INTC | Earnings Review — Q2 2026

Intel's Q2 2026 print is a decisive, broadening acceleration, not a one-line beat. Net revenue of $16,128 million grew +25.4% YoY — the fastest print in the entire 10-quarter window and, per management, the "strongest revenue growth in more than 15 years" — and beat Street's $14.435B by +11.7% (+$1.693B), the largest revenue surprise of the last 12 quarters. Non-GAAP diluted EPS of $0.42 doubled the $0.21 Street estimate and marks the 7th consecutive quarter Intel has exceeded its own guidance (management's own framing, verified: the streak count increments by exactly one every quarter back to Q4'24). DCAI (Data Center & AI) revenue of $6,262M grew +59.0% YoY, the single largest swing factor behind the breakout, while Intel Foundry crossed +30.5% YoY for the first time in this series. GAAP diluted EPS of $(2.16) is not an operating signal — it reflects a ~$12.5B non-cash mark-to-market loss on CHIPS Act/Secure Enclave escrowed shares, unrelated to the operating beat.

Guidance: Q3 2026 revenue is guided to $15.8B–$16.8B (midpoint $16.3B, non-GAAP EPS $0.38) — a guide that implies +19.4% YoY growth, a deceleration from Q2's +25.4%. That is math, not a demand signal: Q3 2025 was already the strongest 2025 comp, whereas Q2 2025 was the weakest. FY2026 capex was raised for the second time this year, to >$20B (from an implied ~$17–$18B outlook, +~$3B), while PC-unit and server-unit guidance were both left unchanged despite DCAI's blowout quarter — management's own tell that the binding constraint is supply, not demand.

Tone: the call shifted from Q1's credibility-rebuilding framing ("a year ago the conversation was about whether we could survive") to a scale-execution and supply-severity framing ("one of the most severe supply constraints in [semiconductor] industry history"). Capital-allocation confidence rose materially — capex raised again, plus a new (unquantified) pre-announcement that FY2027 capex will be "significantly above" FY2026 — even as the risk register shifted almost entirely from demand durability (Q1) to supply severity spanning wafers, substrates, T-glass, and now memory (underscored by hiring SK Hynix's former CEO).

Contradictions (3 found, transcript-only — see below): (1) 2026 capex magnitude and phasing both reversed between the Q4'25 and Q1'26 calls before being raised again in Q2'26; (2) the Q2'26 call's own "$1.2B prior disclosure" baseline for ASIC run-rate doesn't match what Q4'25/Q1'26 actually disclosed (~$1B / "north of $1B"); (3) the Q3'25 call cites two irreconcilable 2025 capex totals ($18B and $27B) in one sentence, and $27B fails to reconcile against the reported FY2025 actual ($17.7B).

Catalysts into Q3'26 (reports Oct 22, 2026): AMD reports first (Aug 4, 2026) and is the earliest read on server demand; 14A PDK 0.9 is due October 2026, the cleanest test of management's guidance-accuracy track record; the preliminary Apple foundry engagement and Google/18A cloud commitment are the largest management-vs-Street expectation gaps; and the memory shortage (DRAM/NAND pricing shock) is simultaneously a client-margin risk and a strategic-priority signal.

*CCPG (formerly CCG, renamed "Client Computing & Physical AI Group" this quarter, presented like-for-like) growth is largely ASP/mix, not units — Zinsner conceded to Rasgon (Bernstein) that the underlying PC unit market is down YoY post-Windows refresh; a ~$173M stranded-inventory charge sits inside the print. Non-GAAP EPS YoY% is omitted above where the prior-year base is near-zero or sign-flipped (Q1'25–Q2'26 all involve at least one such quarter) since the percentage math becomes economically meaningless — the dollar trend (loss → $0.42 record) is the cleaner read.

Verdict — accelerating, on every line. Revenue growth troughed at -7.4% YoY (Q4'24), wobbled through a second false bottom of -4.1% (Q4'25), then inflected to +7.2% (Q1'26) and +25.4% (Q2'26, series high). DCAI (+59.0%) is the lead driver and Foundry crossed +30% YoY for the first time; CCPG's +12.8% is the one line where growth quality (price/mix vs. units) still needs scrutiny.

