---
ticker: "ALAB"
company: "Astera Labs, Inc."
date: "2026-10-23"
view: "Avoid"
fair_value: 149.19
price: 261.24
price_source: "Robinhood last trade 10:18:23 ET, regular session; previous official close $257.82 (2026-10-22)"
price_time: "2026-10-23T14:18:23Z"
price_vs_fair_value: "+75.1%"
bear: 38.5
bull: 363.93
exit_price: 266.82
p_price_justified: "under 20%"
restated: "2026-10-26"
method: "case-spread-2026-10-24"
original_fair_value: 149
original_view: "Avoid"
page: "https://alpha.dming.si/research/2026-10-23-alab"
---

# ALAB · Astera Labs, Inc. · Avoid

Research report as of 2026-10-23, blind. Page: https://alpha.dming.si/research/2026-10-23-alab

We value ALAB at $149 a share. The report's $261.24 price (Oct 23, 2026) was 75% above fair value: Avoid.

_Restated Oct 26, 2026 (method case-spread-2026-10-24); the report's original value was $149, Avoid._

| Lens | Range | Marks |
| --- | --- | --- |
| Our report | $38.50 to $364 | fair $149; original $149 |
| Our simulation | $35.29 to $314 | median $122 |
| Street targets (8) | $153 to $500 | average $370, median $388 |
| Price | | $261.24 |

Bands run from our bear to bull (the hollow mark is the original $149), our simulation's 10th to 90th percentile and the Street's low to high. Street targets forecast future prices, not value today. Market data, captured Oct 23, 2026.

| Figure | Value |
| --- | --- |
| Fair value | $149 |
| Price | $261.24 (Robinhood, Oct 23, 7:18 AM PT) |
| Five-year return | 0.9% (a year, at the report's price) |
| Bear / bull | $38.50 / $364 |
| Exit price | $267 (on Dec 31, 2031) |
| Confidence | 20% |
| Price against our scenarios | The price is justified in under 20% of our scenarios (4,000 simulated outcomes, each drawing the uncertain drivers from ranges we chose: our judgement, not a measured chance) |

**What the price needs.** Any one of these, on its own, would justify the price:

- Growth 2028-31 vs our path: needs 19.0%, we have 0.0%

_This text argues the original call. The current figures are at the top._

**Original call, Oct 23, 2026 (before the Oct 26, 2026 restatement)**

_As of 2026-10-23 · blinded, with four small exposures named in assumptions.md (blind commit). The blind value was $138; the model audit moved it to $149 (section 9)._

- **Fair value today:** $149 per share, discounting at a 12.91% cost of equity. Weighting the pessimistic, central and optimistic cases 25/50/25 (our weights, not calibrated) gives $158.
- **Price:** $261.24, Robinhood, last regular-session trade 2026-10-23 10:18 ET (previous official close $257.82) → 75% above fair value
- **Five-year exit price:** $267 a share on 2031-12-31: the average of two estimates, one from letting Astera's returns fade toward its cost of capital ($137) and one from what similar chip companies have traded at ($397, at 36.5x operating profit before acquisition amortization). Fading over 10 or 20 years instead gives $263 or $270.
- **Five-year return:** +0.4% a year from $261.24 to the central exit (no dividends or buybacks), against a 12.9% cost of equity
- **Bear / bull today:** $47 / $288: our pessimistic and optimistic cases, each one consistent business outcome.
- **Confidence:** About 20%: our judgment of the share of simulated values within 20% of $149 (subjective, not calibrated; odds.json)
- **Moat:** Our judgment: Astera is the leading independent supplier of the small chips that keep signals clean between AI accelerators, processors and network cards (PCIe retimers), and is becoming a second source of the switches that link accelerators inside a rack. Its edge is being first with each new link standard and a software layer (COSMOS) that customers build their fleet monitoring on; it owns no factory and competes with Broadcom, which is far larger.
- **Market expects:** Far more than our central: revenue growth in 2028-2031 about 19 points a year above ours (+44%, +29%, +29%, +29%, about $10.8B in 2031), with everything else as in our central case
- **Price needs:** Growth 2028-31 at +19.0 points a year over our path. No other single driver reaches the price within its bracket: 2027 revenue at 1.5x ours gives $218, operating cost at 15% of revenue $227 (odds.json)
- **Odds:** P(price justified): under 20% (13%), the share of our simulated scenarios in which the value is at or above the price. Eight in ten simulated values fall between $54 and $282. Subjective, not calibrated
- **Key driver:** Revenue growth after 2027: 5 points a year slower or faster in 2028-31 is -$22 / +$25 a share. Next, operating cost: 32% / 25% of revenue is -$31 / +$14. Each $100M of 2027 revenue above our $3,493M, kept through 2031, is worth about +$3.90; a 20% shortfall is -$28.
- **Street:** Average 12-month target $370 (median $388, range $153-500): 31 targets over the last year averaging $287, the last quarter's 8 at $401, none in the last month (market data, fetched 2026-10-23; the date of the latest target was not checked). An expectations reference, not validation

