Intel 3Q26: PCs Break the Guide; Maintain SELL at Lower TP

We maintain SELL and lower our FY2028 target price to USD65 from USD98. The revised target implies -8.3% downside from USD70.87, but the potential return is too small for a capital-intensive turnaround still lacking proof of external Foundry commercialisation. Server demand supports earnings, while a sharper PC downturn, uncertain fab utilisation and higher capex make Intel’s path to sustainable cash generation more difficult.                                                                                               

Figure 1: Intel’s share price

3Q26 Earnings Review: a data-centre beat, a client miss

Intel delivered its eighth consecutive beat, but the quarter was carried almost entirely by servers. Revenue grew +20.9% YoY to USD16.51bn, in line with the USD16.54bn consensus and +1.3% above the guidance midpoint. Non-GAAP gross margin rose +410bps YoY to 44.1%, 200bps above consensus, on better mix and lower Intel 3 and 18A costs, while non-GAAP EPS of USD0.52 beat the USD0.41 consensus by +26.8%. Operating leverage did most of the work, with non-GAAP operating income of USD3.23bn +22.5% above consensus as OPEX ran -6% below the guided run-rate.

DCAI revenue jumped +77.8% YoY to USD7.32bn, +9.4% above consensus, at a 41.5% operating margin. CCPG fell -5.1% YoY to USD8.10bn, -8.0% below consensus and short of a “flattish” guide, as memory-driven price increases hit PC demand. Intel Foundry revenue grew +58.1% YoY to USD6.70bn, but only USD411mn was external and the segment still lost -USD1.85bn. Adjusted free cash flow of USD1.64bn was -45.0% below consensus as gross capex rose +42.7% YoY.

The results reinforce our view that Intel’s product recovery is progressing faster than expected, but its manufacturing transformation has yet to deliver sustainable profitability.

Table 1: Intel’s 3Q26 earnings

Table 2: Intel’s 3Q26 segment revenue

Weak guidance triggered the sell-off

Despite a 26.8% EPS beat, Intel shares fell -23% after hours because the Q4 outlook challenged expectations of continued improvement. Revenue guidance of USD15.0–16.0bn missed the USD17.05bn consensus by 9.1% at the midpoint. Non-GAAP gross margin of 38.0% was 480bps below consensus, while EPS guidance of USD0.25 was 41.9% below expectations.

The Fab 28 accident in Israel is not the whole explanation. The estimated USD300mn revenue impact accounts for only about one-fifth of the USD1.55bn guidance shortfall, while its 200bps margin impact explains less than half the margin miss.

The larger problem is weaker PC demand. Management expects CCPG revenue to fall more than 20% QoQ in Q4 as customers reduce inventories. Lower client wafer starts also spread fixed factory costs across fewer units, while the 2027 capex step-up pushes positive free cash flow further away.

Our concern is whether stronger server earnings can offset weaker PCs while funding a commercially viable Foundry business.

Table 3: Intel’s management guidance

Progress without commercial proof

In July, we raised our target to USD98 while retaining SELL. We set three conditions for reconsidering HOLD: 1) a tier-1 external Foundry commitment, 2) sustained gross margin above 45%, and 3) positive free cash flow through the capex step-up. 3Q26 provides stronger evidence of operational progress, but none of the three has been met.

Table 4: Our HOLD conditions, 2Q26 vs 3Q26

Over the past three months, Intel raised USD23bn of equity, secured a multiyear Oracle Xeon agreement and delivered the 14A 0.9 process design kit on schedule. Against those advances, the PC outlook deteriorated and the higher diluted share count reduced earnings per share. We now expect a stronger DCAI mix, but a more difficult path to sustainable shareholder returns.

DCAI: strong AI demand does not guarantee market-share recovery

DCAI is now the strongest part of the investment case. Q4 guidance for mid-teens sequential growth implies around USD8.4bn of revenue, with an annualised exit run-rate near USD33.7bn. The Oracle agreement, rising purpose-built silicon revenue and new Xeon platforms support this momentum. We model DCAI growth of 37% in FY2027 as new Intel 3 capacity comes online.

Yet demand is not Intel-specific. Industry server CPU units are expected to grow more than 30% in both 2026 and 2027, while AMD continues to gain share. Mercury Research shows AMD reached 34.5% of x86 server units in 2Q26, up from 27.3% a year earlier. Intel remains constrained by Intel 3 output and advanced substrates, and management expects supply tightness into 1H27. New capacity can slow the pace of share loss, but it will not reverse it on its own. Part of the share loss was due to product competitiveness. AMD’s chips have offered more cores while using less power. Intel’s Diamond Rapids will match AMD’s 256 cores in 2027, but without multithreading it has half the threads of AMD’s Venice, which will be rolling out from late 2026. We expect DCAI earnings to improve, not necessarily market share.

Figure 2: AMD and Intel x86 server CPU unit market share

CCPG: a weakening PC market threatens margin recovery

The PC market is the main negative revision to our outlook. IDC estimates global 3Q26 PC shipments fell -20.1% YoY, particularly among large commercial PC vendors. Intel now expects PC units to decline -12% in 2026 and another -10% in 2027. Q4 CCPG guidance implies revenue below USD6.5bn, supporting our FY2027 forecast for a -12% segment decline.

