---
ticker: "INTC"
company: "Intel Corporation"
date: "2026-10-15"
view: "Avoid"
fair_value: 31.23
price: 95.37
price_source: "market data, official 2026-10-15 close (SIP); last regular trade $95.30 at 19:59:59 UTC"
price_time: "2026-10-15T20:00:00Z"
price_vs_fair_value: "+205.4%"
bear: 1.91
bull: 65.03
exit_price: 53.8
p_price_justified: "under 20%"
restated: "2026-10-26"
method: "case-spread-2026-10-24"
original_fair_value: 26
original_view: "Avoid"
page: "https://alpha.dming.si/research/2026-10-15-intc"
---

# INTC · Intel Corporation · Avoid

Research report as of 2026-10-15, partly unblinded. Page: https://alpha.dming.si/research/2026-10-15-intc

We value INTC at $31.23 a share. The report's $95.37 price (Oct 15, 2026) was 205% above fair value: Avoid.

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

| Lens | Range | Marks |
| --- | --- | --- |
| Our report | $1.91 to $65.03 | fair $31.23; original $26 |
| Our simulation | $9.39 to $81.68 | median $31.33 |
| Street targets (20) | $60 to $200 | average $112, median $110 |
| Price | | $95.37 |

Bands run from our bear to bull (the hollow mark is the original $26), 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 | $31.23 |
| Price | $95.37 (market data, Oct 15, 1:00 PM PT) |
| Five-year return | -10.5% (a year, at the report's price) |
| Bear / bull | $1.91 / $65.03 |
| Exit price | $53.80 (on Dec 27, 2031) |
| Confidence | 30% |
| 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:

- Server CPU unit growth, 2027-2031: needs +31.0% a year, we have +9.0% a year
- Server CPU price growth, 2027-2031: needs +17.8% a year, we have +3.4% a year
- External foundry volume in 2031, $B: needs 149, we have 10.5

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

**Original call, Oct 15, 2026 (before the Oct 26, 2026 restatement)**

```
INTC — Intel Corporation — research view: Avoid (new money at $95.37)
As of 2026-10-15 · partly unblinded: we committed the model (498979b, blind value
$29.31) before looking at any INTC price, but the masked Q2 10-Q still showed a
liability that implied a June price far above it (section 7). After the commit the
August offering and two rounds of review moved the value; section 7 prices each step.

Fair value today      $26 per diluted share (cost of equity 10.84%, the yearly return
                      investors require); blind $29
Price                 $95.37, official close 2026-10-15 (SIP, the consolidated
                      trade feed; last regular trade $95.30 at 19:59:59 UTC). Fair
                      value is 73% below the price.
Five-year exit price  $45 on 2031-12-27, end of fiscal 2031: the 15-year long fade (10
                      years $43, 20 years $47). Checks, not averaged in: outside view
                      $73 (peers' 23.2x price-to-earnings), value-driver formula $41
Five-year return      -13.5% a year from today's price on our central case (no payouts)
Purchase price        $25 to earn the 12% hurdle (the yearly return we require)
Bear / bull today     $2 / $49, each a coherent operating case (weighted 25/50/25: $26);
                      success world $65, range $47-$78 (section 3)
Confidence            30% that the fiscal 2031 exit value lands within ±25% of $45
                      (subjective, not calibrated); higher that it lands below the price
Moat                  The x86 server and PC incumbent with US leading-edge fabs and
                      packaging the US government has bought into; no moat yet in
                      foundry (making chips for other companies)
Market expects        The success world, nearly for certain, with excess returns that
                      last far longer than our fade allows: BofA's (Bank of America
                      analysts') volumes (server CPUs about $45B and foundry about $40B
                      by 2030) and bull operations reach $95 only at an 8.89% cost of
                      equity
Price needs           One of: server CPU units +34.2% a year for five years (central +9.0%);
                      server CPU prices +20.0% a year (central +3.4%); or $198B of outside
                      foundry volume in 2031 at 2026 prices (central $10.5B). Jointly: the
                      success world plus $94B of foundry volume, or the success world at
                      an 8.89% cost of equity
Odds in our scenarios Under 1% (0.17%) that fair value reaches the price
                      (P(price justified); subjective, not calibrated): three simulated
                      worlds, 75% base and 20% bull-leaning (each under 1%) and 5% success
                      (2.6%), weights our judgment from fetched sources (section 2);
                      1.9% on the Street's volumes. A diagnostic of our scenario
                      spread, not the odds of a 12% return
Key driver            Whether Intel becomes a scaled outside foundry. Inside the base
                      world: server CPU units 4% a year gives $17, 12% a year $32;
                      outside foundry $2.2B in 2031 gives $18, $22B gives $30
Street                Mean target $120.28 (market data, 52 analysts: 19 buy or
                      overweight, 31 hold, 2 sell; range $80-$200; read 2026-10-16).
                      Targets forecast a future price, so they show expectations, not
                      validation or value today
```

**We rate Intel Avoid for new money at $95.37: our central fair value is $26 a
share, 73% below the price.** Intel is in a real turnaround and the central case
credits all of it. The price needs more: a near-certain, decades-long success
as an outside chip manufacturer, which we give a 5% weight.

