Intel
The Turnaround, Still Unproven

Snapshot: Executive Summary

Intel violently re-rated through the first half of 2026, running roughly 4x off its ~$30–40 Q1 base to a June peak near $128¹ as a string of foundry catalysts, the Q1 FY26 beat with Foundry losses narrowing, the reported Google order for >3M custom AI chips (2028 delivery), Apple and Nvidia 18A/14A interest, 18A-P hitting risk production on schedule, Panther Lake ramping, and US-government equity backing, reframed Intel as the only credible US-based leading-edge alternative to TSMC⁴. The stock has since pulled back to $101.65, and the Q2 FY26 print (July 23) was a decisive operating beat, revenue $16.1B (+25% YoY) led by Data Center & AI at $6.3B (+59%). The re-rate rests on real numbers, not hype.

The debate is now about conviction, not price. After the pullback, our SOTP base target of $110 sits only ~8% above spot, in line with the Street consensus average (~$115)⁶, the Street turned constructive after Q2. We stay Hold, because the residual upside rests on the one thing Q2 did not settle: whether the reported Google/Apple/Nvidia interest converts from intent into committed, dollar-quantified external wafer volume. On ~37x EV/EBITDA and ~9x sales, with Foundry still losing ~$2.1B/quarter and GAAP deeply negative on a non-cash CHIPS mark, the modest upside is not high-conviction, it is a wide, execution-gated distribution (bull $135 / bear $65).

Three SOTP pieces, three valuation regimes:

• Intel Foundry (~$30B FY28E rev): the re-rate engine, valued on a forward EV/Sales multiple (~10.3x base) rather than a breakeven-margin multiple. We now apply a TSM-like sales multiple with the earlier scarcity premium removed, because the 2026 Google/Apple/Nvidia wins are still intent, not committed wafer volume. It is ~55% of total enterprise value, and the single piece the whole PT hinges on. Q2 FY26 operating loss narrowed to $(2.1)B (revenue ~$5.8B), tracking toward ~2027 breakeven.
• Intel Products (~$55B FY28E rev): Client Computing (Panther Lake on 18A) + Data Center & AI (Xeon; DCAI +59% YoY in Q2). The x86 cash engine, valued on EV/EBITDA (~32% margin × 14x base).
• Equity stakes (~$10B): Mobileye (Intel ~80% economic of MBLY) plus the retained 49% of Altera after selling 51% to Silver Lake. Valued at market, not on an operating multiple. A modest floor, small versus the foundry-driven EV.

The optical ~65x forward P/E (~93x on the nearer FY26E EPS) is meaningless, earnings are deeply depressed by Foundry burn, so the multiple is an artifact of trough EPS plus a capitalized foundry call option. Net debt is ~$12.2B (total debt $45.0B less cash $17.2B + ST investments $15.5B), far lighter than the ~$25B carried before the 2025 raises. We rate Intel Hold, 12-month PT $110 (bull $135 / bear $65): the upside to base is now modest and gated by binary Foundry execution and a still-deep GAAP loss.

Tactical: INTC is trading at $101.65, ~8% below our $110 12-month base target, roughly fair value and in line with the ~$115 Street average. We stay Hold: the modest gap to target is not a conviction call, it reflects a Foundry turnaround that is still unproven (losses ~$2.1B/quarter, GAAP deeply negative on a non-cash CHIPS mark). We value Foundry at a TSM-like ~10.3x EV/Sales, not a premium; the bull ($135) needs the reported wins to convert into committed wafer volume, the bear ($65) is a de-rate below TSM. Wide, two-sided, execution-gated. Rating Hold.

