful.
- Estimate catch-ups and factoring: none identified in the reviewed 10-K and 10-Q. Capital spending is rising ($73M paid in the first nine months against $19M a year earlier), which the 10-Q ties to production equipment, with $66M carried as construction in progress. Distributor price adjustments and rebates are estimated and have matched historical provisions (10-K).
- Goodwill vs equity (Q3 10-Q balance sheet): $92M against $2.0B (5%). The purchase of an unnamed data-centre acceleration team for $74M early in 2026 was booked almost entirely as goodwill, i.e. an acqui-hire. The late-2025 aiXscale deal cost $31M (10-K; 8-K).
- Material weaknesses (10-K FY2025 and PwC’s audit report): two (risk assessment and segregation of duties; IT general controls) were reported at end-2024 and remediated by the end of 2025, with a clean PwC opinion on its controls over financial reporting. PwC’s critical audit matter on the aiXscale accounting notes that a material weakness “existed during the year”. There have been no restatements in the filings we reviewed. The Q3 10-Q reports a new enterprise resource planning system as a change in its controls over financial reporting, and concludes its disclosure controls were effective.
- Accounting-policy changes (10-Q): none material. New guidance on share-based payments to customers (ASU 2025-04, effective 2027) is expected to be immaterial (Company).
Share-count bridge (Fact, 10-Q; Our arithmetic). Single share class, no converts, no preferred.
| Shares (M) | |
|---|---|
| Basic outstanding (end of October 2026, 10-Q cover) | 174.8 |
| Options (avg strike $0.89) | 1.7 |
| RSUs + PSUs outstanding | 8.8 |
| Amazon warrants, 2022/23 ($20.34); 1.87M shares across all three Amazon warrants vested at the end of September 2026 (the 10-Q does not split this by warrant) | 2.3 |
| Amazon warrant, 2026 ($142.82; performance-vesting) | 3.3 |
| Fully diluted, gross | ~191 |
| Treasury-method diluted (Q4 guide) | ~187 |
Net dilution has run at about 4–5% a year (162.0M at end-2024, 170.2M at end-2025, 174.8M in October 2026; 10-K and 10-Q). No share repurchases were reported in FY2025 or the first nine months of 2026. The 2024 plan’s pool rises by up to 5% of outstanding shares each January; it rose the full 5% (8.51M shares) in January 2026, and the ESPP pool rose a further 1% (1.70M; S-8).
Confidence: high on the facts, medium on interpretation.
6. Valuation today
After a 55% fall from the June intraday peak (54% from the $483.02 peak close), ALAB is priced at about 34× next year’s non-GAAP earnings and 13× next year’s sales, well above its fastest-growing peers. What the price needs depends mostly on the multiple. To reach the $319 hurdle, Astera needs in 2031:
- about $1.3B of operating profit after stock pay (roughly $4.3B of revenue at a ~30% margin) if investors still pay ~44× at the end of 2030, as our simulation’s middle case assumes; or
- about $2.8B (roughly $9B of revenue) at a peer-style 20×.
Both figures assume about $6.3B of net cash and 195M shares by then. Our simulation’s middle case for 2031 revenue is $6.3B, so on our model’s multiple the price asks for less than the middle case delivers; on peer multiples it needs growth far above consensus.
The facts behind this
Fact: market value is $39.0B on 174.8M common shares, $41.8B on about 187M diluted shares. EV is about $40.4B (diluted) after $1.38B of cash and securities.
| Company (early November) | EV/sales (TTM) | Fwd P/E | EV/EBITDA (TTM) |
|---|---|---|---|
| Astera Labs | 24× | 36× | 89× (GAAP) |
| Marvell | 15× | 28× | 49× |
| Broadcom | 18× | 19× | 31× |
| Credo | 16× | 19× | 45× |
| Rambus | 10× | 22× | 24× |
Peer data: StockAnalysis, early November 2026 (medium confidence; forward P/E bases vary by provider). EV is on basic shares in the table; forward P/E is StockAnalysis’s forward basis, not FY2027 consensus (34×). On our diluted EV of $40.4B, ALAB’s EV/trailing sales is 25.9×. It trades at about 19.9× 2026 sales (nine months plus the Q4 guide midpoint, $2.03B), 13.3× FY2027 consensus sales, 55× FY2026 non-GAAP EPS and 34× FY2027 non-GAAP EPS. Charging stock pay, our middle case for 2027 earnings is about $5.05–5.29 a share (excluding or including interest income), about 42–44×. On the next four quarters (Q4 2026 to Q3 2027), the basis used in section 7, it is about 45×.