**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. The success world is a fourth case: the most coherent world in which
today's price is right (section 3). The fade is how we value Intel after 2031:
growth and the return on each new dollar of investment drift over 15 years
toward long-run levels. The outside view applies the price-to-earnings (P/E)
multiple that peers have traded at. The value-driver formula is a third
estimate of the exit value from growth, return on capital and the cost of
capital. A fab is a chip factory. A foundry makes chips for other companies; 18A and
14A are Intel's current and next manufacturing processes. Capex is capital spending, ASP is
average selling price, and WACC is the weighted average cost of capital, the
rate at which the whole business's cash flows are discounted. The Street means
the analysts who follow the stock. Two reviewers checked this report: an
auditor, who looks for errors, and an advocate, who argues for a higher value
(section 8).

## 1. Why $26: a real turnaround, credited in full

Intel is in a real turnaround. Revenue is growing again (Q2 2026 +25% year on
year, server CPU prices +48%), the 18A node is in volume, Intel has committed
to finish 14A (10-Q, Q2 2026), and the US government and the August equity
buyers have funded the fabs. Our central case gives Intel all of that:
revenue from $63.6B in 2026 to $102.8B in 2031, GAAP gross margin back to 43%
and operating margin to 20%.

That is worth about $26 a share. Intel spends $137B of gross capex in
2027-2031 in the central case. In the long fade after 2031, the return on each
new dollar falls from what the 2027-2031 build earns toward Intel's own long
record: 10.8% after tax on invested capital from 2014 to 2025 (market data
statements), against a 9.9% WACC. And after 2031 Intel pays full price for
its fabs: the 48D credit, a US tax credit for chip-factory investment, that now
pays 35% of qualifying US investment ends for construction begun after 2026 (26 USC 48D(e)).

Dong, who owns the research program, pointed out that a 0% chance on the
odds makes it obvious something is wrong. He is right: a 0% chance that the
price is right meant our ranges were one hump, not that the market was wrong.
So section 3
builds the most coherent world in which $95.37 is right and prices it: Intel
becomes a scaled outside foundry and rides the CPU wave, with BofA's figures
of $45B of server CPU revenue and $38B of foundry revenue in 2030, bull
operations and a 20% return on capital. Our model values that world at $65,
within a range of $47 to $78 depending on its capex and depreciation. Plainly:
the rest of the gap to $95 is mostly about how long excess returns last and
how the long run is discounted, not only about volumes. A 25-year fade with
4% terminal growth gives $78, and an 8.89% cost of equity gives the price. We
think that is unlikely: no foundry has grown 30-fold in four years, and only
the leader, TSMC, has kept excess returns that long. We weight the success
world at 5%. The price is betting on near-certain success held for decades;
we are not.

## 2. Working backwards from $95.37: what the price needs

This works backwards from $95.37 on the central model, a reverse discounted
cash flow (`uv run alpha research odds 2026-10-15-intc`, `odds.json`).
Everything not named stays central. Gross capex rises $1.5 for each extra $1
of yearly foundry volume (the advocate's tie, section 8), and year-five
depreciation is normalised on gross capex (the auditor's rerun, section 8).

| Driver | Central | Needed alone for $95.37 | Against the record |
| --- | --- | --- | --- |
| Server CPU units, a year, 2027-2031 | +9.0% (+21% in 2027 fading to +3%) | **+34.2%** | Units grew 9% year on year in Q2 2026 and 2% in the first half (10-Q). Needed: more than four times today's units by 2031, against $26.1B of server revenue at the 2020 peak |
| Server CPU price, a year, 2027-2031 | +3.4% | **+20.0%** | Q2 2026's +48% came mostly from a mix shift to high-core-count parts (market data note, 2026-08-12); five more years at +20% is 2.5 times today's ASP |
| Outside foundry and packaging volume in 2031, $B at 2026 prices | 10.5 | **198** | $0.293B in Q2 2026. BofA's model, the most bullish we found, has "nearly $40 billion" of foundry revenue in 2030 (Benzinga, 2026-08-12) |
| Product unit cost | central | **not reached**: 30 points a year lower gives $63 | |
| Capex | central | **not reached**: 20% less gives $38 (below that the method cannot value it) | |
| Mature return on new capital | 10.8% | **not reached**: 100% gives $32 | Intel 2014-2021 21.7% |
| Growth after 2031 | 6% | **not reached** below the WACC: 9.5% gives $29 | |

**Combinations that reach the price:**

- **A.** Our bull's server path (units +12.7%, price +4.8% a year), a 16%
  mature return and 9% growth after 2031 need **$122B** of outside foundry
  volume in 2031.
- **B.** Our bull's $22B of foundry volume with the same 16% return and 9%
  growth needs server units **+28.8% a year**.
- **C.** The Street's server path (units +15%, price +6% a year) and 9%
  growth need **$109B** of foundry volume, at our central costs.
- **D.** The whole success world (bull operations, the bull's server path, a
  20% mature return, 9% growth) needs **$94B** of foundry volume in 2031,
  against BofA's $40B in 2030.

