---
ticker: "DELL"
company: "Dell Technologies Inc."
date: "2026-10-16"
view: "Avoid"
fair_value: 269.76
price: 461.295
price_source: "Robinhood last trade (intraday; prior official close $473.39 on 2026-10-15, SIP)"
price_time: "2026-10-16T19:20:50Z"
price_vs_fair_value: "+71.0%"
bear: 129.79
bull: 578.54
exit_price: 327.62
p_price_justified: "under 20%"
restated: "2026-10-26"
method: "case-spread-2026-10-24"
original_fair_value: 291.63
original_view: "Avoid"
page: "https://alpha.dming.si/research/2026-10-16-dell-2"
---

# DELL · Dell Technologies Inc. · Avoid

Research report as of 2026-10-16, partly unblinded. Page: https://alpha.dming.si/research/2026-10-16-dell-2

We value DELL at $270 a share. The report's $461.30 price (Oct 16, 2026) was 71% above fair value: Avoid.

_Restated Oct 26, 2026 (method case-spread-2026-10-24); the report's original value was $292, Avoid._

| Lens | Range | Marks |
| --- | --- | --- |
| Our report | $130 to $579 | fair $270; original $292 |
| Our simulation | $157 to $554 | median $283 |
| Street targets (26) | $289 to $735 | average $565, median $570 |
| Price | | $461.30 |

Bands run from our bear to bull (the hollow mark is the original $292), 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 16, 2026.

| Figure | Value |
| --- | --- |
| Fair value | $270 |
| Price | $461.30 (Robinhood, Oct 16, 12:20 PM PT) |
| Five-year return | -1.6% (a year, at the report's price) |
| Bear / bull | $130 / $579 |
| Exit price | $328 (on Jan 30, 2032) |
| Confidence | 35% |
| Price against our scenarios | The price is justified in under 20% of our scenarios (20,000 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:

- AI server sales, share of AI chip spending: needs 44.2%, we have 17.0%
- AI server margin: needs 13.9%, we have 5.0%
- AI chip spending path (low -1, high +1): needs 2.38, we have 0

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

As of 2026-10-16 · partly unblinded, two ways (explained below the box) · reviewed once
(section "Review round")

**Original call, Oct 16, 2026 (before the Oct 26, 2026 restatement)**

```
Fair value today      $292 per diluted share (cost of equity 9.96%, the yearly return
                      investors require); $279 at the blind commit, before the review
                      round (section "Review round")
Price                 $461.295, Robinhood last trade, 2026-10-16T19:20:50Z (intraday; official
                      2026-10-15 close $473.39, SIP)  →  fair value 36.8% below the price
Five-year exit price  $346 on 2032-01-30 (mean of the long fade $333 and the outside view $359;
                      at 10/20-year fades $342/$349; value-driver formula 15.6x against
                      the fade's 15.2x year-six NOPAT)
Five-year return      -0.7% a year from today's price on our central case, payouts included
                      (five_year_return.py); below the 12% hurdle (the yearly return we require)
Purchase price        $269 to earn the 12% hurdle
Bear / bull today     $160 / $508, each a coherent operating case (case-weighted $313,
                      25/50/25, our weights)
Confidence            35% that value on 2032-01-30 lands within ±20% of $346 (subjective, not
                      calibrated)
Moat                  scale buyer and integrator of servers, storage and PCs with a large enterprise
                      installed base and its own financing arm; thin margin on AI racks, and
                      elsewhere its prices follow memory prices
The disagreement      Whether earnings at the Street's FY30 level (its FY30 EPS implies segment
                      income of $37.6B, 12.6% of its revenue; ours is $19.9B) are the cycle peak
                      or the new base. Held as a base from FY30 they are worth $557-581 a share,
                      above the price; as a peak that fades with the memory premium, $292. The
                      quarterly observable is the AI server margin: Dell's 10-Q (2026-09-08)
                      expects "margin rate pressure" from the mix shift to AI servers, TD Cowen
                      (2026-09-29, as reported) argues richer racks. It is the flip assumption,
                      the one that would change our view: about 10.2% against our 5.0% lifts fair
                      value to the Hold line ($392)
Market expects        A normal server and storage margin of 35.3% (our 15.5%), or Dell taking
                      43.7% of AI accelerator spend (our 17.0%), or a 13.8% AI server margin (our
                      5.0%), or an accelerator spend path far above the sector note's bull
                      (position 2.38 on a scale where its bull is +1), or memory staying short
                      2.8 years longer than the house path; each alone, everything else central
Price needs           One of: normal server and storage margin 35.3% (central 15.5%); AI server
                      share of accelerator spend 43.7% (central 17.0%); AI server margin 13.8%
                      (central 5.0%); accelerator path at 2.38 (bear -1, proposed 0, bull +1);
                      memory normalisation 2.78 years later (central 0). Four more cannot reach
                      the price alone: traditional server volume (best end +30% a year gives
                      $429), PC margin (20% gives $395), mature return (80% gives $296) and
                      growth after FY2032 (+9% gives $321)
Odds in our scenarios 8.7% that fair value reaches the price (P(price justified); subjective, not
                      calibrated): simulated median $320, 10th to 90th percentile $208 to $453;
                      2.8% in the base world, under 1% in the bear-leaning one and 29% in the
                      bull-leaning one (10th to 90th $357 to $514). A diagnostic of our
                      scenario spread, not the odds of a 12% return
Key driver            AI server margin: 3.5% / 8% moves fair value -$29 / +$58
Alpha model           Capfolio's stock-ranking model: rank 10 of 713, score 98.7 (Strong);
                      momentum A, expectations A, growth B, valuation B, no quality grade in
                      the row (as of 2026-10-16)
Street                Mean target $564.65, median $570, range $289-$735 (Street consensus, read
                      2026-10-16): 3 analysts in the last month at $578.33, 26 in the last
                      quarter at $559.85, 59 in the last year at $420.53, so targets are rising.
                      Targets forecast a future price, so they show expectations, not
                      validation or value today
```

**We rate Dell Avoid at $461.295: our central fair value is $292 a share,
36.8% below the price.** The price is above the $269 purchase price that earns
our 12% yearly return, and fair value is below 0.85 times the price ($392.10),
our line for Avoid. The whole disagreement with the price is whether the
earnings the Street expects from FY30 on are a cycle peak or the new base. Our
central case treats them as a peak that fades as the memory-price premium
ends: $292 a share. Held as a base they would be worth $557-581.