Investment Thesis

Rating: Hold. The probability-weighted SOTP (45% base / 30% bull / 25% bear) blends to a fair value of ~$106 (0.30×$135 + 0.45×$110 + 0.25×$65), only ~4% above the $101.65 spot and just below the $110 base PT. The skew is roughly two-sided (+33% bull / (36%) bear), and the residual upside is offset by a binary catalyst: the entire thesis turns on whether 2026's reported foundry wins convert from intent into committed wafer volume. With Foundry still losing ~$2.1B/quarter, GAAP deeply negative on a non-cash CHIPS mark, and the stock at ~37x EV/EBITDA, we will not pay up for a Foundry scarcity premium the turnaround has not earned, so neither a Buy (conviction not earned) nor a Sell (Q2 was a genuine beat, breakeven path intact) is warranted. Hold is the disciplined call.

Business Overview

Intel reports two operating segments, Intel Products (the x86 design business) and Intel Foundry (the manufacturing business), and carries two material public-company equity stakes (Mobileye and Altera). For valuation the useful frame is three SOTP buckets, because a profitable design franchise, a re-rated leading-edge fab, and a basket of marketable securities each deserve a different valuation regime. In Q2 FY26, the AI-driven Data Center & AI segment accelerated to +59% YoY.

Piece-by-piece commentary

• Intel Foundry: The manufacturing business, Q2 FY26 operating loss narrowed to $(2.1)B (revenue ~$5.8B), tracking toward management's ~2027 breakeven target. 18A (RibbonFET gate-all-around + PowerVia backside power) and the next-gen 18A-P and 14A nodes are the process-leadership bet. Reported external interest now includes Google (>3M TPUs, 2028), Apple, and Nvidia, with Microsoft and Amazon confirmed on 18A. It competes directly with TSMC and is valued on a TSM-like EV/Sales multiple.
• Intel Products (x86): the ~$15.5B cash engine, comprising Client Computing Group (~$8.9B, +13%; Core Ultra and Panther Lake, the first high-volume client part on 18A, across 200+ OEM AI-PC designs) and Data Center and AI ($6.3B, +59%; Xeon plus the AI-accelerator efforts that have not closed the gap to NVIDIA/AMD). Total Intel Products revenue ~$15.5B (+~20% YoY), operating income ~$4.8B in Q2 FY26.
• Mobileye (~80% economic): Intel retains a controlling economic interest (~80%) and ~98.6% voting via Class B in the publicly traded ADAS company MBLY (mkt cap ~$7.3B). Intel sold ~$1B of Class A shares in July 2025 and recognized non-cash impairment charges through 2026. Valued at market.
• Altera (49%): Intel sold 51% of its FPGA business to Silver Lake at an ~$8.75B valuation and retains 49% (~$4.3B). The retained stake is valued at market within the equity-stakes bucket.

Intel Foundry: 18A & External Customers

The single most important section in this report, and the source of the entire 2026 re-rating. Intel Foundry is the manufacturing business spun into its own reportable segment, and in 2026 it stopped being purely the reason the company looked broken and became the reason the stock ran ~4x into mid-2026. Q2 FY26 Foundry revenue was ~$5.8B (+31%) with the operating loss narrowing to $(2.1)B², visibly improving toward management's ~2027 breakeven target. The bull case rests on 18A / 18A-P / 14A reaching process leadership and converting reported external interest into committed, dollar-quantified wafer volume, the metric that would justify any scarcity premium above the TSM-like multiple we now apply.

What 18A / 18A-P / 14A actually are

18A is Intel's first node to combine two architectural firsts at high volume: RibbonFET (gate-all-around / nanosheet transistor) and PowerVia (backside power delivery). 18A-P: a performance-optimized variant, hit risk production on schedule with better perf/power than 18A. 14A is the next node; its PDK 1.0 is due this fall, and customer commitments gate further 14A investment (management's "no more blank checks" discipline). Panther Lake, a client CPU, is the first high-volume 18A product, validating the node on Intel's own designs before external customers commit at scale.

