Intel CEO Lip-Bu Tan has discussed possible collaboration with Lisa Su, CEO of rival AMD. Tom's Hardware reported the news on October 26, 2026. Tan raised the subject during a CASPA conversation held on the 24th. However, no specific areas of collaboration, contract terms or amounts have been disclosed. The two companies are already working together on shared AI software for x86 CPUs, so cooperation between the rivals has in fact begun.
To understand Tan's remarks, it helps to distinguish among the three areas Intel is pursuing: custom chip design for customers, advanced packaging and its foundry business. Each has a different ordering process and revenue structure, and a collaboration with AMD does not necessarily mean AMD would outsource chip manufacturing to Intel. Intel Foundry's external revenue is also growing, but accounting factors lie behind part of that growth. To judge Tan's customer-focused turnaround, it is necessary to look beyond the content of any collaboration to actual orders, production plans and the breakdown of revenue.
What Is Known So Far About the AMD Collaboration
At the CASPA event, Tan said he had discussed areas in which Intel could work with Su, but he did not go into specifics. The official CASPA agenda confirms that a conversation between Tan and CASPA chairman Jimmy Cheng was scheduled for the evening of October 24, but it gives no details about collaboration. At present, there is nothing to suggest that AMD plans to place chip manufacturing or advanced packaging orders with Intel.
The two companies do, however, have a formally announced collaboration with concrete activity. According to Intel's official announcement, they are consolidating x86 CPU optimizations for the PyTorch AI development framework into a shared software library, oneDNN. AMD is integrating technology developed in its own optimization library, ZenDNN, into oneDNN, and it also participates as a Steering Member in the UXL Foundation, which promotes open AI software foundations. This is software cooperation aimed at making AI computation more efficient on Intel and AMD CPUs, and it is a different kind of effort from outsourcing chip manufacturing.
The effort also includes future support for AI Compute Extensions. Not all planned features have been implemented, though, and the performance gains have not been disclosed. It is also unclear whether Tan's conversation with Su referred to this software collaboration. Even so, the fact that the two companies are developing shared technology while remaining competitors is an important precedent when considering the potential for broader collaboration.
Custom Chip Design and Advanced Packaging Expand Order Opportunities
This is not the first time Intel has put customers at the center of its strategy. In Intel's official announcement about Intel Foundry on April 29, 2025, Tan stated plainly that listening to customers is the top priority. Two efforts tie that policy to concrete business: design services tailored to customers' applications, and stronger advanced packaging technology that combines multiple chips.
According to the second-quarter 2026 earnings presentation of July 23, 2026, revenue from Intel's Design Services business roughly tripled from a year earlier. Tan explained that the business is expanding from networking into computing and will eventually extend to accelerators. One example he cited is the joint development of a security processor with Fortinet. By taking on the design of ASICs (application-specific integrated circuits) tailored to customers' needs, Intel is trying to open revenue opportunities different from its traditional in-house CPUs. However, Design Services revenue itself has not been disclosed, so the roughly threefold growth rate alone does not show the size of the business.
Preparations to win more customer orders are also under way in advanced packaging. In the same earnings presentation, Intel said its EMIB-T backlog has grown and that yield and reliability have reached target levels. It is also building out volume production capability to support production ramps of customer products planned for 2027. Its 2025 announcement introduced 2.5D packaging and 3D stacking technologies, including EMIB-T and Foveros-R/B, and indicated a cooperative relationship with Amkor.
Still, designing chips for customers, assembling multiple chips into advanced packages and manufacturing silicon wafers are distinct businesses. Winning customers in these areas does not automatically translate into more orders for leading-edge manufacturing processes. Nor is there an official announcement linking EMIB-T production preparations to any AMD project. Judging the results of the customer-focused strategy requires looking separately at order status in each business and at when customer products actually enter production.