**The odds, as three worlds.** A turnaround either works, half works or does
not, so `drivers.json` draws from three worlds and P, the chance that fair
value reaches the price, is the mixture. The
weights are our judgment, set from fetched, dated sources
(`data/foundry-base-rate-and-48d-sources-2026-10-16.txt`), not from memory:

- **Base, 75%:** Intel stays a CPU company with a sub-scale foundry. Our
  central paths with ranges from Intel's record (server units around +9.0%
  with a 4-point deviation, set from data-center revenue growth of about 10%
  a year in 2014-2020; foundry 2031 volume $2.2B to $30B; mature return 8% to
  18%). P under 1%; median $23, 90th percentile $43.
- **Leaning to the bull, 20%:** one or two outside 14A customers commit and
  foundry grows toward the bull's $22B, without the success world's scale
  (foundry mode $18B, server units around +11%, mature mode 14%). Intel
  expects 14A design commitments "beginning in the second half of 2026 and
  expanding into the first half of 2027" (Q1 call, 2026-04-23). Eight
  companies were reported evaluating 14A with none committed (Piper Sandler,
  as reported by TechSpot, about September 2025; secondary). TrendForce
  (2026-01-13) notes that "with 18A, Intel has yet to land a major outside
  client with meaningful volume demand". P under 1% (0.01%); median $30.
- **Success, 5%:** the bull's operations, server units around +12.8% and
  price +4.8% a year, foundry 2031 volume $20B to $60B (mode $42B), mature
  return 12% to 25%. P 2.6%; median $53, 90th percentile $80.

The 5% comes from a speed test, the auditor's suggestion. The success world
needs outside foundry revenue to grow about 30 times in four years ($1.27B
in 2026 to $38.2B in 2030). From 2Q21 to 2Q26, quarterly revenue at TSMC grew
3.0 times ($13.3B to $40.2B) and at SMIC 2.2 times ($1.34B to about $3.0B).
Samsung shrank from $4.33B (17.7% of the market, our arithmetic) to $3.26B
(5.9%) (TrendForce, 2021-08-31 and 2026-09-09). GlobalFoundries stopped 7nm
on 2018-08-27 (eeNews Europe; IEEE Spectrum). No foundry in that record came
close to 30 times in four years. The auditor would defend 3-8%; we take 5%.
SMIC (5.4% share) belongs in the class but grew with state support and
sanctioned demand, which Intel does not have in the same form.

The mixture gives **P = 0.17%** (shown as under 1%), mean $27, median $25,
90th percentile $49. On the Street's volumes alone P is 1.9% (median $49). P
is subjective, not calibrated, and is not the probability of a 12% return.
The simulated median ($25) is 3.4% from the central value; no gap to explain.

## 3. The price is betting on a scaled foundry and decades of excess returns

**The most coherent world in which $95.37 is right**, built only from dated
sources that do not depend on the price:

- **Servers.** Intel takes the CPU wave: BofA has Intel server CPU sales at
  "about $45 billion by 2030" (Benzinga, 2026-08-12), about our bull's path
  (units +12.8% and price +4.8% a year; $44.8B in 2030). Support: record
  server growth and CPU-to-GPU ratios moving toward parity (Q2 call,
  2026-07-23); 3-5 year volume and price agreements, including Google (Q1
  call, 2026-04-23); $1.7B of customer deposits under long-term agreements
  (10-Q, Q2 2026); the CEO meeting about half of CPU demand (Globe and Mail,
  2026-09-27, secondary).
- **Foundry.** Intel becomes a scaled outside foundry: $38B of outside
  revenue in 2030 and $51B in 2031 (BofA: "nearly $40 billion" in 2030).
  Support: the commitment to finish 14A (10-Q, Q2 2026); the CFO's
  "conviction around customers on 14A externally" (Deutsche Bank conference,
  August 2026); Apple on 18A-P from 2027 and a Tesla 14A project, both
  secondary reports. Capex rises $1.5 for each extra $1 of foundry volume, so
  2030 gross capex reaches $56B.
- **Operations and returns.** The bull's client, cost and opex paths; a 20%
  mature return, about Intel's 21.7% of 2014-2021 when it led on process;
  9% growth after 2031.

The result: revenue $160.7B in 2031, GAAP gross margin 48%, operating margin
31.5% (below TSMC's 35-51% of 2016-2025, market data ratios), free cash flow $33B in
2031. **Our model values it at $65.39, in a range of $46.56 to $77.90.**

**The success world's range** (each row changes one thing from $65.39):

| Change | Value | Effect |
| --- | --- | --- |
| Year-five depreciation on 2031 net capex, before the auditor's rerun fix R1 (section 8) | $77.82 | +$12.43 |
| 2031 gross capex held at $50B instead of the tie's $44.5B | $56.26 | -$9.13 |
| 2031 gross capex held at $56B (the 2030 level) | $46.56 | -$18.83 |
| Depreciation normalised at 6% growth instead of 3% | $72.36 | +$6.97 |
| Depreciation normalised at 9% growth (the world's own growth after 2031) | $77.90 | +$12.51 |
| Capex tie $1.0 or $2.0 per $1 of foundry volume instead of $1.5 | $69.21 / $61.63 | +$3.82 / -$3.76 |
| Mature return 15% or 25% instead of 20% | $62.90 / $67.03 | -$2.49 / +$1.64 |

The point value depends on the 2031 capex dip: the tie adds capex for each
year's added volume, and the success path adds $18B of foundry volume in
2030 but $9B in 2031, so 2031 gross capex falls from $56B to $44.5B while
volume still rises. Holding capex at the 2030 level is the low end; reading
the normalisation at the world's own 9% growth is the high end.