**The short version.** Astera Labs has gone from $116M of revenue in 2023 to a company
guiding $540-560M for a single quarter (Q3 2026), on the back of AI server build-outs.
We think its business is worth about $149 a share today. Three things decide whether that
is too high or too low:

1. **How long the step-up lasts after 2027.** We hold Astera's share of AI accelerator
   spending (0.47% in 2026) through 2027 and let it rise to 0.65% by 2031 as new switch
   customers, a new open standard for linking accelerators (UALink) and optical products
   arrive. That gives revenue of $3.5B in 2027 and $5.8B in 2031.
2. **One customer.** In 2025 one end customer bought more than 70% of what Astera sold
   (10-K 2025). The filings do not name it. If its switch program slips or is shared with a
   second supplier, 2027 can be a fifth lower, worth -$28 a share.
3. **How much of the profit owners keep after stock pay.** Stock-based pay is about half of
   Astera's operating cost (48% in the Q3 2026 guide). We charge it in full as a cost, and
   count the awards already granted in the share count once.

**Avoid: at $261.24 Astera trades 75% above our $149 fair value and close to our
optimistic case ($288).** The price needs revenue growth after 2027 about 19 points a year
faster than ours, or the Street's own revenue and margins plus a much lower required return
than we use (section 5). Our blind value, committed before we saw the price, was $138; the
model audit found that our cost assumption fell behind our own Q3 forecast and that three
tax assets were missing, together +$11 (section 9).

Terms. The **central case** is our best estimate; the **pessimistic** and **optimistic**
cases are coherent alternative outcomes we weight 25% each. **Non-GAAP** figures are the
company's own, before stock pay and a few other items; **GAAP** figures follow US accounting
rules. **EBITA** is operating profit before amortization of intangible assets bought in
acquisitions. **Accelerator spend** is the revenue of the AI chip makers (NVIDIA's data
center line plus Broadcom's AI line), our sector note's measure of AI hardware spending.

## 1. What Astera sells and what limits it

Astera sells four families of connectivity products, all used inside AI servers and racks
(10-Q Q2 2026; calls):

- **Aries** retimers clean up PCIe signals so that accelerators, processors and network
  cards can sit farther apart. This was the founding product.
- **Scorpio** switches route traffic among many chips. The P-Series links accelerators to
  processors and network cards ("scale-out"); the X-Series links accelerators to each other
  inside one rack ("scale-up"). Management says Scorpio becomes the largest product family
  from Q3 2026 and that X-Series alone carries $1,000 or more of content per accelerator
  (management claim, Q2 2026 call 2026-08-04 and Citi conference 2026-09-09).
- **Taurus** Ethernet cable modules and **Leo** CXL memory controllers, smaller today; Leo
  is expected at two large cloud customers in 2027 (management claim, Q2 2026 call).