Intel 14A Design Readiness and Volume Production Are Separate Stages
For Intel's leading-edge 14A process, too, preparing the design environment must be distinguished from starting volume production. Under the plan Intel presented on July 23, 2026, version 0.5 of the 14A PDK (Process Design Kit) was complete, and version 0.9 was due in October. A PDK is a development kit that bundles the information and models chip designers need to design and verify circuits for a given manufacturing process. It is essential for customers to move ahead with product designs on a new process technology, but a PDK update does not mean volume production has begun at a fab.
In the production plan Intel presented at the same time, risk production for its own products is to begin in the second half of 2027, with volume production ramping in 2028. Risk production is a stage in which process maturity and yield are verified ahead of volume production, and it differs from stable commercial volume production. These are only plans as of July, and the actual delivery date of PDK 0.9, which was scheduled for October, could not be confirmed. There is also no basis for concluding that the remarks about collaborating with AMD have moved up the 14A production schedule.
More important is that Intel is pacing its capital investment in 14A to actual demand. In its Form 10-Q for the second quarter of 2026, the company says it is continuing work to complete 14A development, while deciding the scale and pace of manufacturing capacity expansion based on committed demand. That demand includes both its own products and demand that would arise if a potentially large external customer adopts 14A. In other words, technical readiness to enable designs on 14A and winning enough orders to justify factory expansion are separate challenges.
It cannot be ruled out that collaboration with AMD could eventually lead to 14A orders, but no specific manufacturing contract has been confirmed at this stage. For Intel's foundry strategy, what matters more than the possibility of collaboration itself is whether customers actually adopt 14A and whether firm demand emerges to support expanded manufacturing capacity.
Altera's Deconsolidation Affects the Growth in Foundry External Revenue
In evaluating Intel's customer acquisition strategy, the composition of foundry revenue also deserves attention. In the second quarter of 2026, Intel Foundry's external revenue was $293 million, equal to about 5.1% of segment revenue of $5.765 billion. The figures below are based on the segment information and management's discussion and analysis (MD&A) in Intel's Form 10-Q.
| Item | Q2 2026 | Scope |
|---|---|---|
| External revenue | $293 million | Includes wafer manufacturing, assembly and test |
| Intel Foundry segment revenue | $5.765 billion | Total segment revenue, including intragroup transactions |
| External revenue ratio | About 5.1% | External revenue ÷ segment revenue |
The period covered is the three months ended June 27, 2026. Expressed in millions of dollars, external revenue and segment revenue are 293 and 5,765, and the ratio calculated as 293 ÷ 5,765 × 100 is about 5.1%. This is the share of Intel Foundry's segment revenue that comes from external customers. It is not Intel's share of the foundry market, nor its ratio to Intel's total consolidated revenue.
External revenue rose $271 million from a year earlier, but this growth cannot be read directly as a result of winning new customers. In the Form 10-Q, Intel cites the deconsolidation of Altera, carried out in the third quarter of 2025, as the main factor in the increase. Sales to Altera that were previously treated as intragroup transactions are now classified as sales to an external customer as a result of the deconsolidation. The increase in external revenue therefore does not directly represent growth in new orders, and the underlying growth rate excluding the Altera effect cannot be calculated from these figures alone.
External revenue also includes income from assembly and testing, not just wafer manufacturing. The fact that external revenue increased therefore does not reveal whether manufacturing orders for leading-edge processes such as 14A grew or whether advanced packaging projects increased. Nor can it be assumed that orders from AMD have already materialized as a result of these remarks. The figure of about 5.1% shows only the share of external customer revenue within the entire Intel Foundry business, which includes manufacturing for internal use.
Whether Tan's customer-focused turnaround translates into actual results will become clear through future orders and progress in production. Particularly important are securing the committed demand that supports 14A capital investment, the production ramps of advanced packaging customers heading into 2027, and external revenue growth separated from the effect of Altera's deconsolidation. If the customer relationships Intel has built in design services and advanced packaging lead to continuing manufacturing orders, it could reduce its dependence on its own products and support its manufacturing base with external demand as well. The possible collaboration with AMD draws attention in that context, but for now it should be evaluated separately from concrete business results.