Legend: 🟢🟢🟢 large beat · 🟢🟢 medium-large beat · 🟢 medium beat · 🟡 small beat · 🔴/🔴🔴 miss (magnitude). Starred quarters = headline EPS depressed by one-time restructuring/impairment charges excluded from guidance; ex-charge results beat the guide in both flagged quarters (Q3'24 ex-charge ≈ +$0.15 vs. a -$0.03 guide; Q2'25 ex-charge non-GAAP EPS was $0.10, ahead of guide).

| Window | Revenue beat rate | Non-GAAP EPS beat rate (headline) | EPS beat rate (ex one-time charges) | |---|---|---|---| | L12Q (Q3'23–Q2'26) | 10/12 = 83% | 9/12 = 75% | 11/12 = 92% | | L4Q (Q3'25–Q2'26) | 4/4 = 100% | 4/4 = 100% | 4/4 = 100% |

| Rolling 4Q block | Avg revenue surprise | Avg EPS surprise (cents) | |---|---|---| | Q3'23–Q2'24 | +1.33% | +6.3¢ | | Q3'24–Q2'25 | +3.86% | -10.4¢ | | Q3'25–Q2'26 | +6.58% | +19.1¢ |

Pattern: CONSISTENT BEATER — and accelerating. Q2 2026 was the largest double beat of the cycle: revenue +11.7% vs. Street, non-GAAP EPS 2.0x Street. All three EPS misses in the trailing 12 quarters were charge-driven and all now sit outside the trailing four. Beat magnitude has stepped up on both lines, not just held — the average revenue surprise has roughly doubled block-over-block, and the last two quarters (+9.33%, then +11.73%) are the two largest revenue surprises in the window. Watch item: the bar into Q3 is now high — Street has reset to essentially the Q3 guide midpoint ($16.3B / $0.38), a materially harder comp than the $14.435B / $0.21 Intel just cleared.

(a) Retrospective bridge — the just-completed quarter (Q2 2026), calibrating confidence in the new Q3 guide:

(b) Forward bridge — the new Q3 2026 guide (no prior guide exists to bridge from, since Intel guides only one quarter ahead):

(c) FY2026 capex — the cleanest true prior→new company-guide bridge this call (no consensus line tracked):

Q3 2026 (FQ+1): new vs. prior vs. consensus

| Metric | Prior guide (issued Q1 call) | New guide low | New guide mid | New guide high | Consensus | vs. Consensus | |---|---|---|---|---|---|---| | Revenue | N/A — not pre-guided | $15.8B | $16.3B | $16.8B | ~$16.1–$16.4B (derived proxy) | Roughly in line | | Non-GAAP gross margin | N/A | 42.0% | 42.0% | 42.0% | Not tracked at quarterly granularity | N/A | | Non-GAAP EPS | N/A | $0.38 | $0.38 | $0.38 | ~$0.55–$0.58 implied for H2 (derived) | Guide is the only hard quarterly number |

YoY context (never QoQ): Q3 2026 guide midpoint $16.3B vs. Q3 2025 actual $13.653B implies +19.4% YoY — a deceleration from Q2 2026's +25.4% actual. Non-GAAP EPS guide of $0.38 vs. Q3 2025 actual $0.23 implies +65% YoY, also decelerating vs. Q1 2026's +123% actual. Net read: every guided YoY rate for Q3 decelerates off Q2's actual pace — normal math given Q2 2025 was the weakest 2025 comp — not, on its own, a demand red flag.

FY2026 (current fiscal year): new vs. prior vs. consensus

| Metric | Prior (Q1 2026 call) | New (Q2 2026 call) | vs. Prior | |---|---|---|---| | Revenue (Street consensus) | $58.7B (pre-print) | $60.43B (post-print) | +$1.73B / +2.9% | | EPS (Street consensus) | $1.088 (pre-print) | $1.272 (post-print) | +$0.184 / +16.9% | | Opex, non-GAAP | "directionally $16.0B, likely higher" (qualitative) | $16.5B (firm) | +~$0.5B / +3% — the "likely higher" caveat materialized | | Capex | "flat to last year" (~$18B) | >$20B, "+~$3B vs. prior outlook" | Raised for the 2nd time this year, +~15% | | Gross margin | Quarterly cadence only | "comfortably in the 40s" through 2026 | Directionally strengthened, not a numeric change | | PC unit TAM | Down low double digits | Down low double digits | Unchanged | | Server CPU unit growth | "Strong double-digit" | "Strong double-digit" | Unchanged — notably not raised despite DCAI printing +59% YoY |