**What in our structure stands between that world and $95:**

| Structure | Success world | Change from $65.39 |
| --- | --- | --- |
| Terminal growth 4% instead of 3% | $70.27 | +$4.88 |
| A 25-year fade instead of 15 | $73.39 | +$8.00 |
| Both together | $78.37 | +$12.98 |
| Both together, with a beta of 1.2 instead of 1.32 | $87.30 | +$21.91 |
| Exit at the peers' 23.2x ten-year median P/E instead of the fade | $85.37 | +$19.98 |
| Cost of equity that makes it worth $95.37 | 8.89% (beta 0.86 against our 1.32) | +$29.98 |

Plainly: **the gap between the success world and the price is mostly how
long excess returns last and how the long run is discounted, not only
volumes.** Before the auditor's rerun fix R1 (section 8), the auditor showed that a
25-year fade with 4% terminal growth reached $94.10 and, with a beta of 1.2,
$97.16; that is, the structure alone nearly closed the gap. Once the 48D
credit is no longer counted in the steady state, the same changes give
$78.37 and $87.30. Even then, structure is worth more than any volume
assumption we could defend: about $22 of the remaining $30.

- **The fade does not start from a depressed return.** It starts at the
  larger of the 2027-2031 return on new capital and the mature return
  (central 29.3%), so today's losses do not hold the value down, and the
  mature return moves the success world by less than $3.
- **Is a 25-year fade at 4% terminal growth unlikely for Intel? Yes, in our
  judgment.** Long excess returns in leading-edge manufacturing have
  belonged to the leader. TSMC kept a 35-51% operating margin through
  2016-2025 (market data ratios) and holds 72.5% of foundry revenue (TrendForce,
  2026-09-09). The others lost theirs. Intel itself earned 21.7% on capital
  in 2014-2021 and lost money in 2022-2025 (market data statements). Samsung's
  foundry share fell from 17.7% to 5.9% in five years (TrendForce).
  GlobalFoundries left the leading edge in 2018. In the success world Intel
  would be a distant second to TSMC, not its replacement.
- **The cost of equity.** 8.89% would need a beta of 0.86. The market does
  not see Intel as safer than its peers: Intel's own weekly beta is 1.58 raw,
  1.39 Blume (2021-10 to 2026-10), above the peer median of 1.32.
- **Strategic value the cash-flow model cannot hold.** Intel is the only
  company doing both leading-edge logic R&D and high-volume manufacturing in
  the US (10-K 2025), and the government holds shares and warrants (10-Q, Q2
  2026). That is a floor, not upside: it protects against the bear (a
  rescue, a forced partner), not the central or the price. At most it lifts
  the bear to the central, which adds 25% x ($26.01 - $1.82) = $6.05 to the
  case-weighted value.

Valuing the central case on the same 23.2x exit gives $39.92. Mixing it with
the success world at that multiple cannot reach $95.37 at any weight: the
success world itself is $85.37 there.

**Which part of that world we think is unlikely, and why:**

1. **Foundry at $40B by 2030 (most unlikely).** Outside foundry revenue was
   $0.293B in Q2 2026 (10-Q) and $1.27B is our 2026 estimate; $38B in 2030
   is about 30 times that in four years, against 3.0 times in five years for
   TSMC and 2.2 times for SMIC (TrendForce). No filing names a significant
   14A customer; the 10-K says 14A may be paused without one. BofA's own
   base case has "no specific customer deals" (Benzinga, 2026-08-12).
2. **Excess returns that last 25 years or more.** As above, only the leader
   has done it.
3. **Server CPUs compounding at the bull's pace for five years.** Intel's
   best stretch was about 10% a year of data-center revenue (2014-2020, market data
   segments). The bull needs 12.8% units plus 4.8% price a year, and Q2
   2026 units grew 9% year on year and 2% in the first half (10-Q).

**Does our structure miss something the market prices?** In part, yes: the
market prices returns that persist longer than our 15-year fade. We show
what that is worth ($13 to $22 in the success world) but keep the method,
which we apply to every company, and say that this is where we and the
market most differ.

## 4. Our out-years are well below the Street's

The Street's Street consensus (fetched 2026-10-15; adjusted EBIT, which excludes stock pay
and restructuring) against our central on the same basis:

| Year | Street revenue (range, analysts) | Our central | Our bull | Street EBIT margin | Our central | Our bull |
| --- | --- | --- | --- | --- | --- | --- |
| 2026 | $63.3B (60-64, 34) | $63.6B | $63.6B | 17.3% | 15.8% | 15.8% |
| 2027 | $73.4B (66-80, 37) | $71.4B | $75.1B | 20.6% | 18.1% | 20.1% |
| 2028 | $84.1B (73-100, 35) | $80.5B | $90.1B | 25.8% | 18.7% | 22.6% |
| 2029 | $98.4B (89-112, 17) | $88.4B | $105.5B | 29.8% | 21.1% | 26.6% |
| 2030 | $125.3B (113-143, 17) | $95.7B | $120.9B | 33.2% | 22.6% | 29.4% |

The Street's 2030 revenue is our bull's; its margin is above our bull's. Market data
gives no segment split, so the volume, price and mix behind it come from the
broker figures we could find (one Bigdata search, 2026-10-16,
`data/bigdata-broker-assumptions-2026-10-16.txt`):