A single day’s price says little, so we also look at the average. On next-four-quarter operating profit after stock pay, the multiple is 38× at the Nov 4 close and averaged 44.5× over Q4 2025 to Q3 2026, using each quarter’s daily-average price and adjusted for the market’s level (our calculation, a geometric average; past quarters use reported results and our current estimates, not what investors expected then). June’s quarter counts at about 49× on its average price, against about 90× at the $483 quarter-end price.
Against its own history (Our estimate, medium confidence): at the $499.48 peak in June, EV was roughly 40× the then-consensus FY2027 sales (about $2.2B). Today it is 13.3× a 2027 consensus that has risen 38% since. The multiple has more than halved while estimates rose.
Analyst views. The consensus is Buy, but it is split, and targets moved in both directions after the Q3 results. S&P Global via StockAnalysis counts 25 analysts: 10 Strong Buy, 7 Buy, 8 Hold, 0 Sell. The average 12-month target is $376 (median $382, range $190–$500). After the results:
- Morgan Stanley’s Joseph Moore (Buy) raised his target to $389, Northland’s Gus Richard moved to Buy with a $315 target, and BMO’s Harsh Kumar started at Buy with $375.
- Among other Buy ratings, RBC’s Srini Pajjuri cut to $400, and Jefferies’ Blayne Curtis and Roth’s Suji Desilva cut to $350. Among Hold ratings, UBS’s Timothy Arcuri cut to $290, Susquehanna’s Mehdi Hosseini to $258 and TD Cowen’s Sean O’Loughlin to $225.
Data sites conflict on Citi’s current rating, so we do not rely on it. Short interest is 5.1% of float.
Confidence: medium.
7. Value at the end of 2030: simulation
The simulation’s middle outcome is about 46% above the 9% hurdle ($466.7 vs $319.4); the full result is in the Our view box at the top. These are the spread of outcomes our inputs imply, not a forecast.
Method: Valued directly at end-2030.
Implied end-2030 multiple: our end-2030 value implies 43.8× (P25–P75 31.2–62.5×) 2031 operating profit after stock pay (EV/EBIT, the basis of our exit multiple); last four quarters’ market-adjusted average, on daily-average prices: 44.5×; at the Nov 4 close on the next four quarters’ figures, not trailing: 38.0×. Also: 13.0× (P25–P75 8.7–19.5×) 2031 sales (at the Nov 4 close, on the next four quarters: 12.5×) and 55.7× (P25–P75 40.7–78.1×) 2031 earnings (at the Nov 4 close, on the next four quarters: 45.4×).
Method in one sentence: each of 50,000 replays builds Astera’s revenue and costs quarter by quarter to Dec 31, 2030, then values it on its next-year operating profit after stock pay at a multiple that, because the recent average is below the longer-run anchor, is centred 75% on its average over the last four quarters and 25% on that anchor, moving with growth versus today’s expectations and the market, and fades its premium toward peers only while growth lags expectations (half of the gap per five years behind).
Implied yearly return P10 / P25 / P50 / P75 / P90: −10% / 4% / +19% / +36% / +53%.
Top assumptions (our model uses 133 company inputs in total, plus a shared economy-wide layer; the main ones are shown here):
- AI spending: growth of 30% in 2027 and 14% in 2028, then 12%, 9% and 8% in 2029–31, which is our judgment after comparing outside forecasts that differ in scope. Its cycle is shared with Nvidia; a pause starts in 5–30% of years and lasts 2–8 quarters. Astera swings about 1.5× Nvidia’s through a temporary customer destock that refills. It is the second-largest driver in the run: about 16% of the spread on our association measure, not a strict split, after the exit-multiple range (21%) and ahead of Astera’s market-share change (11%).
- Astera’s share of AI spending: rises to about 0.28% by late 2026 and about 0.29% in 2028 as Scorpio X ramps, then flat on average with about ±12% a year of variation.
- Exit multiple: centred at about 49×, a geometric blend of 75% the last four quarters’ market-adjusted average (44.5×) and 25% the longer-run anchor, its own history including today (65×), on all economic shares; the blend applies because the average is below the anchor, and its weights are our judgment. Then adjusted for growth versus expectations, a fade toward peers only while a replay’s growth is behind the expected path (in the middle case it runs only briefly), and the market at the exit ≈ 44×.