This report revises [DELL 2026-10-15](https://alpha.dming.si/research/2026-10-15-dell), which
stays as published. Model: model.py, output
output.txt, blind record assumptions.md (blind
commit `aa12f11`, not edited since), bridges bridge.py and
bridge-output.txt, market-side checks flip.py
and flip-output.txt.

**Terms.** Fair value is our estimate of what a share is worth today. We value
each company three ways. The central case is our best estimate, the bear case
is a severe but plausible bad outcome, and the bull case is the matching good
outcome. FY is Dell's fiscal year, which ends at the end of January: FY27 is the
year to January 2027, and the exit date, 2032-01-30, ends FY32. Dell reports two segments: ISG (Infrastructure Solutions
Group: servers, networking and storage) and CSG (Client Solutions Group: PCs).
Segment income is those segments' operating profit before stock pay and
corporate costs; EBIT subtracts those; NOPAT is operating profit after tax,
and DFS is Dell Financial Services, the financing arm. AI servers are racks of
other firms' accelerators (AI chips), memory and networking that Dell
integrates and ships. The sector note is our research note on semiconductors
and AI; the house memory path is our forecast of memory prices from our latest
Micron report (2026-10-15). The fade values Dell after FY32 by letting growth
and the return on new investment drift over 15 years to long-run levels; the
outside view applies the multiples peers trade at; the value-driver formula
is a third estimate of the exit value. The bridge turns the value of the
business into value per share: it adds cash and receivables and subtracts
debt and other claims. The Hold line is 0.85 times the price: below it the
view is Avoid. The Street means the analysts who follow the stock, and the
blind commit is the model we committed before reading any market data.

All figures are the model's (`output.txt`), per diluted share of 652M.

Partly unblinded in two ways. No quote, multiple, estimate, target or rating
for DELL was read in the blind step. But the analyst who wrote this revision
read the old report's `assumptions.md`, which records an average buyback price
of about $400 a share for May-July 2026 and insider and Silver Lake sales that
imply about $530-560 a share in mid-September 2026; and it read our latest
Micron report's `output.txt` for the house memory path, which prints MU at
$931.40 and MU's Street revenue and EPS estimates (MU is not in the peer set).
Both are price levels; neither was used as an input.

## What changed and why

Our chief investment officer's evaluation of the old report (2026-10-16) found that the old central case was a downturn
dressed as a central: revenue fell from $235.5B (FY28) to $201.5B (FY30) on an
AI server dip and a memory price fall that no source dated, and the server
and storage margin ended at 15%, below the 16.3% Dell earned in FY2026. This
revision builds the central as the most likely path normalised through the
cycle, with every timing tied to a dated source:

- **AI servers do not dip.** They hold Dell's FY27 share (17.0%) of the
  sector note's proposed accelerator spend: $127B, $138B, $138B, $143B,
  $152B in FY28-32. The old $95B in FY30 had no dated cause.
- **Memory is modelled as what it is.** Each non-AI line is revenue at
  normal component prices, plus memory content passed through at 5%, plus a
  pricing premium Dell earns while memory is short. Solved on Q2 FY27 against
  Q2 FY26: memory at 3.46x normal and a premium of 18.5% of revenue in
  servers and storage; 2.84x and 1.6% in PCs. Two segments landing on a
  similar memory index is the check that the structure fits.
- **The fade of memory and premium is dated.** H2 FY27 is solved to the
  guide; from there the excess follows the house memory model (our Micron report): up 20%
  in FY28 (Michael Dell, 2026-09-09: the shortage is "probably worse in 2027
  than in 2026"), then back to normal by FY31-32 as new fabs land from 2028
  (TrendForce, in the sector note). The Micron report's fab table closes the shortage at
  the end of 2030, which is where Michael Dell put it on CNBC on 2026-10-13
  ("Maybe it's a four-year problem because there's just not enough supply",
  tight through 2030; a management claim, read in a Stocktwits write-up, not
  the transcript), so the path sits inside that claim. Micron (2026-09-30,
  press summaries of the call) has demand above supply through calendar 2028
  with new fabs from 2028. The old report's "about 2030" had no date; the
  blind step did not find it again, and the CNBC interview is the passage.
  Memory never goes below normal in the central; the bust is the bear's.
- **Normal margins come from history.** Servers and storage 15.5% (about the
  FY25-26 mean on our split), PCs 6.4% (the FY24-26 mean).
- **The method is Capfolio's current one** (`fade-peer-2026-10-16`):
  exit at the mean of the long fade and the outside view, cost of equity from
  the dated pair in `method.py`.
- **The new debt is in the bridge from business value to value per share.** The $5.0B of notes issued 2026-09-15
  (5.10% 2029 to 5.90% 2037) and the $1.75B of 4.900% First Lien Notes due
  2026-10-01 they repay are both counted, from the 424B2's as-adjusted
  capitalization: cash $14.805B, debt $37.997B at principal. Net debt is
  unchanged but for $36M of fees; interest rises on the larger gross debt.
- **EBIT is defined.** The old output's "op margin" was segment income less
  stock pay and corporate costs, over revenue; that is why it sat about one
  point under segment income over revenue. This output prints both, named.
- **After review (one round, section "Review round").** Shareholders now
  receive 80% of adjusted free cash flow, Dell's stated policy, instead of
  dividends only (+$13.82); the bridge drops leased equipment (its income is
  already in NOPAT), adds long-term investments and deducts operating lease
  liabilities (net -$1.21); the bear and bull share the central's growth
  after FY32 and mature return. Central $279.02 to $291.63; bear $145.66 to
  $160.17; bull $515.83 to $507.50. The view is unchanged.