Tone read: management's language on 18A/14A execution stepped up from "yields running ahead of internal projections" (Q1) to "increasingly confident that 14A will be highly competitive" (Q2). The risk register shifted almost entirely from demand durability (Q1) to supply severity (Q2) — a more comfortable problem for management to have, but a harder one for the Street to underwrite, since "supply-constrained forever" is difficult to distinguish from "guidance built with permanent headroom."

| Metric | Q3'24 | Q4'24 | Q1'25 | Q2'25 | Q3'25 | Q4'25 | Q1'26 | Q2'26 | |---|---|---|---|---|---|---|---|---| | Revenue YoY % | -6.2% | -7.4% | -0.4% | +0.2% | +2.8% | -4.1% | +7.2% | +25.4% | | Rev accel (bps QoQ) | -528 | -127 | +699 | +65 | +258 | -689 | +1,129 | +1,824 | | Non-GAAP EPS ($) | -0.46 | 0.13 | 0.13 | -0.10 | 0.23 | 0.15 | 0.29 | 0.42 |

Non-GAAP EPS YoY%/acceleration figures are omitted from this table (rather than shown misleadingly) where the prior-year base is near-zero or sign-flipped — which is true of every quarter in this window. The dollar trend below is the economically meaningful read.

xychart-beta
    title "INTC Revenue YoY % (Q3'24 - Q2'26)"
    x-axis [Q3'24, Q4'24, Q1'25, Q2'25, Q3'25, Q4'25, Q1'26, Q2'26]
    y-axis "YoY %" -10 --> 30
    line [-6.2, -7.4, -0.4, 0.2, 2.8, -4.1, 7.2, 25.4]

xychart-beta
    title "INTC Non-GAAP Diluted EPS $ (Q3'24 - Q2'26)"
    x-axis [Q3'24, Q4'24, Q1'25, Q2'25, Q3'25, Q4'25, Q1'26, Q2'26]
    y-axis "EPS $" -0.5 --> 0.5
    line [-0.46, 0.13, 0.13, -0.10, 0.23, 0.15, 0.29, 0.42]

Inflection points:

1. Trough — Q4'24 (-7.4% YoY): the worst print in the series; PC-market inventory digestion and a soft data-center compare bottomed the cycle.
2. First reacceleration leg — Q1'25→Q3'25: YoY climbs from -0.4% to +2.8% over three quarters (DCAI recovery, Foundry top-line growth) — the first sign the trough had passed.
3. Relapse — Q4'25 (-4.1% YoY, -689bps): growth backslides; the recovery is not yet secular at this point.
4. Breakout — Q1'26 (+7.2% YoY, +1,129bps): first quarter above +5% in the series; supply starts catching up to already-strong demand.
5. Sharper still — Q2'26 (+25.4% YoY, +1,824bps): the sharpest acceleration in the 8-quarter window by a wide margin — "strongest revenue growth in more than 15 years" per management, driven by DCAI +59% YoY.

On EPS: a real loss quarter (Q3'24, -$0.46), a second relapse into loss (Q2'25, -$0.10), then four straight quarters of sequential dollar improvement (0.23→0.15 dip→0.29→0.42), with Q2'26 the highest print in the window — more than double the $0.21 Street estimate.

Plain-English trajectory assessment: Intel's growth engine spent the back half of 2024 decelerating into contraction, then spent three quarters clawing back to flat-to-slightly-positive before a genuine relapse in Q4'25. What changed in 2026 is not gradual improvement but a step-function break: two consecutive quarters of double-digit-plus acceleration that took YoY growth from -4.1% to +25.4% in two prints — the sharpest two-quarter swing in the series. Both revenue and EPS now point the same direction after two years of chop — the open question is whether this is a durable inflection or a supply-catch-up sugar high.

Key drivers per management commentary: DCAI +59% YoY is framed explicitly as supply finally arriving against demand that was already there ("demand continues to outpace our growing supply"), not share recapture — the binding constraints are substrates, T-glass, and memory, not wafers. Capex raised twice on the same call (FY26 to >$20B, FY27 flagged "significantly above" that) signals management is underwriting the demand strength as durable enough to justify committing capital ahead of it. CCPG growth is largely ASP/mix, not units — the PC market itself was down YoY even as CCPG revenue rose 13%. Foundry external revenue remains de minimis ($293M of $5.8B segment revenue) — growing (+31% YoY) but still overwhelmingly an internal-volume story. The memory-supply hire (ex-SK Hynix CEO) is a tell that management sees memory, not wafer capacity, as the tightest near-term constraint on how far this acceleration can run into Q3/Q4 2026.