Astera designs chips and buys manufacturing from one foundry (10-K 2025). Gross margin was
73.7% (non-GAAP) in Q2 2026 and is guided to about 72% in Q3, falling toward management's
long-term 70% target as switches and modules, which carry lower margins, grow in the mix
(management claim, Q2 2026 call).

**Who pays.** Three layers, which the filings keep separate:

| Layer | What the filings say |
| --- | --- |
| End customers (who deploys the product) | One more than 70% of 2025 revenue; the top three about 86% (10-K 2025). Not named. |
| Contracting parties (who is invoiced: distributors and contract manufacturers) | Q2 2026: A 29%, B 25%, C 15%, D 13%. 2025: A 20%, B 20%, C 17%, D 16%, E 11% (10-Q note 2, 10-K). The letters need not be the same companies across periods. |
| Named in a filing | Amazon, as a customer and holder of three share warrants; the 2026 warrant vests 3,262,299 shares as Amazon buys up to $6.5B of Astera products through 2033-02-05 (8-K 2026-02-10). |

The filings do not say whether Amazon is the end customer above 70%; we do not assume it.
Billing addresses are China 39%, Singapore 29% and Taiwan 27% of Q2 2026 revenue, which are
the contract manufacturers' locations; management says Chinese end demand is in the single
digits (management claim, Q2 2026 call).

**What limits it.** (a) The pace of AI hardware spending at a handful of cloud companies.
(b) Broadcom, which sells PCIe switches and retimers and is many times larger; management
names it as the main switch competitor (Citi 2026-09-09). (c) Accelerator makers building
the link in-house (NVIDIA's NVLink is the model for scale-up). (d) A single foundry.

## 2. Recent results against the bar

Astera has beaten the midpoint of its own revenue guidance in each of the last eight
quarters, by +4.2% to +16.0% (median +9.6%), and none came in below the midpoint (company
releases, Q3 2024 to Q2 2026; preview.txt). Q2 2026 revenue was $392.4M against a $355-365M
guide; non-GAAP gross margin 73.7% against about 73%; non-GAAP operating margin 39.1%
(release 2026-08-04). Operating cost has run above its guide in six of the eight quarters,
by up to +10.4%, so the profit beats are smaller than the revenue beats.

The Q3 2026 guide is the largest step Astera has given: revenue $540-560M, 40% above Q2
(earlier guides stepped 8-27%), non-GAAP gross margin about 72%, operating cost $156-160M,
non-GAAP diluted EPS $1.16-1.21 and GAAP EPS $0.87-0.92 (release 2026-08-04).

## 3. Our forecast and value

The model is `model.py` on the shared method (`output.txt` is its
saved run); every input, with its source, is in `assumptions.md`.

| Central case ($M) | 2026 | 2027 | 2028 | 2029 | 2030 | 2031 |
| --- | ---: | ---: | ---: | ---: | ---: | ---: |
| Revenue | 2,031 | 3,493 | 4,366 | 4,802 | 5,283 | 5,811 |
| Growth | +138% | +72% | +25% | +10% | +10% | +10% |
| Non-GAAP gross margin | | 71% | 70% | 70% | 70% | 70% |
| Non-GAAP operating margin | | 43.8% | 42.8% | 42.8% | 42.8% | 42.8% |
| Stock pay, % of revenue | | 13.1% | 12.5% | 12.0% | 11.4% | 10.9% |
| Free cash flow | | 854 | 958 | 1,068 | 1,159 | 1,312 |