Figure 3: Global PC shipments by vendor in 3Q26

Intel’s products have improved, but that has not stopped the share loss. AI PC revenue rose +35% YoY and Wildcat Lake units +70% QoQ, yet Intel’s client unit share slipped to 69.7% in 2Q26 from 75.9% in 1Q25. The weakness also matters beyond revenue, because fewer client wafers leave fabs underloaded and offset yield gains. We expect CCPG to remain a drag on consolidated growth and manufacturing efficiency through at least 1H27, until memory prices normalise.

Intel Foundry: process gains are not proof of commercial scale

Foundry execution is becoming more measurable. Intel 3 costs are down -15% and 18A costs -40% since 1Q26, and 18A products now make up nearly a third of client CPU revenue. The 14A 0.9 PDK was released on schedule in October, with the 1.0 PDK due in 1Q27.

Manufacturing progress should not be confused with commercial success. External Foundry revenue rose to USD411mn, but most of the USD118mn quarterly increase came from a non-recurring customer programme. No binding high-volume wafer commitment from a tier-1 external customer has been disclosed. Intel also provides no absolute node-level yield or utilisation data, while management acknowledged that 18A loading remains below its desired level. A USD1.85bn quarterly Foundry loss shows the distance to breakeven. We need committed volumes, competitive yields and improving utilisation before assigning significant Foundry option value.

Figure 4: Top 10 global foundries 2Q26 revenue ranking (USD Mn)

What the market may be overlooking

First, advanced packaging is progressing as we expected and may still be an earlier catalyst than 14A. In our 1Q26 note we called Intel’s packaging franchise “increasingly the underrated asset”, and in July we described it as “the most credible bridge between today’s product recovery and tomorrow’s external Foundry ambition”. Third-party evidence now supports that view. With TSMC’s CoWoS sold out into 2028, the market estimates EMIB-T costs about USD0.25 per mm², against USD0.39 for CoWoS-L and USD0.29 for TSMC’s CoPoS, which does not reach mass production until 2H28. The market also sees Google’s TPU, already an EMIB customer, alongside Nvidia’s NVLink switch and Apple’s Broadcom-designed AI server chip as potential 2028 adopters. The test is now conversion, as Intel expects to announce a couple of packaging customers within two quarters.

Second, mix is diluting margins for a good reason. Purpose-built silicon, which management expects to triple again over the next few years, carries below-average margins, so gross margin can stall while gross profit grows. Third, the guide cut sits mostly in client and legacy capacity. Fab 28 builds Intel 7 products rather than 18A or Intel 3, and excluding the fab the 4Q26 gross margin guide would be 40%.

Table 5: Advanced packaging cost comparison

Capex remains the obstacle to sustainable returns

Cash generation remains Intel’s biggest unresolved financial issue. Adjusted free cash flow was negative USD8.8bn for 9M26, including the Fab 34 buyout, and management now considers positive free cash flow in 2027 unlikely. Our model assumes FY2027–FY2028 gross capex of USD30bn and USD32bn, respectively. The USD23bn equity raise strengthens liquidity but also raises the diluted share count, with management guiding 5.44bn shares for Q4. Returns on the new fabs, not capex alone, will determine shareholder value.

Risk

1. Prolonged PC weakness – A memory-led PC downturn extending beyond 1H27 would depress CCPG revenue and reduce fab utilisation, potentially offsetting server-led profit growth.

2. Failure to secure external Foundry volume – Without a binding anchor-customer commitment, improved 18A and 14A technology may not translate into profitable loadings. Underutilisation could extend operating losses and weaken returns on deployed capital.

3. Capex, dilution and execution risk – Higher spending on manufacturing capacity, unexpected yield setbacks or additional financing could delay positive free cash flow and dilute per-share earnings. A severe supply disruption would compound these risks.

Valuation

We cut FY2027E EPS by -24% to USD1.77 from USD2.34 and FY2028E EPS by -21% to USD2.59 from USD3.26 in July as weaker CCPG demand and a higher share count outweigh part of the DCAI upgrade. The deeper FY2027 cut reflects the PC inventory correction and memory-driven demand weakness we expect to peak in 1H27. We forecast revenue of USD61.72bn, USD69.20bn and USD80.73bn for FY2026–FY2028, with gross margins recovering to 43% in FY2027 and 46% in FY2028. Those assumptions still require better utilisation and disciplined manufacturing costs.

We lower our fair P/E to 25x from 30x. EPS growth for FY2026–FY2028 has slowed to about 33% a year from 45% in July, and earnings are not yet converting into cash, with free cash flow negative in FY2027 as capex rises. Applying 25x to FY2028E EPS gives a target price of USD65, down from USD98, implying -8.3% downside from the post-results price of USD70.87.

Table 6: Valuation

Key takeaway

We maintain SELL with a lower target price of USD65, implying -8.3% downside. After the post earnings sell-off, the share price still does not adequately reflect uncertain external Foundry commitments, underutilised fabs and the funding needed to scale 14A.

We would need two things before changing our SELL call: 1) a binding, high-volume 14A wafer commitment in Intel’s own fabs, and 2) positive free cash flow through the capex cycle without further equity funding. We drop our previous 45% gross margin condition, as purpose-built silicon adds profit while diluting the margin, making free cash flow the better test.