- New products: five programs with 30–50% odds each, launching 2027–28 and adding about $1.1B of revenue in 2030 in the middle case.
- Gross margin: falls toward about 70% as switches and modules grow, as management has guided.
The most influential events:
- The lead customer’s program is lost or shrinks (15%): core revenue falls 10–40%.
- The second Scorpio X customer does not reach volume (15%), which lowers Astera’s share of AI spending by about a tenth.
- Ethernet wins scale-up networking outside Nvidia (40%): core growth falls by up to 2 points a year.
- A PCIe switch and retimer price war (30%) lowers revenue and gross margin.
- Open scale-up switching (UALink) succeeds (50%), the largest of the five new-product bets.
Sensitivities: a calm economy lifts the outcomes (below); a peer-style multiple would cut them sharply. We also test a gross margin one point lower in every year from 2027, because the Q3 10-Q cites higher inventory-related charges and overhead and gives no amount. One point is our choice: it treats the whole fall from Q2 to Q3 (73.3% to 72.3%) as lasting. It is a test only, and the headline figures do not use it.
| Value per share, Dec 31, 2030 (undiscounted) | P10 | P25 (bear) | P50 (middle) | P75 (bull) | P90 | Average | Chance to beat 9%/yr | Chance of a loss |
|---|---|---|---|---|---|---|---|---|
| Calm economy (no recessions or AI-spending pauses) | $230 | $343 | $530 | $830 | $1,235 | $655 | 78% | 9% |
| Gross margin 1 point lower from 2027 (test only) | $142 | $254 | $455 | $786 | $1,277 | $620 | 66% | 21% |
Our view: the result rests on the exit multiple more than on the business. The middle-case business path (2030 revenue $5.9B) holds Astera’s share of AI spending about flat on average and adds new products. A 44× exit multiple on a business growing about 9% into 2031 in the middle case is above the 38× at the Nov 4 close, just below the last four quarters’ 44.5× average, and generous by peer standards. It reflects our model’s rule, which leans mainly on those recent quarters and partly on the company’s longer history. If investors valued Astera like its chip peers, the result would be far lower.
The facts behind this
| 2030 (simulation, P25 / P50 / P75) | |
|---|---|
| Revenue | $4.3B / $5.9B / $7.8B |
| of which new products (open switching, custom, optical, CXL, Orion) | $0.7B / $1.1B / $1.7B |
| Gross margin | 65% / 68% / 70% |
| EPS after stock pay (13% tax, no interest income) | $5.03 / $8.03 / $11.91 |
| Exit multiple (operating profit after stock pay) | 31× / 44× / 63× |
The results in short, for the end of 2030:
- Middle case (P50: half of the replays end higher, half lower): $465.
- Bear case (P25: a quarter end lower): $260.
- Bull case (P75: a quarter end higher): $805.
- Average: $635.
- The hurdle is $319: today’s price grown at 9% a year to Dec 31, 2030, roughly the stock market’s long-run return. 67% of replays beat it, and 20% end below today’s price.
The simulation replays 2026–2030 50,000 times with ranges for the AI cycle, Astera’s share of AI spending, its new products, customer risk and the end-2030 valuation. Price erosion sits in the price-war event and in the gross-margin path, not in the share assumption.
Near-term fit: our model centres Q4 2026 at the top of the $730–750M guide, because every quarter since the IPO has beaten the top. For 2027 and 2028 its revenue sits about 12% and 17% above the consensus in section 2, because it holds Astera’s share of AI spending flat rather than following those estimates. The EPS lines differ because our model charges stock pay.
Confidence: the inputs are low–medium for 2029–30; the exit multiple is a judgment call.
8. Management and governance
The founders are credible operators with a clean guidance record. But they and the board have sold heavily into the run-up (about $786M in a year, no purchases), and pay rewards adjusted measures that exclude stock pay, so nothing in the pay plan pushes per-share results. The trading-plan timing is a weak flag, not a finding.
The full facts are in Astera Labs’ Long View → /stocks/alab/long-view/
The facts behind this
- Leadership (Fact). Co-founders Jitendra Mohan (CEO) and Sanjay Gajendra (President and COO) are former Texas Instruments interface-chip executives who founded the company in 2017. Every quarter since the 2024 IPO has beaten the top of guidance.