- **2026** is the reported first half ($700.8M) plus our Q3 ($603M) and Q4 ($727M) from the
  guidance record (section 7).
- **2027** holds Astera's 2026 share of accelerator spend (0.47%) while that spend grows 72%
  on our sector note's proposed path (semis-ai 5.3, 2026-10-15).
- **2028-2031** lift that share to 0.65%. Our sector note has accelerator spend growing only
  8%, 0%, 4% and 6% in those years, so most of our growth is Astera taking share: new switch
  customers, UALink switches in 2027, optical links from late 2027 and CXL at two cloud
  customers (management claims, Q2 2026 call). In 2030 our central is 21% of the $25B market
  management says Astera can serve (Q4 2025 call); Astera had about 34% of its 2025 market.
- **Margins** follow management: gross margin at the 70% target from 2028 (we do not let it
  sit above the target), and operating cost at our Q3 forecast's 27.2% of revenue (the
  guide's $158M plus Astera's usual overrun, on our $603M) with no further leverage. Management has given no operating-margin target that we found.
- **Stock pay** is charged as a cost: 48% of non-GAAP operating cost in 2027, falling to 40%
  by 2031 (51% in 2025, 48% in the Q3 guide). The awards already granted (restricted units,
  performance units and options, 11.2M shares) are in the 186.4M share count, so the $680.9M
  of expense still to be booked on them (10-Q note 10) is taken out of future charges
  ($151M, $312M, $161M and $57M in H2 2026-2029) rather than counted twice.
- **Amazon's warrants**: the 1.69M shares vested or probable are in the count. The 3.89M not
  yet vested are left out because their cost already lowers revenue (about 2 points of gross
  margin, Q1 2026 call); counting them as well is worth -$3.05.
- **Cash and tax assets**: $1,253M of cash and securities, a $60M tax refund due (10-Q), the
  federal tax losses ($270M, worth $57M at 21%) and $109M of federal research credits
  (10-K 2025), less $7M of acquisition holdbacks and $150M kept for operations, is $1,323M
  ($7.09 a share). Leases are not deducted because rent stays in the operating cost. No
  debt, no dividend, no buyback; free cash is kept and earns the risk-free rate.

**Cost of equity.** Astera listed in March 2024, so its own share-price history is about two
and a half years long and noisy (raw beta 3.69). We use the median of five connectivity and
accelerator peers instead (raw 2.22, adjusted toward 1 to 1.815; market data 2026-10-23), giving
12.91%. On Astera's own adjusted beta (2.80) the cost of equity is 17.1% and the value
$118 (-$31); on the US semiconductor industry's beta of 1.52 (Damodaran, January 2026,
secondhand via the advocate review) it is about $163.

**The exit.** The five-year exit is the average of two estimates that disagree sharply:
letting Astera's high returns fade toward its cost of capital over 15 years gives $137; the
multiple peers have historically traded on gives $397. They are far apart because peers'
historical multiples (36.5x operating profit before acquisition amortization) imply a much
lower required return, or much longer high growth, than our 12.91% with returns fading to
25%. We take the method's average rather than pick one. The peer multiple is one change
from the method's default: the default divides peers' value by GAAP operating profit, which
for Broadcom and Rambus is after billions in acquisition amortization, while Astera's
operating profit has almost none. On a like basis the multiple is 36.5x, not 43.3x
(`anchor_ebita.txt`, SEC filings); using the default would add $18.

**Sensitivities** (central fair value $149, one change at a time; output.txt):

| Change | Value | Effect |
| --- | ---: | ---: |
| 2027 revenue 20% lower (lead program slips) | $122 | -$28 |
| Growth 2028-31 5 points a year lower / higher | $127 / $174 | -$22 / +$25 |
| Operating cost 32% / 25% of revenue | $119 / $164 | -$31 / +$14 |
| Gross margin 3 points lower | $135 | -$14 |
| Stock pay held at 48% of cost / falling to 30% | $140 / $162 | -$9 / +$13 |
| Peer multiple 30x / 45x | $132 / $172 | -$17 / +$22 |
| Astera's own beta | $118 | -$31 |

The simulated median in odds.json ($126) is 16% below the central. A quarter of the weight
sits on a bear-leaning world whose median ($71) is further below the central than the
bull-leaning world's ($186) is above it, and inside the base world the ranges for 2027
revenue and later growth reach further down than up (base-world median $139).