- **Server CPUs.** BofA: Intel server CPU sales "about $45 billion by 2030",
  roughly 20% of a $220B market, with double-digit unit and ASP growth
  through 2030 (Benzinga, 2026-08-12). Our central 2030 server revenue is
  $36.4B; our bull's $44.8B. Raymond James puts the server CPU market at
  $201B in 2030, 44% a year (Globe and Mail, 2026-09-02). AMD's $220B is a
  rival's claim, quoted by an analyst on Intel's Q2 call.
- **Foundry.** BofA: Intel Foundry revenue from about $1.1B in 2026 to nearly
  $40B in 2030, 8-10% of a $380B foundry market, with no specific customer
  deals in the base case (Benzinga, 2026-08-12). Our central has $9.0B in
  2030 and our bull $18.5B.
- **Margin.** Management: gross margin now "comfortably in the 40s", aiming
  for the mid- and high-40s and ultimately above 50%, with foundry,
  packaging and custom chips closer to 40% (Deutsche Bank conference, August
  2026, as reported by Nasdaq on 2026-08-27 and in our transcript).

**Why we differ.** On servers, a $220B market in 2030 would be six times the
$30-35B of 2025 (market data note, 2026-08-12, on company claims). We take
management's own words, double-digit industry units through 2028, and fade
after: no CPU cycle in Intel's record compounded for five years at 20% or
more. On foundry, see section 3. On margin, the Street's 33% adjusted margin
is at the top of Intel's 22-33% GAAP range of 2014-2021, when it led on
process and had no foundry losses to absorb.

## 5. Margins against history, same reported (GAAP) basis

| | GAAP gross margin | GAAP operating margin |
| --- | --- | --- |
| Intel 2014-2021 (market data) | 55-64% | 22-33% |
| Intel 2022-2025 (market data) | 33-43% | -22% to 4% |
| 2026E (our base) | 40.6% | 11.4% (before acquisition amortization) |
| 2031 bear | 29.9% | -1.1% |
| 2031 central | 43.4% | 20.2% |
| 2031 bull | 48.0% | 28.9% |
| 2031 success world | 48.0% | 31.5% |

Each margin is a result: revenue is units times price per line, and cost of
sales is units times unit cost plus depreciation from the capital schedule.
The only calibration is one 2026 cost ratio, set so that 2026 GAAP gross
profit matches H1 actual plus the Q3 guide (assumptions.md, sections 3-4).
Central depreciation rises from 18.9% of revenue in 2026 to 21.3% in 2031,
above Intel's 2014-2025 average D&A of 16.2%: that is the price of the capex.

## 6. From enterprise value to a share price: the August offering, claims and payouts

**The offering, verified from the filings.** On 2026-08-10 Intel agreed to
sell 210,526,315 shares at $95.00 to underwriters led by J.P. Morgan, Goldman
Sachs, Morgan Stanley and Citigroup, who pay $93.4325 a share (8-K filed
2026-08-12, Item 8.01; Underwriting Agreement, Schedule I). The launch release
proposed $15B; the pricing release says the offering "was upsized to $20
billion from the previously announced offering size of $15 billion" (both
2026-08-10, exhibits 99.1 and 99.2). The underwriters exercised their
31,578,947-share option in full on 2026-08-11 (8-K). Closing was set for
2026-08-12. In total: 242,105,262 shares, $23.0B at the offering price, $22.62B
to Intel before offering expenses. Use: general corporate purposes, including
capital expenditures and working capital (exhibit 99.2); the CFO tied it to
supplier commitments for 2027 capex and 14A (Deutsche Bank conference).
Files: `data/8-K-2026-08-12-offering-filing.txt`.

**"Heavily oversubscribed": not confirmed by a primary source.** Neither the
8-K, Intel's two releases nor the CFO's conference remarks say the book was
oversubscribed. The claim comes from Bloomberg, citing people familiar with
the matter, as relayed by MT Newswires on 2026-08-11 ("multiple-times
oversubscribed"; about a third of ordering investors got no shares) and by the
Charlotte Observer on 2026-08-13 (more than $100B of orders)
(`data/bigdata-offering-press-2026-10-16.txt`). The filings do show demand
consistent with it: a one-third upsize on the day and the full option
exercised the next day. Selling stock at $95 is accretive at our value: each
share brought in more than three times what we think it is worth.

**Debt and other claims today** (2026-06-27, 10-Q, plus the offering): cash and short-term
investments $29.7B less $5.0B operating cash; AMIC (advanced manufacturing investment credit) refundable tax credits
receivable $8.8B; equity
investments $8.4B; Altera receivable $0.5B; Secure Enclave cash for 143M
escrowed shares $2.9B; the offering $22.6B; less debt $50.5B, customer
deposits for future supply $1.7B, partners' interests at book $15.6B (Arizona
fab $13.4B, Mobileye $2.1B), accruals for the VLSI patent case and the EC (European Commission) fine
$1.3B and the payable to SLP for the Altera separation
$0.3B. Net -$1.5B. The offering counts the $22.62B Intel received at the underwriters' price; other offering expenses are not deducted (the 8-K gives no figure; the pricing release's net for the firm shares is within rounding of the underwriters' price). Shares: 5,043M outstanding, 143M escrowed government
shares, 117.5M RSUs and PSUs at target (10-K 2025 table,
`data/10-K-FY2025-rsu-psu-table-filing-excerpt.txt`) and the offering's
242.1M: 5.5456B.