- Credibility events (Fact). The 2024 material weaknesses have been remediated. CFO Mike Tate retired in March 2026 (“no disagreements”). His successor, Desmond Lynch, was CFO of Rambus, a named competitor. A new Chief Accounting Officer (Germaine Cota, ex-BILL) joined in January 2026. There have been no restatements or guidance withdrawals. CXL timing has slipped.
- What pay rewards (Fact, DEF 14A). The annual bonus is 50% revenue and 50% non-GAAP operating margin, which excludes stock pay; it paid 180.6% of target for 2025. The CEO and COO received IPO-time RSU grants worth about $49.5M each (2024), vesting quarterly to about early 2028, and no new equity in 2025. PSUs start in 2026 (metrics not yet disclosed). Nothing rewards per-share results or cash flow after stock pay.
- Insider sales, Oct 2025–Oct 2026 (Fact, our tally of 64 Forms 4): no open-market purchases; officers and directors sold about $786M, mostly under Rule 10b5-1 plans or to cover tax on vesting RSUs. The breakdown by seller and type is in the Long View.
- Plan timing vs events (Forms 4; 10-Q Item 5). The founders adopted plans in early December 2025, about nine weeks before the Amazon warrant agreement and the Q4 report. The market did not treat that report as good news: the stock fell 21% the next day and was around $110–130 when plan selling began. Plan sales started only after the cooling-off period, at $132–290 a share, well below the June high of $499, and both founders have now sold the maximum their plans allowed. Alba’s 2026 plan covers the same 1,412,000 shares as his 2025 plan, an annual pattern that eases the concern; its first sales ($51M) came after the Q2 results were public, and the stock fell 12% the day after that report. About 1.23M shares remain under it through August 2027, close to all of the roughly 1.38M shares he still reports holding. In Q3, three more insiders adopted plans, all small: Chief Accounting Officer Germaine Cota (13,114 shares) and directors Jack Lazar (25,000) and Bethany Mayer (1,063).
- Holders (13G/A, DEF 14A): FMR owned 9.1% (15.8M shares) at the end of September (Schedule 13G/A), down from the 13.7% shown in the April proxy, which relied on an older filing. BlackRock owned 7.4% at the end of June. Directors and officers owned 10.4% at the April record date (DEF 14A), less after the sales since then.
- Board (DEF 14A, 8-K vote results): single share class; staggered (three classes) with plurality voting. At the June 2026 meeting, compensation-committee chair Michael Hurlston, the sitting CEO of Lumentum, got only about 68% support (78.7M for, 36.5M withheld); the other two nominees got 84–85%. He resigned from the board effective Nov 3, 2026; the company says there were no disagreements (8-K). The proxy has a “Director Commitments” section defending his availability, which suggests the withheld votes reflect concern about over-boarding rather than a protest over pay. Say-on-pay passed with 97% (111.5M for, 3.6M against).
Confidence: high on facts; medium on what the plan timing means.
9. Signposts
Q3 passed the test we set: revenue of $585M against at least $550M, a Q4 guide above $650M and Scorpio as the largest family. Next, we watch Q4 against the top of the guide, whether the second Scorpio X customer reaches volume, whether quarterly revenue holds through 2027 if the lead customer’s builds pause, a falling share for the largest end customer in the FY2026 10-K, and revenue from at least two of the 2027 launches.
The facts behind this
| # | Signpost | On track | Broken | When |
|---|---|---|---|---|
| 1 | Q4 results and Q1 guide | Revenue around or above $750M; gross margin near 72% | Miss, or a flat Q1 guide; gross margin well below 72% | Around Feb 2027 |
| 2 | Second Scorpio X customer at volume; more customers behind it | Second customer moves from pre-production to revenue in 2027; Q4 shipments to more customers turn into revenue in H2 2027, as promised | Second customer stalls, or new customers move to 2028 | Through 2027 |
| 2b | Lead customer’s generation change | Revenue keeps rising through 2027 on new platforms and customers, as management says | Flat or falling quarters tied to lead-customer timing | Q1–Q4 2027 |
| 3 | FY2026 10-K concentration | Largest end customer below about 60% | Above 70% again | Feb 2027 |
| 4 | Warrant contra-revenue (proxy for Amazon purchases) | Rising steadily ($18.1M in Q3); each $1B bought ≈ $43M contra | Flat or falling | Each 10-Q |
| 5 | Hyperscaler 2027 capex guides | Growth sustained | Cuts or a “digestion” year | Jan–Feb 2027 |
| 6 | 2027 launches: Orion, UALink Scorpio X, Leo CXL, optical | Revenue from at least two in 2027 | Further slips | Q1–Q4 2027 |
| 7 | Receivables and inventory | Collection period back toward 30–40 days from about 45; inventory turns as Q4 ships | Both keep climbing | 10-K, Q1 10-Q |
| 8 | Insider selling and new 10b5-1 plans | Pace slows as IPO grants finish vesting | Large new plans before big events | 10-Q Item 5; Forms 4. Founders’ December 2025 plans are fully used; Q3 brought only three small new plans; watch Alba’s roughly 1.23M remaining plan shares |
10. Open questions
The biggest unknowns for our model are who the over-70% end customer is and whether Astera can follow scale-up networking if it moves to Ethernet.