The external-customer question: intent vs committed volume

A merchant foundry only works at scale if external customers fab there at volume. The 2026 rally priced a wave of reported wins: a Google order for >3M custom AI chips (2028 delivery), an Apple US-build commitment (M-series on 18A-P rumored for 2027), and Nvidia evaluating 18A as a TSMC backup, with Microsoft and Amazon confirmed on 18A. The critical caveat, and the heart of the bear case, is that most of these are intent, not committed wafer volume. The bull view is that geopolitical supply-chain diversification plus a competitive 18A/14A converts them into binding orders. The bear view is that they shrink or slip and the foundry multiple de-rates hard, below TSM. This is the single binary the stock turns on through 2026–2027.

The capex backdrop

Products: x86 & the AI-GPU Gap

Intel Products is the franchise that pays the bills. It is the x86 design business, Core Ultra client CPUs and Xeon server CPUs, and in Q2 FY26 it generated revenue of ~$15.5B (+~20% YoY) and operating income of ~$4.8B. In the SOTP it is a steady ~$246B-EV piece, the dependable cash engine under the Foundry option, but it carries two structural wounds: ongoing x86 share loss to AMD, and a near-total absence from the AI-accelerator market that is reshaping data-center economics.

Client Computing: Core Ultra & Panther Lake

• Core Ultra / Panther Lake: the notebook/desktop CPU line. CCG revenue was ~$8.9B (+13% YoY) in Q2 FY26. Panther Lake is strategically critical because it is the first high-volume client part built on Intel's own 18A node, ramping across 200+ OEM AI-PC designs: validating the foundry's leading edge on Intel's own silicon.
• Client remains the largest revenue line in Products, but it is cyclical with the PC market and structurally pressured by AMD share gains and the emergence of ARM-based Windows laptops.

Data Center & AI: Xeon momentum, and the accelerator gap

• Data Center & AI: $6.3B (+59% YoY) in Q2 FY26, the standout growth line, accelerating on AI-server momentum and Xeon attach. Xeon still holds the majority of the server-CPU installed base, though AMD's EPYC has taken share for years.
• The AI-GPU gap. The defining strategic failure. Intel's Gaudi accelerators failed to gain meaningful share against NVIDIA and AMD; Falcon Shores was cancelled as a commercial product; and the rack-scale "Jaguar Shores" remains unproven. For now Intel is effectively a non-participant in the merchant AI-accelerator market, the single largest profit pool in semiconductors.

Network and Edge

• The networking and edge-compute portfolio, a smaller, steadier contributor that rounds out the Products franchise.

The risk lens

The base case for Products is not heroic: a steady x86 franchise that throws off cash, with DCAI providing genuine growth. The downside is an acceleration of share loss, AMD in both client and server, plus ARM-on-Windows, that turns slow erosion into structural decline and de-rates the Products multiple toward the bear 11x. The AI-GPU gap is largely a sunk strategic loss; we do not credit Jaguar Shores with material accelerator share in the base case. This piece's job is to be the cash floor; the foundry is the story.

Financial Health & Trends

The Q2 FY26 print: the blowout

Q2 FY26 (quarter ended June 27, reported July 23) was a decisive operating beat. Revenue was $16.1B (+25% YoY), roughly $1.7B ahead of the ~$14.4B consensus and the biggest upside surprise of the turnaround, led by Data Center & AI at $6.3B (+59% YoY). Non-GAAP gross margin was 41.8% and non-GAAP diluted EPS $0.42, doubling from $0.21 a year ago. The headline GAAP figure looks alarming, GAAP diluted EPS of $(2.16) on a ~$11B net loss, but it is almost entirely a non-cash mark: an impairment tied to the CHIPS/US-government equity arrangement, not a deterioration in the operating business. Lead with the operating story, revenue, margin, and non-GAAP EPS all beat, while the GAAP loss is a one-time non-cash charge.