| # | Catalyst | Timing | Read-through | |---|---|---|---| | 1 | 14A PDK 0.9 release | October 2026 | The cleanest test of management's 7-straight-quarter guidance-accuracy track record — a specific date, repeated across two consecutive calls; a miss would be a credibility event. | | 2 | 14A external foundry customer commitment | H2 2026 into H1 2027 | Two prospective customers evaluating 14A test chips; a named win would be thesis-confirming, silence is already priced as "pending." | | 3 | Apple foundry engagement (preliminary, per Jun 2026 reports) | Ongoing, no confirmed timeline/node/scope | Barely in Street models — the largest management-vs-Street expectation gap on the foundry side; watch for detail on the Q3 call. | | 4 | Google/Xeon 18A cloud-foundry commitment | Announced pre-print; execution through H2 2026 | External foundry revenue $174M (Q1'26) → $293M (Q2'26), +68% QoQ — the closest real-time proof point; bar is that it keeps compounding. | | 5 | AMD Q3 2026 print (reports first, Aug 4, 2026) | ~11 weeks ahead of Intel's own Q3 print | First read on H2 server demand. AMD x86 server revenue share ~46.2% vs. Intel ~54.9% (down ~950bps YoY) — but AMD DC revenue +57% YoY is comparable to Intel DCAI +59% YoY, meaning Intel's reacceleration is not (yet) coming at AMD's direct expense. | | 6 | Memory supply constraint / DRAM-NAND shortage | Live now through 2027–2028 per industry trackers | Simultaneously a client-margin risk (already visible in the $173M stranded-inventory charge) and a strategic-priority signal (ex-SK Hynix CEO hire); PC OEMs have announced 15–20% price hikes industry-wide. | | 7 | 18A-P risk production → volume | Risk production entered Jun 16, 2026; complete by year-end | Lower-risk than 14A — a drop-in evolution of already-qualified 18A, not a new node; consensus bar is simply "on track." | | 8 | U.S. government 10% equity stake / CHIPS Act overhang | Ongoing, episodic headline risk | Stake marked ~$36–$40B vs. ~$8.9B invested — a real but non-fundamental sentiment/headline-risk catalyst, not a Street model input. | | 9 | Panther Lake (Core Ultra 300, 18A client) volume ramp | Already ramping through Q3/Q4 2026 | AI PC revenue +26% QoQ, ~2/3 of client revenue; Street's implicit bar is CCPG "roughly flat" in Q3 as edge growth offsets client softness. |

Reading the set: #1 (14A PDK 0.9) is the highest-conviction, easiest-to-mark-right-or-wrong near-term item. #3 (Apple) and #2 (14A customers) are the largest management-vs-Street gaps — the "management says it, Street hasn't priced it" pattern per house style, not a contrarian call already made. #5 (AMD) is the most negative consensus framing outstanding on INTC and is falsifiable against AMD's Aug 4 print, which lands before Intel's own.

Seven analysts asked two questions each (14 pairs); 8 well answered with specifics, 6 deflected or left materially incomplete — a higher deflection rate than typical, concentrated on capex-allocation specifics.

| Analyst | Topic | Verdict | |---|---|---| | Ben Reitzes (Melius) | Capex increase — which foundry customers, hard orders? | 🔴 Deflected — no customer names or order specifics given | | Ben Reitzes (Melius) | CPU TAM toward $220B by 2030? | 🔴 Deflected — declined to engage with the specific figure | | Joe Moore (Morgan Stanley) | Server share vs. AMD/Arm — is owning the fab an advantage? | 🟢 Well answered — "constraint is supply, not share," named roadmap (Clearwater Forest, Diamond Rapids, Coral Rapids) | | Joe Moore (Morgan Stanley) | Capex net vs. gross, internal/external split | 🟢 Well answered on net/gross mechanic; split left open | | Stacy Rasgon (Bernstein) | Is client strength pricing- or demand-driven? | 🟢 Well answered — "largely ASP," unit market down YoY volunteered | | Stacy Rasgon (Bernstein) | Q3/Q4 client outlook, inventory write-down detail | 🔴 Deflected — declined to itemize the $173M charge | | Timothy Arcuri (UBS) | Q3→Q4 undershipping dynamic | 🟢 Well answered — named bottlenecks (substrates, T-glass, memory) | | Timothy Arcuri (UBS) | Is the 40–60% GM drop-through rule still valid? | 🟢 Well answered — direct yes | | Vivek Arya (BofA) | External foundry customer timing, 14A capex split | 🔴 Deflected — both specific asks unanswered | | Vivek Arya (BofA) | Balance sheet capacity for elevated capex | 🟢 Well answered — ~$40B liquidity, multiple levers quantified | | CJ Muse (Cantor Fitzgerald) | Shape of server recovery — units or ASP? | 🟢 Well answered — both addressed, specific node named | | CJ Muse (Cantor Fitzgerald) | Capex framework vs. free cash flow | 🟢 Well answered — direct framework, FCF drag caveat given | | Aaron Rakers (Wells Fargo) | ASIC business diversity, growth/margin profile | 🔴 Deflected — run-rate given ($2B), margin/growth explicitly withheld | | Aaron Rakers (Wells Fargo) | Memory strategy, internal memory role | 🔴 Deflected — "further detail to be announced" |