## 4. Market expectations and how the price reacted

Weekly closes (Robinhood, regular session): $112-120 through March 2026; a rise to $417 in
the week of 2026-06-15 (intraday high $499 in the week of 2026-06-29); $292-334 through
August and September; $342 in the week of 2026-10-05 after an intraday $402; $286 in the
week of 2026-10-12. On 2026-10-22 the official close was $257.82, and at 10:18 ET on
2026-10-23 the last trade was $261.24 (SOXX $476.71). The company's market value at the
price is about $45B (Alpha's data, 2026-10-23).

Running our model backwards (odds.json), the price is justified by revenue growth in
2028-2031 about 19 points a year faster than ours, which reaches about $10.8B in 2031: 1.2%
of our sector note's accelerator spend, against 0.47% today and our optimistic case's 1.1%;
in 2030 it is $8.4B, about 33% of management's $25B served market (Astera had about 34% of
its 2025 market). No other single driver gets there inside its bracket.

Alpha, our momentum model, ranks Astera 13th of about 700 on 2026-10-23: growth A,
momentum A, quality A, expectations B, valuation E (670th). It ranks the business's growth
and the stock's trend; we value the business. They differ because the price has run far
ahead of the value we find, which Alpha's own valuation grade also says.

Timing, a quick chart read: after a fivefold rise from the April low ($98) to the late-June
high ($499), the stock has made lower highs ($402 in early October) and is back near the
August-September lows of $251-266; below them the late-July low of $245 is the next level.
Grade D: a broken uptrend sitting on support. Timing is beside the view, never an input
to it.

## 5. Where we differ from the Street

The Street is below us for 2027-2028 and well above us after it. Its averages (market data,
2026-10-23) are revenue of $1.9B, $3.1B, $4.0B, $6.4B and $7.1B for 2026-2030 (ours $2.03B,
$3.49B, $4.37B, $4.80B, $5.28B) and non-GAAP EPS of $3.99, $6.49, $8.26, $14.60 and $16.80.
The 2030 revenue figure averages 15 analysts (low $6.0B, high $8.4B; secondhand, as relayed
by the advocate review): our $5.28B is below every one of them. Run through our model
(`street_check.py`, written by the model auditor):

| Step | Fair value |
| --- | ---: |
| Our central | $149 |
| Street 2026 revenue, our later levels | $149 (-$0.3) |
| Street revenue 2027-2031 (2031 +10%), our margins | $194 (+$45) |
| Street margins by year (non-GAAP operating margin 41.6%, 40.6%, 45.6%, 46.9%, implied by its EPS at a 12% tax with interest income removed) | $218 (+$23) |
| Exit at the peer multiple alone, without averaging in the 12.91% fade | $322 (+$104) |
| Street average target, discounted one year at our 12.91% | $328 (+$6 left) |

So about $68 of the gap to the Street is forecast (revenue after 2028 and cost leverage we
do not credit) and about $104 is how the exit is valued: our method averages the peer
multiple with a 15-year fade at 12.91%, and the Street's targets do not. On the method's
averaged exit the Street's path reaches its own target only at a 8.0% cost of equity.
In 2027-2028 the Street's implied margins are at or below ours; the disagreement is about
2029-2031.

What would settle it: 2029-2030 revenue, which shows first in how many customers take
Scorpio X switches during 2027 and in the concentration disclosures; and the cost ratio
over the next two quarters (25% of revenue with revenue above $3.5B would move us to about
$164).

## 6. Why the other side could be right

**The pessimistic case ($47)** is a world where AI spending pauses from 2028, as in our
sector note's pessimistic shape, and the lead customer shares its switch program with a
second supplier: 2027 growth of 35%, then -10% and -15%, gross margin 66% under price
pressure, cost cut but stock pay still near half of it. Nothing in the filings rules this
out: the concentration is extreme, Broadcom is a capable incumbent, and a single program's
timing has already driven a 40% quarterly step.