The gross-margin story: off the trough

Intel's gross margin collapsed from ~60% in 2020 to the ~30–40% range in 2023–25³, driven primarily by Foundry under-utilization. The 2026 inflection is that it is finally turning up: Q2 FY26 non-GAAP GM of 41.8% extends the recovery (Q1 was 41.0%) and beat the ~39% guide, the second consecutive step off the trough. The path back to ~50%+ runs entirely through fab utilization and 18A/18A-P yield; the bull case is that margin was a cyclical trough, the bear that ~40% is closer to the new ceiling.

Q2 FY26 print highlights

Source: Intel Q2 FY26 8-K earnings release and 10-Q (quarter ended June 27, 2026). FY27 consensus non-GAAP EPS ~$1.56, still depressed by Foundry burn, which is why a forward-P/E lens is uninformative and the SOTP governs.

Capital Allocation & Returns

Intel's capital story in 2025 was about survival financing, not returns, and those raises are the reason the balance sheet entering the 2026 re-rate is far healthier than the headlines suggested. The dividend was suspended in 2024; there is no buyback. Net debt is now ~$12.2B (total debt $45.0B less cash $17.2B + short-term investments $15.5B; on a cash-only basis ~$27.8B). The 2025 capital injections diluted shareholders but secured the runway, and brought strategic validators (the US government, NVIDIA, SoftBank) onto the cap table.

The 2025 capital events

• US government ~10% equity stake (2025). CHIPS Act funding (~$8.9B related) was converted into a direct equity stake of roughly 10%, an unprecedented federal ownership position that signals strategic backing of a domestic leading-edge fab (a pillar of the 2026 re-rate) but also introduces governance complexity.
• NVIDIA $5B investment (Sept 2025). Paired with an x86 + RTX collaboration, a notable validation given NVIDIA is simultaneously Intel's chief AI-accelerator rival and a potential foundry customer.
• SoftBank $2B investment (2025). Additional strategic capital into the turnaround.
• Altera 51% sale to Silver Lake (2025). Monetized the majority of the FPGA business at an ~$8.75B valuation ($4.46B cash), leaving Intel with a 49% retained stake, an early example of crystallizing the parts.

Why the capital structure matters now

In the SOTP, Intel's net debt is a per-share subtraction, equity value equals total enterprise value minus the ~$12B net debt. At ~$12B it is a far smaller drag than the ~$25–28B feared a year ago, which is part of why even the bear case ($65) is well above the old ~$18–30 regime. The trade-offs: the entire return case rests on Foundry execution (no dividend, no buyback), and the dilution from the 2025 raises plus the US-gov/NVDA/SoftBank cap-table grew share count to ~5.04B and leaves a governance/dilution overhang.

Valuation Overview

Intel trades at ~65x forward earnings (~93x on the nearer FY26E EPS), but that number is an artifact, not a signal. Earnings are deeply depressed through the turnaround (Foundry losses crush consolidated EPS), so a high P/E on a trough number tells you nothing. The more telling reads are ~37x EV/EBITDA and ~9x sales: the inverse of Intel's 2025 cheapness. INTC is no longer the cheap value name, its multiple is now an artifact of trough EBITDA plus a capitalized foundry call option. A single earnings multiple on a business that is part profitable x86 design house, part re-rated leading-edge fab, and part basket of marketable stakes is the wrong frame. The right frame is SOTP.

Why the SOTP

A blended multiple always loses to a fundamentals-aware SOTP when the underlying mix is heterogeneous, and Intel's is as heterogeneous as it gets. We separate Intel into three pieces, Foundry, Products, and the Mobileye + Altera stakes, apply piece-appropriate valuations (a forward EV/Sales multiple on Foundry, EV/EBITDA on Products, market value on the stakes), sum the enterprise values, subtract net debt, and divide by shares. That work is in the next section, with an interactive scenario tool.

