Amkor Technology delivered a quarter in which the business improved in important ways but still fell short of what investors were paying for. Third-quarter earnings of $0.76 per share missed the $0.79 consensus by 7.3% and fell further below the $0.82 Earnings Whisper number. Revenue of $1.991 billion trailed the $2.02 billion consensus by 1.5%. The year-over-year comparison looks very different: earnings grew 49.0% while revenue grew just 0.2%. That is the central tension. Amkor is earning substantially more on essentially flat sales, which means the improvement is about mix and margin rather than volume. The fourth-quarter outlook, which came in below estimates, suggests the volume problem is not yet resolved.
The quality of the margin gain is real. Gross margin reached 18.3%, up from 16.8% in the second quarter and 14.3% a year ago. Operating margin was 11.3%, compared with 8.0% a year earlier. The cost breakdown shows where the gain came from. Materials fell to 50.7% of sales from 57.5% a year ago, a sign of a richer, less pass-through-heavy mix. Depreciation rose to 8.6% from 7.5%, so heavier investment is already weighing on the cost structure. The end-market shift is striking. Communications fell to 37% of sales from 51% a year ago, while computing rose to 26% from 19% and automotive and industrial rose to 22% from 16%. Advanced products revenue was actually lower year over year at $1.632 billion versus $1.684 billion. Top-ten customer concentration dropped to 66% from 73%, which is healthier, but it also reflects the shrinking smartphone franchise.
The miss is easier to understand against Amkor's own guidance. Gross margin landed below the 18.5% to 19.5% guide. Management attributed roughly 60 basis points to currency and the July Kumamoto earthquake. Computing rose 25% sequentially to a record but trailed the roughly 30% management had guided, because loading the 300-plus-step HDFO process took longer than expected. Consumer grew 10% against a mid-teens guide, and communications fell 8%. None of these shortfalls was large, but together they produced EPS just under the guidance midpoint and well below the whisper.
The forward-looking evidence is better than the quarter. Management now expects 2026 computing growth above 30%, up from above 20% at the start of the year, and said 2027 should be similar or higher. Fourth-quarter computing is guided up low double digits sequentially and more than 50% year over year. A hyperscaler HDFO CPU program began ramping, and two new 2.5D networking and data center programs come online in the fourth quarter. Advance-payment arrangements now cover a dozen customers, which analysts put at roughly $3 billion, with more under discussion. Management also framed packaging as an industry bottleneck with a constructive pricing environment.
Bears still have legitimate ammunition. The fourth-quarter guide of $1.9 billion to $2.0 billion in revenue and $0.65 to $0.75 in EPS implies a sequential step down. Communications is expected to fall high single digits and consumer mid-teens. New component shortages in premium smartphones and wearables add to memory-pricing pressure and an SIP migration headwind that extends into the first half of 2027. The capital burden is also escalating. The Arizona campus is now a roughly $12 billion, two-phase commitment, up from a $7 billion framework. Capex intensity is expected to exceed 30% of sales into 2028-29, and depreciation is set to rise further in 2027. Over nine months, $868 million of operating cash flow did not cover $1.24 billion of capex payments. Long-term debt rose to $2.32 billion from $1.28 billion at year-end, and total debt of $2.5 billion now slightly exceeds $2.4 billion in cash and investments.
The market had already turned skeptical before the report. Shares entered earnings at $39.30, down 24.1% from the open after the July report. That left them 36.8% below the August 17 inter-earnings high of $62.23 and just 2% above the low set on October 23. The stock sits 31.7% below its 200-day moving average. Investor sentiment moved from essentially neutral at 0.02 to modestly negative at -0.19, a meaningful but not extreme deterioration. In short, expectations were deflating, yet the whisper number still sat above consensus, and Amkor missed both.
The bottom line is that Amkor's quarter was a near-term disappointment wrapped around a strengthening long-term thesis. Margins, compute momentum, customer prepayments and capacity commitments all moved in the right direction. Execution slipped modestly, however, and the fourth-quarter guide shows that smartphone and consumer weakness can still overwhelm AI-driven growth at the revenue line. With the stock near its lows and sentiment soured, the debate now centers on whether the compute ramp can outrun a heavier capex and depreciation burden before investor patience runs thinner.