The central now shows revenue of $247.5B (FY28), $256.7B, $243.1B, $244.3B
and $254.3B (FY32): the FY30 dip is memory prices leaving revenue, not lost
business, and segment income falls from $25.7B (FY29) to $18.3B (FY32) as
the pricing premium ends. EPS after stock pay runs $28.03, $28.71, $21.51,
$19.21 and $19.60 in FY28-32 (before stock pay $29.04, $29.78, $22.65,
$20.41 and $20.85), on today's 652M shares: the model counts buybacks as
cash paid to holders at fair value, so the share count it prints does not
fall. At the central's values the FY28-32 buybacks of $49B would retire
roughly a fifth of the shares (our inference), which would lift EPS by about
that much. The old report's "about $18-20 by FY30" did not match its own
model ($17.58).

## Changes from the 2026-10-15 report, step by step

Old central to the blind commit's, cumulative, per diluted share, all at
2026-10-16 (`bridge.py`; steps 1-5a rerun the old model with patched inputs,
5b runs the old forecast through the shared method, 6-8 rerun this model as
committed blind (`aa12f11`) with old inputs put back one at a time):

| Step | Change | $/share | Value |
| --- | --- | ---: | ---: |
| | Old central fair value (published 2026-10-15) | | 183.38 |
| 1 | Valued one day later | +0.05 | 183.43 |
| 2 | $5.0B notes and the 2026 Notes repaid; debt at principal, not carrying value | -0.52 | 182.91 |
| 3 | H2 FY27 cash: amortisation not added back twice, stock pay after tax, H2 dividends paid to holders | +1.23 | 184.13 |
| 4 | Corporate costs $1.2B to $0.7B a year (FY24-26 average $0.74B); 652M shares | +6.08 | 190.21 |
| 5a | Cost of equity 10.7% to the method's 9.96% (old 12x exit kept) | +6.66 | 196.87 |
| 5b | Exit by the method on the old FY32 forecast: fade $358 and outside view $428, against 12x NOPAT | +54.93 | 251.80 |
| 6 | Memory pass-through and premium fading on the house path; this revision's volumes, receivables and payouts (old AI path and end margins kept) | -10.85 | 240.95 |
| 7 | Normal margins from history: servers and storage 15% to 15.5%, PCs 6% to 6.4% | +7.14 | 248.09 |
| 8 | AI servers at 17% of the accelerator path, no FY30 dip | +30.93 | 279.02 |
| | **Total to the blind commit** | **+95.64** | **279.02** |

Most of that change is method (steps 5a-5b, +$62) and the AI path (+$31).
Step 6 is a residual of the new structure against the old, not one input: it
nets the longer-lasting premium against the end of FY28's inflated revenue
and the different interim cash.

Blind commit to the reviewed model (`bridge.py`, second block; R1 to R5 are
the review-round changes):

| Step | Change | $/share | Value |
| --- | --- | ---: | ---: |
| | Blind commit `aa12f11` | | 279.02 |
| R1 | Fees on the new notes: the $14M discount is already in the 424B2 cash, the $22M underwriting discount and $13.6M expenses are not | 0.00 | 279.02 |
| R2 | Leased equipment $3.182B out of the bridge | -2.95 | 276.07 |
| R3 | Long-term investments $2.679B in | +2.49 | 278.55 |
| R4 | Operating lease liabilities $805M out | -0.75 | 277.81 |
| R5 | Payouts: 80% of adjusted free cash flow | +13.82 | 291.63 |
| | **Total, review round** | **+12.61** | **291.63** |

Bear moves from $82 (old) to $146 (blind) to $160, and bull from $434 to $516
to $508 (the bull falls because it now shares the central's growth after FY32
and mature return); the exit price from $293 to $438 to $346 (it falls
because the cash that used to pile up to the exit is now paid out earlier).

## Business economics

Dell sells three things at three economics. AI servers ($74B guided for
FY27) are racks of other firms' accelerators, memory and networking,
integrated and shipped; Dell earns "mid-single digits" on them. Traditional
servers, networking and storage ($36B in FY26) earned 13-16% on our split,
because Dell owns the storage IP, the enterprise relationships and the
service attach. PCs ($51B) earn 5.6-7.6%.

Since late 2025 memory, which Dell buys and passes through, has repriced
the whole non-AI book. Traditional server revenue more than doubled in Q2
FY27 "primarily" on average selling price (10-Q), and PC revenue rose 20%
with units down. Dell repriced its server business in days in December 2025
(Q4 FY26 call) and in Q2 FY27 earned about 26% on servers and storage,
against about 12% in the same quarter a year before (16.3% for FY2026). That
premium is earned while memory is short and leaves with the shortage;
memory content leaves revenue at its thin margin.

Financing matters. Financing receivables rose from $14.3B to $20.4B in six
months and "adjusted" free cash flow adds the $6.4B of new receivables back
(Q2 release); the model counts them as cash spent and adds the book back at
exit less losses. Purchase obligations are $22.4B for the rest of FY27 and
$12.4B after; supplier guarantees of $8.6B carry full recourse (10-Q Note
10). Michael Dell and Silver Lake control the votes; Dell moved to Texas on
2026-07-01, which raises the bar for derivative suits to 3% holders (8-K).

## Recent results against Dell's own guidance

Q2 FY27 (to 2026-07-31): revenue $47.0B, ISG income $4.78B (15.0%), CSG
$1.14B (7.6%), AI orders $60.9B and backlog $95B. Dell raised its FY27
revenue guide from $167B to $192B, AI servers from $60B to $74B and non-GAAP
EPS from $17.90 to $25.50 (release, 2026-09-01). Solving H2 to that guide
gives a premium at 0.79x Q2's and a memory excess at 0.71x Q2's: the guide
already assumes some of Q2's pricing eases, matching the CFO's warning that
not every benefit continues.

That sits against the house memory path, which rises to 5.28x normal by the
end of 2026 (our Micron report). The 0.71x does not depend on the memory share we assume
(it is 0.71x at 12%, 20% and 30% of server and storage revenue, `output.txt`),
so either the guide is low or H2 volumes differ from our flat-to-down units;
the model cannot tell which. Two quarters of Dell's guide record are
measured, both from press summaries of the releases, not the releases:
Q2 FY27 guided revenue of $44.0-45.0B and non-GAAP EPS of $4.80 at the
midpoint (Q1 release, 2026-05-28) and delivered $47.0B and $7.04 (+5.6% and
+47% against the midpoints); Q1 FY27 guided $35.2B and $2.90 (StorageNewsletter,
not checked against the release) and delivered $43.8B and $4.86; the FY27
revenue guide went from $140B to $167B to $192B in two reports. The other
six of the last eight quarters were not measured. Dell has beaten its
guide in the two quarters measured, so the guide may be low; the central
keeps it because it is the one dated forward figure. If H2 held Q2's excess
and premium, FY27 EPS would be $27.86 (+9%) and fair value $303.17
(+$11.54).