**The optimistic case ($288)** has switches broadening to several cloud customers, UALink
and optical arriving on time, and operating cost falling to 25% of revenue (a 45% operating
margin, the level Marvell targets for fiscal 2031). Revenue reaches $9.7B in 2031, about 30%
of management's served market. Astera's record of beating its own guide every quarter is the
best evidence for it.

Against our central from below: our growth after 2027 is mostly share gain, and share gains
in a customer base this concentrated come in lumps. From above, the advocate review's Buy
world: Credo, a smaller connectivity peer, already runs non-GAAP operating cost at 19.9% of
revenue ($95.2M on $479.0M, release 2026-09-01), while we hold Astera at 27.2% through 2031;
and every one of the 15 analysts covering 2030 is above our revenue. With 2028-2030 growth
of +40%, +25% and +15% (2030 at 28% of management's served market) and Credo-like cost, our
own model gives $259; at the industry beta $283. That world needs no change in the exit.
We do not adopt it: cost at a peer's ratio is not Astera's evidence, and the extra growth
rests on management's market claim and on Scorpio X customers not yet named in any filing.
**The one fact that would flip the call:** 2030 revenue on course for about $6.1B or more
(25% of the served market), shown during 2027 by Scorpio X in production at several
customers beyond the lead.

## 7. Next results: Q3 2026, on 2026-11-03 after the market close (confirmed by the company on 2026-10-06)

Each forecast starts from the company's guide and adds Astera's median gap to its guide over
the last eight quarters (`preview.py`). The ranges are wider than that record
alone, because our earlier ranges held only 12 of 29 outcomes, and because this quarter is
the largest step Astera has guided, driven by one product's first volume quarter.

| Metric | Our forecast | Range (about 80%) | Why |
| --- | ---: | ---: | --- |
| Revenue ($M) | 603 | 545-644 | Guide midpoint $550M plus the median beat of +9.6%; never below the midpoint in eight quarters |
| Non-GAAP gross margin (%) | 73.2 | 70.6-75.0 | Guide about 72% plus the median gap of +1.2 points; mix shifts toward lower-margin switches |
| Non-GAAP operating expenses ($M) | 164 | 149-176 | Guide midpoint $158M plus the median overrun of +3.9% |
| Non-GAAP diluted EPS ($) | 1.37 | 1.15-1.51 | Built from the three above with $12M other income, 12% tax, 185M shares (the EPS record alone gives 1.27) |
| GAAP diluted EPS ($) | 1.09 | 0.80-1.45 | Non-GAAP less $76M of stock pay in operating cost; 4% GAAP tax as guided |
| Stock-based compensation ($M) | 78 | 72-88 | Guide reconciliation: $76M in operating cost plus about $2M in cost of revenue |
| Q4 2026 revenue guide midpoint ($M) | 663 | 601-754 | Our Q3 plus 10%; earlier guides stepped 8-40% |

What a surprise is worth: our model carries Q4 2026 revenue at $727M. A Q3 that changes our
2027 view matters more than the quarter itself: each $100M of 2027 revenue above our
$3,493M, kept through 2031, is worth about +$3.90 a share.

## 8. Dates to watch, and what would change the view

- **2026-11-03, Q3 2026 results.** Revenue at or below the $550M midpoint, or a Q4 guide
  below $600M, would be the first break in the guidance record and would point toward the
  pessimistic 2027.
- **2027, UALink switches and CXL ramps** (management claims). A slip moves 2028 growth
  toward our pessimistic path.
- **Customer concentration in the 2026 10-K** (February 2027). The top end customer's share
  falling below 60% with revenue still growing would cut the main risk.
- **Gross margin below 70%** before 2028 would mean the mix shift is faster or pricing
  weaker than management's target allows.