**Capital-return convention.** No dividend since 2024 and no buyback
authorized (10-K). The base pays nothing out: free cash flow from 2026 H2 to
2031 is retained at a 4.0% pre-tax yield (our assumption), deficits are
borrowed at 5.45%, and the share count stays at 5.5456B (stock pay is
expensed in NOPAT, net operating profit after tax). Paying out every surplus at each fiscal year-end instead
gives $26.36 central, $2.05 bear and $49.11 bull: the convention moves cents,
because central free cash flow only turns strongly positive from 2029 ($10.6B,
$13.0B, $14.8B in 2029-2031).

## 7. How blind the work was, what moved the value, and the report check's flags

Blind means we valued the company before reading any price, Street estimate
or target.

- **Blind commit 498979b**: central $29.31, bear -$0.36, bull $64.11. No INTC
  price, multiple, target, consensus or news recap was opened before it.
- **Partly unblinded.** The masked Q2 10-Q kept the $15.6B fair value of the
  forward liability on 143M escrowed government shares and a $12.5B quarterly
  loss "driven by an increase in our stock price", which implies a June price
  of roughly $100. The masked copy of the offering 8-K, fetched before the
  commit and not opened, also kept "$95.00 per share". Both gaps in our masking tool, which hides market information from
  filings, are in tools.md. No input was moved toward either.

**Every step from the blind value, in order** (`output.txt`; each is a filing
or a reviewer's finding, none the price):

| Step | Bear | Central | Bull |
| --- | --- | --- | --- |
| Blind (498979b) | -$0.36 | $29.31 | $64.11 |
| August offering, 242.1M shares, $22.62B (8-K 2026-08-12) | +$4.09 | +$2.80 | +$1.28 |
| Customer deposits $1.7B as debt (10-Q; auditor) | -$0.31 | -$0.31 | -$0.31 |
| 2027 server units +18% to +21% (Q2 call; advocate) | | +$1.11 | |
| Client unit cost lower (Panther Lake, Q2 call; advocate) | | +$0.93 | +$1.42 |
| Year-five depreciation at 3% long-run growth (auditor) | +$0.50 | -$3.02 | -$9.50 |
| Year-five depreciation on gross capex: 48D ends for construction begun after 2026 (auditor rerun) | -$2.11 | -$4.80 | -$8.26 |
| **Now** | **$1.82** | **$26.01** | **$48.76** |

Intel's own beta (Blume 1.39) gives $24.95 and a $23.94 purchase price
against the peer median's $26.01 and $24.63; the method keeps the peer
median.

**Flags that apply from our automated report check (research-check)**, by
its flag numbers:

- **Flag 1, medium**: net move from blind -$3.30 (-11.3%), every step above from
  a filing, a statute or a reviewer's finding.
- **Flag 3, medium**: value 73% from the price.
- **Flag 4, low**: the outside view ($73 at the peers' 23.2x ten-year median P/E)
  is 62% above the fade ($45); section 3 prices what that gap means.
- **Flag 5, medium**: central 2030 revenue 24% below the Street (Street consensus, 17
  analysts); our dated sources are in section 4.
- **Flag 6, medium**: two lenient inputs. The 15% debt weight (8.7% at market)
  lowers WACC; at 8.7% the value is $24.84. The 7-year tool life is at the
  long end of the 10-K's 3-8 years.
- **Flag 8, high**: the 241M government warrants at $20 are not in the share count,
  because they become exercisable only if Intel's ownership of its foundry
  falls below 51% (10-Q, Q2 2026). If exercised: central $25.76, bull $47.56.
- **Flag 13**: answered. Customer deposits are now in the claims counted today (section 6; finding A3 in section 8).
- **Flag 14**: does not apply; the median is 3.4% from the central.
- **Flag 15**: does not apply. Our server path is our sector note's shared path;
  2027 moved from +18% to +21% for a dated reason (Q2 call, 2026-07-23), and
  the sector note's INTC rows are updated to match.
- **Flag 16, high**: P is 0.17% (0.0% before the review). Answered in section 3:
  the price needs the success world nearly for certain, with excess returns
  that outlast our fade; we give that world 5%.

## 8. What the two reviews found, and how we answered

The auditor and the advocate reviewed this report on 2026-10-15; the auditor
then reran it after our fixes. Finding IDs: A is the auditor's first review,
V the advocate's, R the auditor's rerun, and Owner a question from Dong, who
owns the research program. Dollar effects are on the central per share,
measured on the current model unless marked.