## Forecast and valuation: how we reach $291.63

| FY | Revenue | Segment income | EBIT | NOPAT ex DFS | EPS after stock pay |
| --- | ---: | ---: | ---: | ---: | ---: |
| 2027 | 191.3 | 21.5 | 20.2 | 15.9 | guide $25.50 before stock pay |
| 2028 | 247.5 | 25.2 | 23.7 | 18.7 | 28.03 |
| 2029 | 256.7 | 25.7 | 24.2 | 19.0 | 28.71 |
| 2030 | 243.1 | 19.9 | 18.3 | 14.2 | 21.51 |
| 2031 | 244.3 | 18.1 | 16.4 | 12.6 | 19.21 |
| 2032 | 254.3 | 18.3 | 16.6 | 12.8 | 19.60 |

US$ billions except EPS; revenue excludes the small corporate line; EBIT is
segment income less stock pay and other corporate costs, before acquired
amortisation; NOPAT excludes DFS income, which is valued with the
receivables in the bridge; EPS on 652M shares.

At year five servers and storage earn 15.5% and PCs 6.4% at normal memory
prices: the business normalised, not in a trough.

**The fade's start.** The long fade starts at a volume-only return on new
capital of 158.6% and fades to 25.6%, the midpoint of DELL's FY2021-26
returns with and without goodwill. The start is the one our valuation method prescribes
(`valuation.md`): the profit of the new normal-price revenue between FY27 and FY32 at
each line's normal margin (AI servers at their 5%), after stock pay and tax
($4.6B), over the net fixed and intangible reinvestment of FY28-32 ($2.9B).
That reinvestment is our assumption, 6% of the rise in normal-price revenue
(AI servers counted at 30%) plus $0.1B a year; it has no asset life in it.
Its check is history: Dell's capex ran $2.6-3.0B a year in FY2022-26 (market data)
against depreciation and amortisation of $3.0-4.6B, so growth took about no
net fixed capital. The limit is that H1 FY27 ran faster: capex of $2.2B in
six months ($4.4B a year, market data quarterly cash flow) against about $2.7B of
depreciation (our estimate from the annual figure). At $1.2B a year more net
reinvestment the start falls to 52% and fair value to $281.96 (-$9.67);
starting the fade at the mature return would give $285.98 (-$5.65). The
fade is 15.2x year-six NOPAT and the value-driver formula 15.6x.

**The outside view,** at the peers' latest EV over trailing GAAP EBIT
(servers and storage 23.4x, the NTAP and HPE median; PCs 9.9x, HPQ; AI
servers 7.2x, SMCI), gives $359, 8% above the fade ($333). At the peers'
ten-year medians (13.9x, 8.8x, 12.7x) it would be $320 and fair value $280
(-$11.67). The latest multiples sit above their medians at the two server
peers and the exposure is large: those multiples are at or near cycle highs.
Two of the four trailing EBITs are unrepresentative. HPE's trailing four
quarters of GAAP EBIT ($3.46B) include a 4.4% margin quarter beside the
latest quarter's 11.4%; the latest quarter times four ($5.57B) would put HPE
at 15.6x, not 25.2x. SMCI's latest quarter, at a 13.4% margin after 6.1%,
3.7% and 3.6% (its best year in ten is 10.7%), holds $1.49B of its $2.77B
trailing EBIT; at the prior three quarters' 4.6% margin the trailing EBIT
would be $1.79B and the multiple 11.1x, not 7.2x. Using both run-rate
multiples (core 18.6x, AI servers 11.1x) gives fair value $289.94 (-$1.69),
because the lower server multiple and the higher AI multiple offset. The old text said the latest
multiples sat above their medians "not on rates"; the yield was higher
(5.29% now against about 2.5% on average over the ten years), which would
pull multiples down, but we did not test what explains them, so the phrase
is cut.

**The bridge at exit** adds $14.9B: cash $26.8B (the 20% of adjusted free
cash flow not paid out, kept at 3.5%), debt -$38.0B, receivables $24.2B net
of 3% losses, long-term investments $2.7B, operating lease liabilities
-$0.8B. Payouts are 80% of adjusted free cash flow, Dell's stated policy to
return over 80% (Securities Analyst Meeting, 2025-10-07, 8-K exhibit):
$6.0B in H2 FY27 and $15.1B, $14.8B, $10.6B, $9.7B and $10.1B in FY28-32,
counted as dividends ($0.63 a quarter, +10% a year) plus buybacks at fair
value, which are worth the same to a continuing holder as cash paid. They
are $82.32 of the $291.63 (28%); the exit price is the other $209.31. They
run well ahead of the remaining buyback authorization (roughly $10B at
2026-07-31: $15.2B after the 2026-02-26 approval less $5.4B repurchased in
H1, our arithmetic, 10-Q and market data), so the model assumes new authorizations;
$30B has been authorized since 2021 (10-Q). The payouts are 28% of the
value, against the 57% that `valuation.md` records for the payouts in an earlier
Micron report (MU-2) at five times its authorization; here the buybacks of $49B over FY28-32 are
about five times the remaining authorization, the pattern that paragraph
warns about. The capped case is the figure to weigh against the central: at
$6B a year of buybacks (the $30B authorized since 2021 over five years)
fair value is $285.26 (-$6.37). Cash earns 3.5%: Dell's interest income over average
year-end cash was 4.1%, 2.9% and 3.7% in FY2024-26 (market data; 3.6% pooled), below
its 5% on debt because much of the cash probably sits offshore in short
deposits (our inference); at 5.0% fair value is +$1.31.

Where this departs from the sector note: the accelerator path is adopted;
memory follows our latest Micron report (which the note says replaces its
provisional MU v2 path); traditional server volume grows 6% a year, not the note's server CPU
unit path (+21, +12, +6, +4, +3%), because Dell's own units fell in FY27 and
its recovery waits on deferred enterprise refreshes; the note's path would
add $15.