## 9. What changed from the blind value, and the review

The blind value (commit 66154e9, pushed before we looked at the price) was $138, pessimistic
$46, optimistic $287. After the price, two fresh reviewers worked on draft cf57c4537: a
model auditor (the `reviewer` role) and an opposite-side advocate (the `architect` role),
both Claude Opus 5.5, run one after the other without seeing each other's findings. Each
numerical finding was rerun here; `review_bridge.py` reproduces the
accepted changes step by step.

| # | Finding (reviewer) | Answer | Effect |
| --- | --- | --- | ---: |
| 1 | Operating cost held at the guide's 28.7% on the guide's revenue, while our Q3 already carries the usual beat (27.2%): the model cut the margin from 47% in H2 2026 to 42% in 2027 while saying "no further leverage" (auditor) | Fixed: our Q3 forecast's ratio, an internal consistency fix, not a new view | +$9.54 |
| 2 | Leases deducted while rent stays in non-GAAP cost; a $60.3M tax refund due (10-Q), $269.6M of federal losses and $109.3M of federal research credits (10-K) left out; our reason for leaving the losses out was wrong, as the 12% non-GAAP rate excludes stock pay (auditor) | Fixed | +$1.45 |
| 3 | Existing-award schedule summed to $720M against $680.9M in the 10-Q (auditor) | Fixed | -$0.12 |
| 4 | street_check.py overstated the Street's margins (our tax, interest left in, one flat ratio) (auditor) | Fixed: replaced with the auditor's script; section 5 | none |
| 5 | Free cash kept to the exit earns 4.5% after tax but is discounted at 12.91% (auditor +$3.97; advocate +$4.21) | Disclosed: the shared method keeps cash to the exit; a method question for the CIO | not applied |
| 6 | Operating cost could fall toward Credo's 19.9%; 22% of revenue is $183, 20% is $196 (advocate) | Disclosed, not adopted: a peer's ratio, not Astera's evidence; our range runs 24-33% | not applied |
| 7 | Every 2030 Street estimate is above ours; Street 2029-30 revenue on our costs $195 (advocate) | Disclosed in section 5 | not applied |
| 8 | US semiconductor industry beta 1.52 gives $163 (advocate) | Disclosed in section 3; peer beta kept, set before the price | not applied |
| 9 | The odds tool needs each driver's median inside the middle of its range; the gross-margin top moved from +1.0 to +1.7 points, and the cost range's low from 25% to 24% after fix 1 | Format, after the price; moves only the odds | none |

Net: $138 to $149 (+$10.87), the call unchanged (Avoid). The two changes that moved the value
were found by the auditor on internal consistency and filings, and both raise the value.
Flags answered: 3 (value 43% below the price): no risk is charged twice beyond finding 5;
the gap is growth after 2028, cost leverage and the averaged exit. 4 (fade $137 against
outside view $397): the 12.91% discount rate inside the fade; the fade equals the outside
view only at about 6%. 7 (own beta $118): disclosed; the market's gap is not a beta gap, as
our central reaches the price only at a 6.75% cost of equity (advocate). 17 (0.46x the
discounted Street target): section 5. Not checked by either reviewer: the sector note's
accelerator path, the outside-view EV data, and the Citi conference transcript beyond press
summaries.

## 10. Sources

Primary, all saved in `data/` and masked for prices before reading: 10-K 2025 (filed
2026-02-20); 10-Q Q2 2026 (filed 2026-08-05); earnings releases Q1 2024 to Q2 2026; 8-Ks of
2024-12-16, 2025-10-22, 2026-01-20, 2026-02-10 (two) and 2026-06-04. Secondhand: call
transcripts Q4 2025 (2026-02-10), Q1 2026 (2026-05-05), Q2 2026 (2026-08-04) and the Citi
conference (2026-09-09), Quartr via TradingView. Market data statements and peers' betas, fetched
2026-10-23; SEC XBRL amortization for AVGO, MRVL, NVDA, CRDO and RMBS, fetched 2026-10-23.
Sector note `docs/research/sectors/semis-ai/note.md` (5.3, 2026-10-15).