SOTP: Sum-of-the-Parts

The SOTP is the heart of the report. Intel Products is valued on FY28E revenue, EBITDA margin, and an EV/EBITDA multiple. Intel Foundry is valued on FY28E revenue and a forward EV/Sales multiple: we now apply a TSM-like ~10.3x base (TSM ~10–11x sales), the earlier 2026 scarcity premium removed while the turnaround is unproven, not a breakeven-margin EBITDA multiple. A modest premium returns only in the bull. The equity stakes (Mobileye ~80% + Altera 49%) are valued at market. Toggle Bull / Base / Bear / Reverse; drag the Products and Foundry multiples to see the implied PT update. Intel carries net debt, so the model subtracts ~$12B from enterprise value to reach equity.

Sensitivity grid: Products multiple × Foundry EV/Sales (PT in $)

PT calculator (alternative simple-multiple view)

A cross-check on out-year normalized earnings. Because trough/normalizing EPS is depressed, a spot P/E is misleading, this lens uses out-year normalized EPS (e.g. BofA frames ~26x on 2030E EPS ~$6.24 for its $160 target, above our base). Our $110 base ≈ ~$5.50 normalized EPS × ~20x; set normalized EPS and a normalized multiple to triangulate the SOTP.

Risk / Reward calculator

Note: The assistant reasons from the dashboard's data snapshot and thesis sections, it does not browse the web, hit Bloomberg, or access real-time fundamentals beyond what's in data.js. Treat its responses as scenario-modeling support, not as primary research. Author judgments on rating, PT, and probabilities remain with the analyst.

Upcoming Catalysts

Risk Factors

• Deal-vs-volume gap. The pre-eminent risk. The 2026 rally prices foundry wins that are largely reported/intent, not committed wafer volume. If the Google/Apple/Nvidia orders shrink or slip, the foundry multiple (~55% of the SOTP) de-rates below TSM, straight to the bear $65.
• Extreme valuation, thin margin for error. ~37x EV/EBITDA, ~9x sales, ~65x FY27E EPS (~93x FY26E) leaves little room for execution error; the ~$102 pullback restores only a modest cushion, and any 18A/18A-P/14A yield miss is punished given the multiple still capitalizes a successful turnaround.
• Foundry still burns cash. ~$2.1B/quarter operating loss; if breakeven (~2027) slips, the burn reignites against ~$12B net debt and could pressure the multiple and the balance sheet again.
• x86 share loss and the AI-accelerator gap. AMD continues to take client and server share; ARM-on-Windows and hyperscaler in-house silicon erode x86; Gaudi failed, Falcon Shores was cancelled, and Jaguar Shores is unproven, Intel remains a non-factor in the largest semis profit pool.
• GAAP still loss-making. Q2 FY26 GAAP EPS was $(2.16) on a ~$11B net loss, almost entirely a non-cash CHIPS/impairment mark; the reported business is not yet GAAP-profitable even as non-GAAP EPS doubled to $0.42.
• Governance / dilution overhang. The US-gov ~10% stake plus NVIDIA/SoftBank on the cap table, and a share count that rose to ~5.04B, leave dilution and governance complexity. The stock has already pulled back ~20% from the June highs.

Scenario Stress Tests

All stress-test PTs are derived from the same SOTP framework used in the interactive tool. The middle rows are single-lever shocks (only the input noted moves); the Full Bull and Full Bear rows apply the complete scenario (all of Products/Foundry revenue, margin/multiple, stakes, and net debt). The foundry EV/Sales multiple is the dominant lever, it is ~55% of enterprise value, so a single-notch change there moves the PT more than anything else. Note even the full bear ($65) sits well above the old ~$18–30 regime: the franchise has been structurally re-rated.

Bull vs Bear Debate

Technical Analysis

Trader's view

• Price (~$102) has pulled back below its 20/50-day averages but stays well above the 200-DMA (~$88); the gap to the 200-DMA still shows how far the 2026 re-rate carried the stock.
• Key support: the ~$44 area (the pre-gap Q1 2026 consolidation). A retreat there would imply the April gap is filling and the re-rate is unwinding.
• Key resistance: the ~$142 area (52-week high $142.35; ATH close ~$141 in June). A decisive close above unlocks the bull leg toward $135.
• Momentum: MACD rolling over; RSI has unwound from overbought, fitting a Hold that pulled back from the highs and now sits just below the $110 PT.