Sensitivities (central fair value, one at a time, from `output.txt`):

| Driver | Low / high | Fair value |
| --- | --- | --- |
| AI server margin | 3.5% / 8% | $262 / $350 |
| Normal server and storage margin | 13% / 17.5% | $271 / $309 |
| AI share of accelerator spend | 14% / 20% | $272 / $311 |
| Memory path | one year earlier / later | $281 / $310 |
| Traditional server volume | 2% / 10% a year | $278 / $308 |
| Normal PC margin | 5% / 7% | $281 / $296 |
| Growth after FY32 | 2% / 6% | $282 / $302 |
| Mature return | 14.3% / 36.9% | $287 / $293 |
| FY32 NOPAT (fade only) | -10% / +10% | $282 / $301 |
| Beta | 1.0 / 1.25 / 1.5 | $335 / $302 / $277 |
| Without the -1.0 point calibration | cost of equity 10.96% | $270 |

The review round added these (also in `output.txt`):

| Change | Fair value | vs central |
| --- | --- | --- |
| AI share 15% / 13% (central 17%) | $278.67 / $265.63 | -$12.96 / -$26.00 |
| Accelerator path flat at $749B / the note's bear / the note's bull | $276.61 / $259.51 / $363.46 | -$15.02 / -$32.12 / +$71.83 |
| AI server margin 11% | $407.23 | +$115.60 |
| Payout dividends only / 56% / 100% / buybacks capped at $6B a year | $277.81 / $286.86 / $295.53 / $285.26 | -$13.82 / -$4.77 / +$3.90 / -$6.37 |
| Shares 642.8M / 635.8M (10-Q cover, with and without Q2's 7M dilutive) | $294.00 / $295.86 | +$2.37 / +$4.22 |
| Net reinvestment $1.2B a year higher | $281.96 | -$9.67 |
| Fade starts at the mature return | $285.98 | -$5.65 |
| Peers' multiples on run-rate EBIT | $289.94 | -$1.69 |
| Memory path 0.5 / 2 years later | $300.61 / $365.38 | +$8.98 / +$73.75 |
| H2 FY27 at Q2's memory excess and premium | $303.17 | +$11.54 |

Beta is Damodaran's Computers/Peripherals 1.35, as recorded in the HPE
report. Each risk is charged once: the AI air pocket and the memory bust
live in the bear's revenue and margins, and its multiple is the fade's on
that bear business. The long-run levers (growth after FY32 4%, mature return
25.6%) are the same in all three cases, so the cases differ only in their
operating worlds.

## Bear case and why the other side could be right

**Bear, $160.** AI orders fall after the backlog ships (the sector note's
bear accelerator path, Dell's share sliding from 15% to 13%, margin 3.5%),
memory busts below normal in FY31-32 with Dell left holding inventory bought
high, normal margins of 13% and 5%, 12% of receivables never collected. Its
year-five NOPAT is $6.2B. The bear's outside view ($193) is 25% above its
fade ($154): the peers' latest multiples price today's cycle, which the bear
business no longer has; we believe the fade.

**Why the bull could be right, $508.** Michael Dell's shortage lasts to
about 2030 and keeps pricing firm through FY31, enterprise and sovereign AI
broadens the book at 8%, and AI servers follow the sector bull path. The
model rewards a later memory normalisation (+$18 for one year later). The
bull is 10% above the price.

**What argues against this central.** It holds 17% of a path that still
grows after 2027, at a 5% margin no one outside Dell can check, with one
customer at 12% of FY26 revenue and lower-grade financing receivables up
from $1.8B to $4.4B in six months (10-Q Note 4, as recorded in the old
report). Dell's AI servers were 3.1%, 7.1%, 11.4% and 18.1% of NVIDIA's
revenue in FY2024-27, a share that rose every year, and the sector note
itself says its rising accelerator path and falling memory price cannot both
hold. And the step that moved value most is the method's exit, not new
evidence about Dell: on the old forecast alone the method adds $55.

## What would prove us wrong, and the dates to watch

- Every quarter: the AI server margin, the flip assumption. Dell does not
  disclose it; ISG operating margin against the AI share of ISG revenue is the
  nearest read. The 10-Q (2026-09-08) expects "margin rate pressure" from the
  shift in mix toward AI-optimized servers; TD Cowen (2026-09-29, as
  reported) argues richer racks. About 10.2% against our 5.0% lifts fair
  value to the Hold line.
- Q3 FY27 results (late November 2026): revenue against the $49.0B guide;
  ISG margin against Q2's 15.0% tests whether the premium is already easing
  as the guide implies.
- Q4 FY27 and the FY28 outlook (about late February 2027): an FY28 AI
  server outlook below about $110B (86% of our $127B) falsifies the
  share-hold.
- Memory: two quarters of falling DRAM contract prices before the end of
  2027 bring the bear's timing forward (sector note M8, M13).
- Financing receivables above $30B, or lower-grade receivables above half
  the book, with AI orders falling, is the bear's credit leg.

## Sources

10-Q for the quarter to 2026-07-31 (filed 2026-09-08); Q2 FY27 release
(8-K, 2026-09-01); 10-K FY2026 (filed 2026-03-16); 8-Ks of 2026-07-01,
2026-07-06, 2026-09-10 and 2026-09-15; 424B2 of 2026-09-11; Q4 FY26 call
(2026-02-26) and Michael Dell at a conference (2026-09-09), Quartr via
Bigdata.com; market data statements and peer metrics read 2026-10-16 (`data/`);
sector note semis-ai and our Micron report (2026-10-15). Filings were masked with
`scripts/research-mask.py` before reading; the masked copies are in `data/`.
Added in the review round, after the blind commit: Dell Securities Analyst
Meeting presentation (8-K exhibit, 2025-10-07, SEC EDGAR); Q1 FY27 release
(2026-05-28) and Q4 FY26 release, through press summaries; Michael Dell on
CNBC (2026-10-13, Stocktwits write-up, not the transcript); Micron Q4 FY26
call (2026-09-30, press summaries); TD Cowen note on Dell (2026-09-29, press
summaries). Several of these are secondary write-ups and are labelled where
used.