Deflection pattern: six of fourteen questions clustered into three themes — (1) capex specificity (customer names, hard orders, internal/external split behind the >$20B raise), asked twice by different analysts; (2) new-business quantification (ASIC margin/growth rate, $220B CPU TAM); (3) inventory/memory detail (the $173M charge, memory-supply strategy). None of the deflections contradict prior disclosures — they read as management holding back forward-looking specifics rather than avoiding an uncomfortable topic, but the capex-allocation opacity is the most notable pattern given it ties directly to the >$20B/2026 and "significantly above 2026" 2027 capex commitments.

Intel's Q2 2026 call was almost entirely product-cycle and supply-constraint focused — no discussion of interest rates, inflation, or consumer/industrial confidence surveys was present in the transcript available this run.

Macro / end-market commentary:

• PC / consumer: market "down YoY post-Windows refresh," FY2026 PC market guided down low double digits. Client (CCPG) unit demand is soft/declining even as CCPG revenue grew +13% YoY — growth there is priced (ASP/mix), not volume-driven. A headwind read-through for PC OEMs generally, not just Intel.
• Server/enterprise: "strong double-digit unit growth" guided for FY2026; server growth framed as "well north of a double-digit CAGR" over the next few years. Reads as durable, broad-based hyperscaler infrastructure spend — supportive for the data-center capex complex generally.
• Industry-wide supply constraints: "one of the most severe supply constraints in [semiconductor] industry history," with bottlenecks named specifically as substrates, T-glass, and memory — not just leading-edge wafers. A sector-wide capacity signal supportive of substrate/glass-substrate/memory-supplier pricing power, and a constraint (not a demand problem) for every compute OEM/ODM sourcing the same inputs.

Companies mentioned:

| Company | Relationship | Read-through | |---|---|---| | AMD | Competitor (server CPU) | Mixed/net-neutral-to-cautious: Tan denies broad share loss ("constraint is supply, not share") but concedes Intel is "still catching up in some areas" — a soft positive for AMD in specific segments even as the overall server TAM expands for both. | | Arm | Competitor and partner | Continued architectural competition in server CPUs; Arm-based server compute remains a credible alternative Intel must roadmap against. | | SK Hynix | Supplier / talent source (memory) | Intel hired SK Hynix's former CEO to address memory supply and is collaborating with three major memory vendors — confirms memory as a binding, industry-wide bottleneck, supportive of memory-maker pricing power/utilization. | | Samsung / Micron (implied, unnamed) | Supplier (memory) | Same read-through as SK Hynix — supportive for memory-vendor bargaining power amid an acknowledged industry-wide shortage. | | SambaNova | Partner (AI compute) | Expanded partnership is a validation of SambaNova's AI inference/compute technology, giving it a larger distribution channel via Intel. | | Fortinet | Customer (custom silicon) | New security-ASIC collaboration validates Intel's foundry/ASIC design-services pitch to other prospective customers. | | Microsoft / Windows (implied) | Ecosystem / demand driver | "Down YoY post-Windows refresh" implies the prior-year Windows 11 hardware-refresh tailwind has largely played out — a modest headwind for PC OEMs (Dell, HP, Lenovo) reliant on that cycle. |

Data sourced from Daloopa (company_id 103), Intel's Q2 2026 earnings call transcript (2026-07-23), and FMP /stable/earnings and /stable/analyst-estimates for consensus. Data sourced from Daloopa.