| # | Finding | Answer | Before | After | Effect |
| --- | --- | --- | --- | --- | --- |
| A1 | Stale wording: record says "about 30%", "$90B", "$45B"; report rounds free cash flow | **Fixed.** Record and report now quote odds.json as printed (now +34.2%, $198B, joint D $94B) and free cash flow to one decimal from output.txt | | | $0 |
| A2 | RSU count not saved; 8M PSUs left out | **Fixed** (source saved: `data/10-K-FY2025-rsu-psu-table-filing-excerpt.txt`). The 117.5M already includes PSUs at target. **Disclosed:** at the 200% maximum payout | 117.5M | 125.5M | -$0.04 |
| A3 | Customer deposits are debt-like | **Fixed** at $1.7B: the 10-Q says the $1.7B received under Q1 long-term agreements is booked in accrued and long-term liabilities, so the $1.2B long-term balance is most likely part of it, not added to it | $0 | -$1.7B | -$0.31 (at that step) |
| A3 | Operating leases | **Rejected:** rent is in our cash opex, so a lease liability on top would count it twice | | | |
| A3 | Arizona partner bought out above book (Apollo's Ireland stake: $0.14B carried, $14.2B paid, per auditor) | **Disclosed:** no buyout is announced; a 1.29x buyout of the $13.4B claim | | | -$0.70 |
| A4 | Ranges single-humped and capped; a coherent success case gives $77-89 | **Fixed:** worlds in drivers.json, now three after R4; success world priced in section 3. P | 0.0% | 0.17% | central unchanged |
| A5 | Normalisation assumes 6% capex growth while capex is flat | **Fixed:** year-five depreciation normalised at the 3% long-run growth the fade ends at, for every case | central $33.84, bull $66.51, bear $3.43 | $30.82, $57.02, $3.93 | -$3.02 central, -$9.50 bull |
| A6 | Mature return 10.8% does not reproduce (auditor 10.0%) | **Disclosed:** our invested capital is equity plus minorities plus debt, less cash, short-term and long-term investments, year-end, pooled 2014-2025 (market data). The auditor's 10.0% is under $0.5 | | | under $0.5 |
| A6 | A 21% tax from 2029 | **Disclosed:** our path rises from 12% in 2027 to 15% in 2030-2031 (assumptions.md); at 21% | | | -$2.09 |
| A6 | Unvested RSUs on top of stock-pay expense | **Disclosed:** the unvested grants' remaining expense sits in 2026-2028 stock pay, so part is counted twice; we keep the conservative count. At most | | | +$0.66 (auditor) |
| V1 | Server units and price too low; 2027 only 5% above the Q4 run rate | **Fixed in part:** 2027 units anchored to the Q4 2026 exit rate, +18% to +21%. **Rejected** for the central: the advocate's full path (+25% units, +8% price in 2027) is worth +$7.66, but Q2 units were +9% year on year and +2% in the first half (10-Q) and the +48% price is mostly mix (market data, 2026-08-12). Their path is near our bull, used in the success world | +18% | +21% | +$1.11 |
| V2 | Foundry, with capex tied at $1.5 per $1 | **Fixed (structure):** capex now follows foundry volume in `fair_value` and the success world; the foundry-capex correlation is removed. Central volume stays $10.5B (no named customer). At $16B and $22B the value rises $1.94 and $4.09. Foundry needed alone | $72B | $198B (after R1) | central $0 |
| V3 | Opex +3% a year instead of 4-5% | **Disclosed:** the $16.5B guide covers 2026 only; no 2027 source. At 3% | | | +$1.67 |
| V4 | Unit cost 1 point lower a year | **Fixed for client:** Panther Lake cost down "roughly 50%" so far with further reductions in 2027 (CFO, Q2 call): central client -5, -4, -3, -2, -1% to -7, -5, -4, -2, -1%; bull client also lower. Server cost has no such source; the advocate's 1 point on both lines would be +$2.10 | | | +$0.93 central, +$1.42 bull (at that step) |
| V5 | Incentive credit 30% in 2027-2029 | **Disclosed:** no source for the US share of capex that would earn 30%. At 30% | | | +$0.64 |
| V6 | Mature return 16% and 8% growth after 2031 | **Rejected** for the central (+$3.52; the return alone +$1.47): Intel earned 10.8% over 2014-2025. Used in the success world (20%, 9%) | | | |
| R1 | Perpetual 15% capex credit: year-five depreciation normalised on 2031 net capex carries the 48D credit for ever, but the credit ends for construction begun after 2026-12-31 (26 USC 48D(e)) | **Fixed:** normalised on 2031 gross capex for every case. Cash credits in 2027-2031 stay (25% falling to 15%), because 48D follows placement in service and Ohio and Arizona construction began before 2027. Zeroing the 2031 cash credit as well would be -$5.70 (auditor); no dated source for an extension was found, so no extension case | central $30.82, bear $3.93, bull $57.02, success $77.82 | $26.01, $1.82, $48.76, $65.39 | -$4.80 central, -$2.11 bear, -$8.26 bull, -$12.43 success |
| R2 | Success world shown as one point that rests on the 2031 capex dip and on 3% normalisation growth | **Fixed:** section 3 shows the range, $46.56 (capex held at $56B) to $77.90 (normalised at the world's 9% growth) | $77.82 | $65.39 ($46.56-$77.90) | |
| R3 | The structure can nearly reach the price; say so | **Fixed:** section 3 says it plainly. The auditor's $94.10 (25-year fade, 4% terminal) and $97.16 (with beta 1.2) were before R1; after R1 they are $78.37 and $87.30. We judge long-lasting excess returns unlikely for a second-place foundry, with evidence | | | |
| R4 | Base-rate paragraph recalled, partly wrong | **Fixed:** replaced with fetched, dated sources (TrendForce 2021-08-31 and 2026-09-09, eeNews and IEEE Spectrum, TechSpot, Q1 call), saved in data/. Weights moved from 83/17 to 75/20/5 (base, leaning to the bull, success). UMC and Intel Custom Foundry are dropped: no fetched source | P 1.1% | P 0.17% | |
| R5 | assumptions.md sections 5 and 11 still say capex is not tied to volume; sector note line references | **Fixed:** both sections updated; the sector note points to the right lines | | | |
| R6 | Half-year convention in the steady-state depreciation ratio | **Disclosed:** the ratio counts a full year of depreciation in the year spent while the schedule uses half; correcting it is +$0.41 (auditor) | | | +$0.41 |
| R6 | Offering expenses not deducted | **Disclosed:** the 8-K gives $22.62B at the underwriters' price and no expense figure (section 6) | | | under $0.05 |
| Owner | "0% on odds make it obvious that something is wrong" | **Answered** in section 3 and by the three-world odds | P 0.0% | P 0.17% | |