## Market pass (2026-10-16)

The market pass is the step in which we first read the price and the Street. A separate step that checks the analyst's work ran it after the blind commit (`aa12f11`) and reran it after the review round: the model, forecast and figures above are the reviewed ones; `assumptions.md` is unchanged.

**Price and view.** DELL $461.295 (Robinhood last trade, 2026-10-16T19:20:50Z; the official close on 2026-10-15 was $473.39, SIP). The price is above the purchase price ($269.15), so not a Buy; fair value $291.63 is below 0.85 times the price ($392.10), so the view is Avoid. The central case returns -0.7% a year to the exit, from `five_year_return.py` on the saved `forecast.json` (it was -0.2% before the review round: exit $438 with $85B of retained cash, now $346 with the cash paid out).

**What the price requires** (`alpha research odds`, `odds.json`): a reverse valuation (working back from the price), one driver at a time with the rest central. Each of the five headline drivers needs a value far outside its range: a normal server and storage margin of 35.3% against 15.5% (Dell earned 16.3% in FY2026), AI server share of accelerator spend of 43.7% against 17.0%, an AI server margin of 13.8% against 5.0% (management says mid-single digits; the server builders' best year is 10.7%), an accelerator path at 2.38 on a scale where the sector note's bull is +1, or memory staying short 2.8 years beyond the house path. Traditional server volume, PC margin, mature return and growth after FY2032 cannot reach the price alone inside their bounds. Even the bull case, $507.50, is only 10% above the price.

**Odds.** 8.7% that fair value reaches the price, under our scenario spread (20,000 draws, seed 20261016; subjective, not calibrated). It was under 1% before the review round, partly by construction: the accelerator path was fixed and the AI margin capped at 8%. The weights are 50% base (P 2.8%), 25% bear-leaning (under 1%) and 25% bull-leaning (29%), so most of the mass comes from the bull-leaning world; each driver's range ends below the value the price needs, so the 8.7% is joint moves only. The simulated median is $320, 9.6% above the central, and the 10th to 90th is $208 to $453, so the price sits just above our 90th. Variance explained alone: AI server margin 70%, AI share 61%, accelerator path 56%, normal server margin 45%, memory timing 42%.

**Fair value at the company's own beta.** Not part of the analyst's work. Computed here as the OLS slope of DELL's weekly returns on SPY's over the last 104 weeks, from the stored adjusted closes (`readonly-data-run.sh`, `BetaService`, last close 2026-10-15; the same weekly raw beta the repository uses elsewhere): 1.94 (R-squared 0.21, 104 weeks), against the model's 1.35. Putting it into the same forecast and method, the same code that gives $291.63 at 1.35, gives cost of equity 12.44%, fair value $243.50, purchase price $247.57 and exit $306.58. Blume-adjusted (0.67 times the raw beta plus 0.33, beta 1.63) it gives $266.06. The debt weight stays at the analyst's 10%. At either beta the price is far above fair value and the purchase price, so the Avoid does not depend on the choice. The raw beta fits poorly (R-squared 0.21). (Before the review round: $227.72 and $251.91.)

**What Capfolio's ranking model (Alpha) says** (`alpha rank DELL --raw --json --allow-stale-model-cache`, read in place, 2026-10-16; cache status partial, integrity verdict WARN with 1 warning that was not reviewed). Rank 10 of 713, score 98.7, tier Strong. Category grades: momentum A (rank 1), expectations A (47), growth B (81), valuation B (142); the row has no quality grade (coverage 80%). Why we differ is our inference and was not tested: Alpha's top grades are momentum and expectations, which describe what the stock has done and where estimates have gone, while this report values the cash flows after the memory premium fades and finds the price needs earnings to double again. A top rank and an Avoid can both hold, because the rank does not say whether the price already reflects the momentum.

**Street** (market data, read 2026-10-16). Mean target $564.65, median $570, range $289 to $735: 3 analysts in the last month at $578.33, 26 in the last quarter at $559.85 and 59 in the last year at $420.53, so targets are rising. The mean is 22% above the price and 1.9 times our fair value. Consensus revenue is $194.6B for FY2027 (18 analysts), $235.2B for FY2028 (20), $263.6B for FY2029 (18) and $298.4B for FY2030 (17), against our $191.3B (FY2027; $192.0B with the corporate line, the guide), $247.5B, $256.7B and $243.1B (the vendor's year labels are read as Dell's fiscal years: our inference from the FY2027 level). We sit 1% to 2% under the Street in FY2027, 5% over in FY2028, 3% under in FY2029 and 19% under in FY2030, where our memory premium has faded and the Street's revenue keeps rising (our inference about the gap; the Street's drivers are not published). Consensus non-GAAP EPS is $25.89, $30.93, $35.39 and $45.78 for FY2027-30 (14, 8, 5 and 3 analysts), against our EPS before stock pay of $29.04, $29.78 and $22.65 for FY2028-30 on today's 652M shares (the FY2027 guide is $25.50): 6%, 16% and 51% under. The Street's FY2030 EPS implies segment income of $37.6B, 12.6% of its revenue, against our $19.9B (`flip-output.txt`). The bull case's revenue ($296.3B in FY2029, $325.3B in FY2030) is above consensus; its EPS before stock pay is above in FY2029 ($36.21) and below in FY2030 ($39.15). Targets are an expectations reference, not validation.

## Review round (2026-10-16)

Three reviewers read the report: research-eval, which checks the basics
(math that reruns, sourced figures, the bridge), a model auditor, who looks for
errors, and an advocate, who argues the opposite side from ours; one round,
under our review procedure (`references/review.md`). Every finding is answered below as
fixed, disclosed or rejected. An assumption changed only on dated evidence
that does not depend on the price. The operating assumptions (AI share,
memory timing, margins, volumes) were not changed; what changed is the
payout convention, the bridge, the bear and bull levers, the odds drivers
and the disclosures. Dollars are per diluted share of the central fair value
(blind $279.02), from `output.txt` and `bridge-output.txt`.

