Intel: $23 Billion Capital Raise Signals Confidence in 14A Ramp and Data Center CPU Supercycle, CFO Says Gross Margins Now "Comfortably in the 40s"
Deutsche Bank 2026 Technology Conference, August 26, 2026
Intel's Chief Financial Officer David Zinsner used a fireside chat at the Deutsche Bank Technology Conference to explain the logic behind the company's recent $23 billion equity raise, framing it as a direct response to two converging signals: manufacturing execution that is finally tracking ahead of internal targets, and a data center CPU demand cycle that is proving more durable and more supply-constrained than the Street may appreciate. "The capital raise precedes capital investment," Zinsner said, noting that 2026 capital expenditure was already upticked from $18 billion to $20 billion, with a "significant" further increase planned for 2027, in part to secure substrate capacity that he described as severely constrained.
14A Commitment to High-Volume Manufacturing in 2028
The most consequential technical disclosure was Intel's formal commitment to high-volume manufacturing on its 14A process node in 2028, with the 0.9 version of the process design kit landing in October. Zinsner said defect density on 14A is tracking better than Intel's internal target curve and better than any prior node, a level of execution the company has not seen "since 22-nanometer, which is arguably one of the best nodes Intel has ever put out." Notably, he pointed to internal design teams — whom he called "probably the most cynical bunch out of anybody" — now choosing to design products on 14A as a key confidence signal, alongside what he described as a marked increase in external foundry customer engagement, with Intel CEO Lip-Bu Tan and his team now meeting weekly with prospective customers who have moved from evaluating data to negotiating capacity allocation. That shift toward capacity conversations is precisely what necessitated the capital raise, since Intel must begin placing supplier orders now and start risk production in 2027 to hit the 2028 ramp target.
Advanced Packaging Emerging as a Standalone, High-Margin Business
Zinsner offered unusually specific financial detail on Intel's advanced packaging business, particularly EMIB-T, which he framed as a differentiated, high-return complement to the foundry story. Revenue should begin ramping in the back half of 2027, become a "run rate business" in 2028, and hit its stride in 2029, scaling to multiple billions of dollars per customer per year. He guided to gross margins around 40% and operating margins near 30%, adding that because the business is not capital intensive relative to front-end manufacturing, "the ROIC is phenomenal." As a signal of internal conviction, Zinsner noted that Lip-Bu Tan recruited Seok-Hee Lee, the former CEO of SK Hynix, to run the unit: "You don't get a CEO caliber person to lead one of these businesses unless that particular individual sees the growth opportunity in front of us." He also described packaging as a strategic wedge into front-end foundry relationships, since customers who test Intel's operational execution in packaging are increasingly willing to cross over into front-end engagements.
Data Center CPU Demand Outpacing Supply Through at Least 2028
Zinsner laid out a structural argument for why CPU demand in data centers is inflecting upward, driven by the shift from training to inference to agentic workloads, which he said requires four to six times the CPU intensity of a standard training data center. Unlike prior cycles in which unit declines were offset only by core count growth and falling ASP per core (at times down nearly 20%), Zinsner said ASP per core has now stabilized and in some cases is rising, meaning "all the core count growth you get shows up in ASP growth" while unit volumes are also growing double digits. He was candid that the constraint is supply, not competitiveness: "This is going to be less about competitive performance of CPUs and more about can you get the CPUs to the customer." Intel expects to undersupply the market this year and next, with Ireland's Granite Rapids capacity targeted to more than double in output next year, and a return to balanced supply not expected until 2028.
Fab Footprint Strategy Shifting Toward Centralized High-Volume Sites
Zinsner detailed a change in Intel's manufacturing playbook: Oregon will now serve purely as a pilot line for new nodes before volume production is consolidated into one or two centralized high-volume sites to maximize scale and cost efficiency, rather than the historical approach of ramping modestly in Oregon before porting elsewhere. Arizona's Fab 52 is running 18A today, with Fab 62 nearing readiness, while 14A volume production is being routed to Oregon once 18A vacates space there. Ohio remains earlier stage, with Zinsner acknowledging Intel is moving "as fast as we can" on Mod 1 of that site.
Client Business Deliberately Starved of Supply to Feed Data Center
Zinsner confirmed Intel made an early-2026 decision to shift wafer capacity away from client CPUs toward data center products, anticipating that rising memory prices would suppress PC demand. "The client business weakening was probably the best thing we could have given the data center demand because we needed as much to be freed up as possible," he said. Panther Lake was described as performing well competitively, while Arrow Lake in the high-end desktop segment was only an "okay product," with Nova Lake positioned as the more competitive answer next year. Memory-driven softness in client demand is expected to persist as a headwind into 2027 before normalizing.
Rule of 45 Replaces Simple Margin Targets
Perhaps the most notable internal governance disclosure was Intel's adoption of a "Rule of 45" framework, borrowed from Lip-Bu Tan's software background at Cadence, which requires each business unit's revenue growth rate plus operating margin to sum to 45. Zinsner said gross margin is now "comfortably in the 40s," ahead of an internal plan that had called for high-30s, but cautioned that structurally lower-margin, high-growth segments — foundry (front-end and packaging) and the ASIC business, both roughly 40% gross margin — will continue to weigh on the corporate blended rate even as absolute profitability improves. He said the Rule of 45 was a central slide in the capital raise materials and that Lip-Bu Tan has fully embraced it as an accountability tool across every business unit.
Foundry Breakeven Timeline Could Slip to Late 2028
Zinsner reiterated an internal target of foundry breakeven by the end of 2027 but added a caveat that increased customer wins could push that timeline out to as late as the end of 2028 due to onboarding and start-up costs. He characterized current foundry operating losses at roughly $2.5 billion per quarter and said the priority is steady quarter-over-quarter improvement rather than hitting an exact breakeven date. On the topic of an investor day, Zinsner said it will not happen this year and likely falls sometime in 2027, explicitly decoupling the event from any foundry customer announcements, which he said Lip-Bu Tan has ruled out entirely as inconsistent with foundry industry norms.