## 9. Bear case, and why the other side could be right

**Our bear ($2).** Arm and AMD take server share and a CPU price war starts in
2029; PCs shrink; no outside 14A customer, so 14A is paused and half the
construction in progress is written off; Intel keeps spending about its
depreciation to stay on 18A. Revenue is $59B in 2031, operating margin about
zero. Against $50.5B of debt and $15.6B of partner claims, the equity is worth
little beyond the offering's cash. The government's stake makes a rescue
more likely than a wipe-out in this world (section 3).

**Why the market could be right.** The CPU shortage is real: Intel's CEO said
it can meet only about half of customer CPU demand (reported by the Globe and
Mail, 2026-09-27), server ASPs rose 48% in a year, and the agent-driven shift
in the CPU-to-GPU ratio is the kind of demand break our fade may understate.
If a hyperscaler or Apple signs a multi-billion 14A commitment in 2027, the
success world becomes the base case and a peer multiple follows: that is
$85 on our numbers, more if the excess returns last. Our 5% rests on the
record of how fast foundries grow; one named customer would move it more
than any input in the base world.

## 10. Dates to watch, and what would prove us wrong

- **2026-10-29**, Q3 2026 results (earnings calendar, verified): revenue
  against the $15.8-16.8B guide (ours $16.6B), GAAP gross margin against the
  41% guide (ours 41.0%), 2027 capex.
- **2027-01-21** (tentative), Q4 2026: foundry loss path to the end-2027
  breakeven target; any named 14A customer.
- **2028**, the 14A high-volume commitment year (Q2 call).

We are wrong, and should move toward the price, if a named 14A or 18A customer
commits several billion dollars a year by 2028 (that moves weight to the
success world); if Intel's server revenue passes $35B in 2028 (central
$31.5B, bull $35.6B); or if a full year of GAAP gross margin above 47% arrives
by 2028 with capex above $27B. Toward the bear: server ASPs falling year on
year in any 2027 quarter, or outside foundry revenue below $2B in 2028.

## 11. Sources

Filings (EDGAR, masked before the blind commit): 10-K fiscal 2025; 10-Qs Q1
and Q2 2026; earnings releases Q2 2024 to Q2 2026; 8-Ks of 2026-02-27 and
2026-04-08; the offering 8-K of 2026-08-12 with its exhibits, read in full
after the commit; the 10-K's RSU and PSU table, saved after the review; 26 USC
48D, TrendForce foundry rankings (2021-08-31, 2026-09-09), eeNews Europe and IEEE
Spectrum on GlobalFoundries (2018), TechSpot and TrendForce News on 14A
(`data/foundry-base-rate-and-48d-sources-2026-10-16.txt`). All in `data/`. Earnings calls: Q1 2026 (2026-04-23), Q2 2026 (2026-07-23), Deutsche Bank
conference (August 2026). Market data statements, segment revenue, peer ratios and
analyst estimates (fetched 2026-10-15). Market quote (read
2026-10-16T03:26Z) and earnings calendar. Street forecasts (read
2026-10-16). Bigdata: two searches, 2026-10-16. Alpha, Capfolio's stock-ranking model (read-only, as of
2026-10-15): rank 20 of 713, score 97.3; A on expectations, growth and
momentum, C on quality, D on valuation.

## 12. Tooling problems met while writing this report

- Our tool that masks market information from filings leaves in the masked
  text a fair-value liability on the company's own escrowed shares, and the per-share price in an offering 8-K's
  Item 8.01.
- Capfolio's shared valuation method refuses a case whose net reinvestment is negative, so a
  harvest bear cannot be valued; the bear's capex was held at about its
  depreciation (assumptions.md, section 5), and the foundry capex tie only
  adds capex, never removes it, so low foundry draws do not hit that refusal.
- `alpha research odds` refuses to run before report.md and record.json exist,
  and refuses when the record's fair value differs from the last odds.json,
  so the old odds.json had to be moved aside before rerunning.
- The odds drivers can only override the central case, so the success world
  needed an `operations` switch driver to reach the bull's non-driver paths.
- research-check still asks for the "P(price justified)" label while the
  report template now says "Odds in our scenarios"; the block carries both.
- The coding environment that wrote this report refused inline `python -c`
  inside compound shell commands; scripts went to `.tmp` files instead.
- The file-writing tool refused a file named report.md for a helper agent; the
  report was written to `.tmp` and copied into place.