Files. Among the blind-commit files (`aa12f11`), `model.py` changed and
`output.txt`, `forecast.json` and `valuation.json` were regenerated from it;
`assumptions.md` and `data/` are untouched. Changed outside the blind
commit: `drivers.json`, `odds.json`, `record.json`, `bridge.py`,
`bridge-output.txt`, this report, `process.md`; new: `flip.py`,
`flip-output.txt`, `street_path.py` and `street_path.txt`.

| # | Finding | Answer | Before → after |
| --- | --- | --- | --- |
| 1 | Free cash flow piles up to $85B at exit at 3.5%, discounted at 9.96% | **Fixed.** Payouts are now 80% of adjusted free cash flow (dividends plus buybacks at fair value), Dell's stated policy "over 80%" (Securities Analyst Meeting, 2025-10-07, 8-K exhibit; dated, independent of the price). H1 FY27 returned $6.3B against $11.3B adjusted free cash flow (56%, buybacks $5.4B; Q2 release and market data), so 80% is a policy figure above the latest half: at 56% the value is $286.86 (-$4.77), at 100% $295.53 (+$3.90), capped at $6B a year of buybacks $285.26 (-$6.37). Cash at exit falls from $85.4B to $26.8B and the exit price from $438 to $346 | $279.02 → $291.63 (**+$13.82**); the expected +$49.67 is **not reproduced**: even paying out 100% adds $17.72 over dividends only. The gain is the gap between the 9.96% cost of equity and the 3.5% the retained cash earned, on cash held about three years on average (our reading of the arithmetic); the auditor's calculation was not in hand |
| 2a | Memory timing: the CNBC 2026-10-13 quote, Micron 2026-09-30, the old report's "about 2030" claim | **Disclosed; the central is unchanged.** Both dated claims (Michael Dell: tight through 2030, "Maybe it's a four-year problem"; Micron: demand above supply through calendar 2028, new fabs from 2028) sit inside the house memory path (shortage closes end-2030, normal by FY31-32), so the centre stays at 0 years. The claims are management claims; the CNBC quote is from a Stocktwits write-up, not the transcript. The late bound of the odds driver (+2.5 years) is three years past the latest dated claim, the early bound (-1.5) a bust from 2028 | 0.5 year later +$8.98, one year later +$18.47 (the auditor's +$17: close), 2 years later +$73.75, one year earlier -$10.45 |
| 2b | H2 FY27 memory index 0.71x Q2 against the house memory path's 5.28x by end-2026: guide low or the 20% share high | **Disclosed.** The 0.71x is independent of the memory share (0.71x at 12%, 20% and 30%), so the share is not the cause; the guide, or H2 volumes, is. Dell beat both measured guides (Q2 FY27: $47.0B and $7.04 against $44.5B and $4.80; Q1 FY27: $43.8B and $4.86 against $35.2B and $2.90, the guide from a secondary source), so the guide is probably low, but it is the only dated forward figure and the central keeps it | If H2 held Q2's excess and premium: FY27 EPS $27.86, fair value $303.17 (**+$11.54**); not adopted |
| 3 | Our report check's odds flags 16 and 18: path fixed, AI margin capped at 8%, memory delay at +2.5 years | **Fixed in `drivers.json`.** The accelerator path (sector note bear / proposed / bull) is a driver, `accel_path`, triangular -1 / 0 / +1, with the bull side farther in dollars. AI margin: 2.6% / 5.0% / 10.7%, from the server builders' annual margin history FY2017-26 (Supermicro 2.6-10.7%, HPE 4.8-8.3%), management's mid-single digits as the mode; the 8% cap is gone. AI share: 12% / 17% / 24%, the high end from the history of Dell's share (rising every year; 23% needs two points a year). Memory delay: normal, sd 0.8 years, bounds -1.5 and +2.5 with a dated rationale; the price needs 2.9, which sets nothing | P under 1% → **8.7%**; simulated median $282 → $320; 10th to 90th $204-$360 → $208-$453 (the price, $461.30, now just above the 90th) |
| 4 | Headline: peak or new base; the observable | **Fixed.** The decision block states it: whether FY30-level earnings are the cycle peak or the new base. The Street's FY30 EPS implies segment income of $37.6B (12.6% of its revenue); held as a base from FY30 it is worth $581 (at $36B, $557), as a fading peak $292. Observable: the AI server margin each quarter (10-Q "margin rate pressure"; TD Cowen 2026-09-29, as reported). The flip is about 10.2% against our 5.0% (Hold line $392), not 11%: the brief's 11% was computed on the blind value. The auditor's "$230-310" for the base case was not reproduced; `flip.py` values the held earnings by the same method | Flip 10.21% reaches Hold, 13.84% the price (`flip-output.txt`); AI margin 11% gives $407.23 (+$115.60) |
| 5 | AI share held at 17% against Dell's rising history; the note says the path and falling memory must give | **Disclosed, not changed.** 17.0% is the dated FY27 level ($74B guide over the note's $435B 2026 spend); history rose every year (3.1%, 7.1%, 11.4%, 18.1% of NVIDIA's revenue in FY2024-27); we hold it because a rising share needs a dated cause we do not have, while the note's own path already rises. Sensitivity rows are in the forecast section | 15%: -$12.96; 13%: -$26.00; path flat at $749B: -$15.02; the note's bear path: -$32.12; its bull path: +$71.83 |
| 6a | Leased equipment $3.182B added at book while its lease income stays in NOPAT | **Fixed.** Removed from the bridge; the income stays in NOPAT | -$2.95 |
| 6b | Long-term investments $2.68B missing | **Fixed.** Added at book, held flat ($2.7B at 2026-07-31, 10-Q) | +$2.49 |
| 6c | Operating lease liabilities $805M missing | **Fixed.** Deducted, held flat (10-Q lease note) | -$0.75 (the auditor's -$0.91 was not reconciled) |
| 6d | Original issue discount subtracted twice | **Rejected.** The 424B2 as-adjusted cash already nets the $14M discount; the model deducted only the $22M underwriting discount and $13.6M of expenses (total $35.6M), not the discount again | $0.00 |
| 6e | Cash earns 3.5% against 5% debt and 5.29% risk-free | **Disclosed (justified).** Dell's interest income over average year-end cash was 4.1%, 2.9% and 3.7% in FY2024-26 (market data; 3.6% pooled), below the 5% on debt because much of the cash probably sits offshore in short deposits (our inference); 3.5% is a haircut below the pooled history, not a price-driven choice | At 5.0%: +$1.31 |
| 7 | The fade starts at 158.6%; source, window, asset life | **Disclosed** in the forecast section: the figure is `valuation.md`'s prescribed volume-only start; window FY27-32 volume against FY28-32 reinvestment; reinvestment is our 6% of the rise in normal-price revenue plus $0.1B a year with no asset life; its check (capex below D&A in FY2022-26) and its limit (H1 FY27 capex pace is faster) are named | Start at the mature return: -$5.65; $1.2B a year more reinvestment (start 52%): -$9.67 |
| 8 | The bear cuts growth-after and the mature return on top of margin and revenue cuts, against the claim it does not | **Fixed.** Bear and bull now use the central's growth after FY32 (4%) and mature return (midpoint); the cases differ only in their operating worlds | Bear $154.12 → $160.17 (+$6.05, the auditor's figure); bull $533.62 → $507.50 (-$26.12) |
| 9 | Disclosure: header price levels; the old report's "about 2030" and "not on rates"; share count 635.8M against 652M | **Fixed / disclosed.** The header names the buyback level (about $400), insider sales ($530-560) and MU at $931.40 with its Street lines. The "about 2030" passage was not re-found in the blind step; the CNBC interview of 2026-10-13 is it (2a). "Not on rates" is cut: higher rates would pull multiples down, and what explains the multiples was not tested. Shares: 652M is the Q2 FY27 weighted-average diluted count (10-Q: 645M basic plus 7M dilutive), the quarter the model is calibrated on; the 10-Q cover shows 635.8M basic at a later date, after further repurchases, and the model pays buybacks out as cash at fair value rather than retiring shares | 642.8M: +$2.37; 635.8M: +$4.22 with every per-share input rescaled (the auditor's +$1.1 not reproduced; not adopted) |
| 10 | Outside-view peers: HPE trailing EBIT depressed, SMCI's latest quarter high; multiples at cycle highs | **Disclosed** in the forecast section with the run-rate multiples: HPE 15.6x not 25.2x, SMCI 11.1x not 7.2x. The two errors offset | Run-rate multiples: -$1.69; ten-year medians: -$11.67 |