The facts behind this
- Who is the over-70% end customer, and is it the Scorpio X lead customer and the Amazon warrant holder? How much of 2026 revenue is it?
- How much of Scorpio X’s design pipeline (10+ engagements) is PCIe/UALink vs Ethernet vs NVLink Fusion, and does Astera plan a native Ethernet scale-up switch?
- The tranche structure of the 2026 Amazon warrant: how much of the $6.5B is expected by when, and what gross-margin drag management assumes.
- Why receivables jumped to about 45 days, and how much is the lead customer’s contract manufacturers. At the end of September, one direct customer held 58% of receivables and another 13%; the MD&A cites “higher product sales and the timing of customer payments.”
- Whether the lead customer’s builds pause between generations in 2027–28, and how much new platforms and customers would cover. The CFO declined to discuss customer volumes on the Q3 call.
- What the Feb 2026 $74M acquisition was, and what the 2026 PSU metrics are.
- A full FY2029–30 consensus set: what we found is two or three estimates, reportedly below 2028.
- How much China-billed revenue (25% of Q3) is exposed to new export rules, versus management’s single-digit end-demand figure.
Sources
Sources (accessed Oct 25 and Nov 4, 2026): 23 filings, releases and data sources
- Astera Labs 10-K FY2025 (Feb 20, 2026)
- 10-Qs Q1, Q2 and Q3 2026 (May 6, Aug 5, Nov 4, 2026)
- 8-K earnings releases Q2 2025–Q3 2026 (Aug 5, Nov 4, 2025; Feb 10, May 5, Aug 4, Nov 3, 2026)
- 8-K Feb 10, 2026 (Amazon warrant; CFO change)
- 8-K Jan 20, 2026 (CAO)
- 8-K Oct 22, 2025 (aiXscale)
- 8-K Jun 8, 2026 (vote results)
- 8-K Nov 4, 2026 (director resignation)
- DEF 14A Apr 23, 2026
- Forms 4 Oct 2025–Oct 2026 (SEC EDGAR)
- Q2 2026 earnings call transcript (Investing.com, Aug 2026) and Q3 2026 call transcript (MarketBeat, machine-made, Nov 2026)
- Nasdaq price history (ALAB closes to Nov 4, 2026)
- StockAnalysis (ALAB statistics, forecast and ratings, Nov 4; peer statistics, early November)
- Investing.com (Northland downgrade, Oct 9, 2026)
- StreetInsider/Investing.com (Northland May 26, 2026)
- MarketBeat (ratings)
- TIKR (Oct 8 sell-off; long-range consensus)
- TipRanks (RBC on Trainium3)
- Quiver Quant and TokenPost (Oct 22 move)
- 650 Group (scale-up switching forecast)
- Dell’Oro Group (Jul 28, 2026 AI back-end switch forecast)
- Futurum/StorageReview (Orion, Oct 12, 2026)
- TradingView/StockScan (Jun 30, 2026 high)
About Astera Labs: Website (external site)Investor relations (external site)SEC filings (external site)Wikipedia (external site)
+ 3 more
Accessed Oct 25, 2026: Astera Labs 8-K, Feb 10, 2026 (Amazon warrant; CFO change)
Accessed Nov 4, 2026: Astera Labs 10-Q for Q2 2026 · Astera Labs Q3 2026 earnings call (MarketBeat transcript)
Quoting this? You may quote short excerpts under our reuse policy. Please credit WhyMyStockMoved.com and link to this page.