Glossary & Methodology Notes

18A / 18A-P / 14A (Intel nodes)

Intel's leading-edge process nodes. 18A is the first high-volume node combining RibbonFET and PowerVia; 18A-P is a performance-optimized variant that hit risk production on schedule; 14A is the next node, with PDK 1.0 due fall 2026 and committed customers gating further investment. Panther Lake is the first high-volume 18A product.

RibbonFET / PowerVia

RibbonFET is Intel's gate-all-around (nanosheet) transistor, replacing finFET for better performance-per-watt. PowerVia is backside power delivery, routing power on the back of the wafer to free the front for signal routing. Both debut at volume on 18A.

Intel Foundry

Intel's manufacturing business, reported as a separate segment. It serves Intel's own products (internal) and, increasingly, external customers. Q2 FY26 operating loss narrowed to $(2.1)B (revenue ~$5.8B), tracking toward ~2027 breakeven. In this SOTP it is valued on a TSM-like EV/Sales multiple (~10.3x base), the earlier scarcity premium removed while the turnaround is unproven.

SOTP (Sum-of-the-Parts)

A valuation method that values each business piece separately, sums the enterprise values, adjusts for net debt or net cash, and divides by shares. Used when a single blended multiple obscures the mix. Intel carries net debt, so the adjustment is a subtraction.

EV/Sales multiple (Foundry)

For Intel Foundry, a forward EV/Sales multiple stands in for a normal EBITDA multiple (Foundry still loses money). We anchor the base at a TSM-like ~10.3x (TSM trades ~10–11x sales) and removed the 2026 scarcity premium while the reported wins remain intent; a scarcity premium would return only if the Google/Apple/Nvidia 18A/14A wins convert to committed volume (the bull, ~10.7x).

EV/EBITDA

Enterprise value divided by EBITDA. A capital-structure-neutral measure, used here for the profitable Intel Products franchise. INTC's consolidated ~37x is rich because EBITDA is at a cyclical trough and a foundry option is capitalized into the price.

Gaudi / Falcon Shores / Jaguar Shores

Intel's AI-accelerator efforts. Gaudi failed to gain share versus NVIDIA/AMD; Falcon Shores was cancelled as a commercial product; "Jaguar Shores" is the rack-scale pivot, still unproven. Together they define the AI-GPU gap.

Mobileye / Altera (the stakes)

Mobileye (MBLY) is the publicly traded ADAS company in which Intel owns ~80% economic / ~98.6% voting (mkt cap ~$7.3B; Intel recognized non-cash impairment charges in 2026). Altera is the FPGA business: Intel sold 51% to Silver Lake (~$8.75B valuation) and retains 49%. Both are valued at market in the SOTP.

x86

The instruction-set architecture underpinning Intel's (and AMD's) Core and Xeon CPUs. Intel Products is fundamentally an x86 design franchise; the competitive threats are AMD share gains and ARM-based alternatives.

Methodology

• Snapshot anchor: August 7, 2026 close ($101.65). Live price patches via the Cloudflare-Worker quote proxy on page load.
• Fundamentals anchor: Q2 FY26 (quarter ended June 27, 2026; reported July 23, 2026). FY26E/FY27E/FY28E figures are estimates; Foundry/Products FY28E revenue and margin inputs are author triangulations calibrated to reconcile to the signed-off PTs, labeled as model anchors, not consensus line items.
• SOTP: Products on EV/EBITDA, Foundry on a forward EV/Sales multiple (a TSM-like base, a scarcity premium only in the bull), the equity stakes at market value, less net debt. The multiple and margin anchors are the author's view.
• Conclusions are the author's view. Illustrative, not investment advice.