Flags raised by our automated report check (`research-check.py`) after the
round, answered by number:

| Flag | Answer |
| --- | --- |
| 1 (medium): net move +$12.61 from the blind value; name the largest change | The largest change is step R5, the payouts (+$13.82), which rests on Dell's stated policy dated 2025-10-07 and on H1 FY27 actuals, both before and independent of the price; it is a management policy, and the actual H1 payout was 56%, not 80% (-$4.77 at 56%). Weigh it against the capped-buyback case (-$6.37) |
| 16 (high): P 8.7%, under 10% | Not cleared: P rose from under 1% to 8.7% by widening the ranges from history and adding the accelerator path as a driver. The coherent world in which the price is right is the case set already in the model: the sector note's bull accelerator path, AI margin 8% and the shortage to about 2030 gives the bull, $507.50, 10% above the price (the bull-leaning world's P is 29%). The advocate is asked to price a world from dated sources of its own. Separately, held at the Street's FY30 earnings level from FY30 the model gives $557-581 (`flip-output.txt`) |
| 17 (high): fair value is 0.52x the Street average target | Not cleared; our revenue is within 5% of the Street's through FY2029 and 19% under it in FY2030, and our EPS is 6%, 16% and 51% under the Street's in FY2028-30. The gap is the FY30 level: the Street's FY30 EPS implies segment income of $37.6B against our $19.9B; held as a base it is worth $581 (our revenue, FY2030 income only); the full Street path in our model is $585.19. This is a forecast disagreement over whether FY2030 is a peak or a base, not a model error: the scripts reproduce the Street path from its own figures. The driver-by-driver run of the Street's path is the table below the answers (`street_path.py`). The Street target is a 12-month figure, ours a fair value today |
| 18 (medium): the price is outside our 10th to 90th ($208-$453) | Disclosed. The width, $245, is 55% of the Street's target spread ($289-$735, $446), so the first clause (under half) does not fire; the price is $8 above our 90th. The ranges come from history (server builders' margin FY2017-26, Dell's share history, memory upcycle length) and were not set toward the price; they were not widened further |
| 14, 15 | Not raised: the simulated median is 9.6% from the central; the accelerator path is the note's |

The Street's path through our model one driver at a time, which answers
flag 17 (`street_path.py`, `street_path.txt`). Street FY2028-30 revenue $235.2B,
$263.6B and $298.4B (ours $247.5B, $256.7B, $243.1B); segment income from its
EPS as `flip.py` reads it, $25.8B, $29.3B and $37.6B (ours $25.2B, $25.7B,
$19.9B). FY2027 is the company guide in our model and stays; FY2031-32 carry
the FY2030 replacement.

| Step | Fair value | Change |
| --- | ---: | ---: |
| Our central | $291.63 | |
| Street revenue FY2028-30, every line scaled together at our margin | $353.87 | +$62.24 |
| + Street segment income FY2028-29 | $358.86 | +$4.99 |
| + Street segment income FY2030, held in FY2031-32 | $585.19 | +$226.33 |

The FY2030 gap in segment income, $17.7B ($19.9B to $37.6B), is $4.5B revenue
at our 8.2% margin and $13.1B margin (12.6% against 8.2%). The Street's path is
worth $585 in our model, $20.54 above the Street's average target; the target
discounted one year at our 9.96% cost of equity is $513.50. With our revenue and
only the FY2030 segment income held, the value is $581.10 (`flip-output.txt`).

Old against new, per diluted share:

| | Blind commit (market pass) | Reviewed |
| --- | ---: | ---: |
| Central fair value | $279.02 | $291.63 |
| Bear / bull | $145.66 / $515.83 | $160.17 / $507.50 |
| Case-weighted (25/50/25) | $305 | $313 |
| Exit price | $438 | $346 |
| Purchase price (12% hurdle) | $253.85 | $269.15 |
| Five-year return from $461.295 | -0.2% | -0.7% |
| View | Avoid | Avoid |
| P(price justified) | under 1% | 8.7% |

Authorization limit. Nothing here changes a live model or portfolio.